In [1]:
import numpy as np
import pandas as pd
import matplotlib.pyplot as plt
from sklearn.model_selection import train_test_split, cross_val_score
from sklearn.metrics import r2_score, mean_squared_error, mean_absolute_error, f1_score
from sklearn.preprocessing import StandardScaler, LabelEncoder
from sklearn.neural_network import MLPRegressor, MLPClassifier
Multilayer Perceptron in Torch via Skorch¶
Dataset¶
The dataset was generated from a deep learning model trained on the Obesity or CVD risk dataset. https://www.kaggle.com/datasets/aravindpcoder/obesity-or-cvd-risk-classifyregressorcluster
The data consist of the estimation of obesity levels in people from the countries of Mexico, Peru and Colombia, with ages between 14 and 61 and diverse eating habits and physical condition , data was collected using a web platform with a survey where anonymous users answered each question, then the information was processed obtaining 17 attributes and 2111 records. The attributes related with eating habits are: Frequent consumption of high caloric food (FAVC), Frequency of consumption of vegetables (FCVC), Number of main meals (NCP), Consumption of food between meals (CAEC), Consumption of water daily (CH20), and Consumption of alcohol (CALC). The attributes related with the physical condition are: Calories consumption monitoring (SCC), Physical activity frequency (FAF), Time using technology devices (TUE), Transportation used (MTRANS) variables obtained : Gender, Age, Height and Weight.
NObesity values are:
- Underweight Less than 18.5
- Normal 18.5 to 24.9
- Overweight 25.0 to 29.9
- Obesity I 30.0 to 34.9
- Obesity II 35.0 to 39.9
- Obesity III Higher than 40
In [2]:
df = pd.read_csv("./Data/playground-series-s4e2/train.csv")
df
Out[2]:
| id | Gender | Age | Height | Weight | family_history_with_overweight | FAVC | FCVC | NCP | CAEC | SMOKE | CH2O | SCC | FAF | TUE | CALC | MTRANS | NObeyesdad | |
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| 0 | 0 | Male | 24.443011 | 1.699998 | 81.669950 | yes | yes | 2.000000 | 2.983297 | Sometimes | no | 2.763573 | no | 0.000000 | 0.976473 | Sometimes | Public_Transportation | Overweight_Level_II |
| 1 | 1 | Female | 18.000000 | 1.560000 | 57.000000 | yes | yes | 2.000000 | 3.000000 | Frequently | no | 2.000000 | no | 1.000000 | 1.000000 | no | Automobile | Normal_Weight |
| 2 | 2 | Female | 18.000000 | 1.711460 | 50.165754 | yes | yes | 1.880534 | 1.411685 | Sometimes | no | 1.910378 | no | 0.866045 | 1.673584 | no | Public_Transportation | Insufficient_Weight |
| 3 | 3 | Female | 20.952737 | 1.710730 | 131.274851 | yes | yes | 3.000000 | 3.000000 | Sometimes | no | 1.674061 | no | 1.467863 | 0.780199 | Sometimes | Public_Transportation | Obesity_Type_III |
| 4 | 4 | Male | 31.641081 | 1.914186 | 93.798055 | yes | yes | 2.679664 | 1.971472 | Sometimes | no | 1.979848 | no | 1.967973 | 0.931721 | Sometimes | Public_Transportation | Overweight_Level_II |
| ... | ... | ... | ... | ... | ... | ... | ... | ... | ... | ... | ... | ... | ... | ... | ... | ... | ... | ... |
| 20753 | 20753 | Male | 25.137087 | 1.766626 | 114.187096 | yes | yes | 2.919584 | 3.000000 | Sometimes | no | 2.151809 | no | 1.330519 | 0.196680 | Sometimes | Public_Transportation | Obesity_Type_II |
| 20754 | 20754 | Male | 18.000000 | 1.710000 | 50.000000 | no | yes | 3.000000 | 4.000000 | Frequently | no | 1.000000 | no | 2.000000 | 1.000000 | Sometimes | Public_Transportation | Insufficient_Weight |
| 20755 | 20755 | Male | 20.101026 | 1.819557 | 105.580491 | yes | yes | 2.407817 | 3.000000 | Sometimes | no | 2.000000 | no | 1.158040 | 1.198439 | no | Public_Transportation | Obesity_Type_II |
| 20756 | 20756 | Male | 33.852953 | 1.700000 | 83.520113 | yes | yes | 2.671238 | 1.971472 | Sometimes | no | 2.144838 | no | 0.000000 | 0.973834 | no | Automobile | Overweight_Level_II |
| 20757 | 20757 | Male | 26.680376 | 1.816547 | 118.134898 | yes | yes | 3.000000 | 3.000000 | Sometimes | no | 2.003563 | no | 0.684487 | 0.713823 | Sometimes | Public_Transportation | Obesity_Type_II |
20758 rows × 18 columns
In [3]:
df.info()
<class 'pandas.core.frame.DataFrame'> RangeIndex: 20758 entries, 0 to 20757 Data columns (total 18 columns): # Column Non-Null Count Dtype --- ------ -------------- ----- 0 id 20758 non-null int64 1 Gender 20758 non-null object 2 Age 20758 non-null float64 3 Height 20758 non-null float64 4 Weight 20758 non-null float64 5 family_history_with_overweight 20758 non-null object 6 FAVC 20758 non-null object 7 FCVC 20758 non-null float64 8 NCP 20758 non-null float64 9 CAEC 20758 non-null object 10 SMOKE 20758 non-null object 11 CH2O 20758 non-null float64 12 SCC 20758 non-null object 13 FAF 20758 non-null float64 14 TUE 20758 non-null float64 15 CALC 20758 non-null object 16 MTRANS 20758 non-null object 17 NObeyesdad 20758 non-null object dtypes: float64(8), int64(1), object(9) memory usage: 2.9+ MB
In [4]:
df.nunique()
Out[4]:
id 20758 Gender 2 Age 1703 Height 1833 Weight 1979 family_history_with_overweight 2 FAVC 2 FCVC 934 NCP 689 CAEC 4 SMOKE 2 CH2O 1506 SCC 2 FAF 1360 TUE 1297 CALC 3 MTRANS 5 NObeyesdad 7 dtype: int64
In [5]:
# Split the data into features and target variable
target = "NObeyesdad"
y = df[target]
X = df.drop(columns=[target, "id"])
In [6]:
# Convert categorical variables to numeric using one-hot encoding
X[X.select_dtypes('object').columns] = X.select_dtypes('object').apply(pd.Categorical)
X = pd.get_dummies(X, drop_first=True).astype(np.float32).values
In [7]:
# Convert labels to numeric using LabelEncoder
le = LabelEncoder()
y = le.fit_transform(y)
np.unique(y), le.classes_
Out[7]:
(array([0, 1, 2, 3, 4, 5, 6]),
array(['Insufficient_Weight', 'Normal_Weight', 'Obesity_Type_I',
'Obesity_Type_II', 'Obesity_Type_III', 'Overweight_Level_I',
'Overweight_Level_II'], dtype=object)) In [8]:
# Split the data into training and testing sets
X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=0.2, random_state=42)
X_train.shape, X_test.shape, y_train.shape, y_test.shape
Out[8]:
((16606, 22), (4152, 22), (16606,), (4152,))
Dummy Classifier¶
In [9]:
from sklearn.dummy import DummyClassifier
model = DummyClassifier(strategy="most_frequent")
model.fit(X_train, y_train)
y_pred = model.predict(X_test)
f1_score(y_test, y_pred, average="weighted")
Out[9]:
0.06282803118308164
MLP with Sklearn¶
In [10]:
model = MLPClassifier(
hidden_layer_sizes=(32,),
solver="adam",
alpha=0.001,
max_iter=500,
random_state=42,
batch_size=2048,
shuffle=True,
# parameters for early stopping
early_stopping=True, # whether to use early stopping to terminate training when validation score is not improving
n_iter_no_change=50, # maximum number of epochs to not meet tol improvement
validation_fraction=0.1, # proportion of training data to set aside as validation set for early stopping
)
In [11]:
model.fit(X_train, y_train)
y_pred = model.predict(X_test)
f1_score(y_test, y_pred, average="weighted")
/Users/francesco/miniforge3/envs/timeseries_dl/lib/python3.12/site-packages/sklearn/neural_network/_multilayer_perceptron.py:691: ConvergenceWarning: Stochastic Optimizer: Maximum iterations (500) reached and the optimization hasn't converged yet. warnings.warn(
Out[11]:
0.8338739595965879
MLP with Skorch¶
Skorch is a lightweight library that wraps PyTorch with a scikit-learn style interface. It lets you train neural networks, such as an MLP, using familiar methods like fit, predict, and score, while still keeping the flexibility of PyTorch for defining custom models and training logic.
For this notebook, Skorch is useful because it makes PyTorch models easier to integrate into standard machine learning workflows, including preprocessing pipelines, cross-validation, and hyperparameter tuning.
In [12]:
import torch
from torch import nn
import torch.nn.functional as F
from skorch import NeuralNetClassifier
from skorch.callbacks import EpochScoring
In [ ]:
# Define the MLP architecture. cpu will always work, but if you have a GPU you can use it to speed up training.
DEVICE = 'mps'
DEVICE = 'cuda'
DEVICE = 'cpu'
Vanilla¶
In [14]:
class VanillaMLP(nn.Module):
def __init__(self, in_features, out_features):
super().__init__()
self.model = nn.Sequential(
nn.Linear(in_features, 32),
nn.ReLU(),
nn.Linear(32, out_features),
nn.Softmax(dim=-1),
)
def forward(self, X, **kwargs):
return self.model(X)
First attempt: no scaling, mostly default parameters¶
In [15]:
net = NeuralNetClassifier(
VanillaMLP,
module__in_features=X_train.shape[1],
module__out_features=len(le.classes_),
max_epochs=200,
lr=0.01,
batch_size=2048, # default is 128, but we can increase it for better performance on GPU
callbacks=[
EpochScoring(
scoring='accuracy',
name='train_acc',
on_train=True,
),
],
device=DEVICE,
)
net
Out[15]:
<class 'skorch.classifier.NeuralNetClassifier'>[uninitialized]( module=<class '__main__.VanillaMLP'>, module__in_features=22, module__out_features=7, )In a Jupyter environment, please rerun this cell to show the HTML representation or trust the notebook.
On GitHub, the HTML representation is unable to render, please try loading this page with nbviewer.org.
<class 'skorch.classifier.NeuralNetClassifier'>[uninitialized]( module=<class '__main__.VanillaMLP'>, module__in_features=22, module__out_features=7, )
In [16]:
net.fit(X_train, y_train)
epoch train_acc train_loss valid_acc valid_loss dur
------- ----------- ------------ ----------- ------------ ------
1 0.1431 4.2396 0.1942 1.9763 2.0555
2 0.1973 1.9649 0.1671 1.9494 0.0686
3 0.2054 1.9431 0.1942 1.9290 0.0739
4 0.2138 1.9208 0.2008 1.9087 0.0683
5 0.2298 1.9030 0.2026 1.8917 0.0672
6 0.2053 1.8856 0.2050 1.8748 0.0655
7 0.2090 1.8680 0.2083 1.8568 0.0655
8 0.2145 1.8494 0.2092 1.8382 0.0678
9 0.2190 1.8303 0.2137 1.8197 0.0664
10 0.2209 1.8122 0.2146 1.8025 0.1078
11 0.2238 1.7947 0.2143 1.7855 0.0720
12 0.2255 1.7773 0.2164 1.7695 0.0719
13 0.2291 1.7607 0.2191 1.7537 0.0699
14 0.2326 1.7447 0.2243 1.7387 0.1487
15 0.2389 1.7293 0.2255 1.7250 0.0663
16 0.2505 1.7153 0.2402 1.7111 0.0670
17 0.2585 1.7139 0.2065 1.8079 0.0781
18 0.2471 1.7763 0.2634 1.7099 0.0842
19 0.2617 1.7216 0.2080 1.7953 0.1189
20 0.2679 1.7290 0.2435 1.7071 0.0804
21 0.2778 1.7205 0.2306 1.7286 0.1005
22 0.2852 1.7119 0.2547 1.6916 0.0750
23 0.2906 1.7091 0.2267 1.7600 0.0698
24 0.2958 1.6928 0.2565 1.6741 0.0708
25 0.3019 1.6899 0.2321 1.7587 0.0750
26 0.3063 1.6846 0.2643 1.6624 0.0881
27 0.3107 1.6767 0.2357 1.7522 0.0780
28 0.3128 1.6776 0.2703 1.6529 0.0762
29 0.3146 1.6651 0.2508 1.7129 0.0720
30 0.3193 1.6690 0.2884 1.6481 0.0680
31 0.3214 1.6593 0.2559 1.7071 0.0670
32 0.3226 1.6600 0.2842 1.6359 0.0691
33 0.3219 1.6508 0.2610 1.6961 0.0682
34 0.3260 1.6517 0.2941 1.6289 0.0681
35 0.3245 1.6434 0.2601 1.6875 0.0680
36 0.3278 1.6427 0.3004 1.6246 0.0672
37 0.3266 1.6386 0.2628 1.6815 0.0676
38 0.3296 1.6336 0.3323 1.6148 0.0709
39 0.3379 1.6298 0.2631 1.6933 0.0707
40 0.3321 1.6306 0.3209 1.6096 0.0671
41 0.3376 1.6201 0.2613 1.6806 0.0672
42 0.3348 1.6230 0.3453 1.6023 0.0719
43 0.3424 1.6113 0.2667 1.6683 0.0690
44 0.3381 1.6122 0.3609 1.5889 0.0693
45 0.3469 1.6014 0.2691 1.6630 0.0689
46 0.3397 1.6026 0.3799 1.5755 0.0677
47 0.3528 1.5793 0.2818 1.6184 0.0680
48 0.3434 1.5876 0.3597 1.5757 0.0669
49 0.3463 1.5903 0.2845 1.6147 0.0767
50 0.3454 1.5915 0.4127 1.5700 0.0687
51 0.3604 1.5747 0.2706 1.6493 0.0703
52 0.3493 1.5808 0.3949 1.5595 0.1871
53 0.3583 1.5715 0.2763 1.6440 0.0699
54 0.3504 1.5798 0.4208 1.5583 0.0708
55 0.3649 1.5639 0.2818 1.6144 0.0706
56 0.3517 1.5745 0.3874 1.5487 0.0713
57 0.3630 1.5604 0.2893 1.6029 0.0696
58 0.3554 1.5698 0.3934 1.5432 0.0666
59 0.3659 1.5554 0.2932 1.5938 0.0675
60 0.3578 1.5632 0.4214 1.5401 0.0663
61 0.3735 1.5441 0.3016 1.5821 0.0681
62 0.3587 1.5582 0.3811 1.5386 0.0666
63 0.3692 1.5489 0.2926 1.5958 0.0683
64 0.3628 1.5449 0.4133 1.5283 0.0676
65 0.3733 1.5358 0.2917 1.6094 0.0662
66 0.3628 1.5487 0.4160 1.5231 0.0660
67 0.3772 1.5290 0.2998 1.5999 0.0679
68 0.3652 1.5377 0.3986 1.5209 0.0734
69 0.3704 1.5327 0.3049 1.6083 0.0765
70 0.3683 1.5372 0.4275 1.5139 0.0706
71 0.3755 1.5187 0.3137 1.5886 0.0789
72 0.3696 1.5309 0.4028 1.5089 0.0795
73 0.3723 1.5247 0.3116 1.5827 0.0855
74 0.3683 1.5204 0.4118 1.5052 0.0695
75 0.3718 1.5251 0.3260 1.5898 0.0668
76 0.3709 1.5160 0.4064 1.5073 0.0698
77 0.3560 1.5125 0.3733 1.5419 0.0680
78 0.3736 1.4965 0.3715 1.5282 0.0672
79 0.3699 1.4942 0.3874 1.5251 0.0683
80 0.3789 1.4944 0.4091 1.5076 0.0685
81 0.3739 1.4875 0.3173 1.5750 0.0679
82 0.3436 1.5610 0.3691 1.5125 0.0685
83 0.3408 1.5531 0.3028 1.5761 0.0677
84 0.3540 1.5189 0.3916 1.5033 0.0692
85 0.3585 1.5087 0.3570 1.4925 0.0681
86 0.3619 1.4924 0.4040 1.5049 0.0690
87 0.3628 1.4797 0.3793 1.5327 0.0676
88 0.3730 1.4957 0.4094 1.4914 0.0672
89 0.3662 1.4878 0.3299 1.5312 0.1449
90 0.3556 1.4996 0.4061 1.4848 0.0750
91 0.3506 1.5042 0.3251 1.5543 0.0709
92 0.3511 1.5045 0.4052 1.4879 0.0702
93 0.3557 1.4893 0.3591 1.4830 0.0688
94 0.3622 1.4777 0.4067 1.5045 0.0678
95 0.3727 1.4635 0.3835 1.5098 0.0751
96 0.3547 1.4810 0.4359 1.4642 0.0674
97 0.3621 1.5033 0.3140 1.5609 0.0673
98 0.3491 1.5037 0.4181 1.4820 0.0674
99 0.3523 1.4760 0.3805 1.4788 0.0671
100 0.3683 1.4684 0.4115 1.4821 0.0667
101 0.3633 1.4577 0.3766 1.5016 0.0693
102 0.3508 1.4764 0.4314 1.4680 0.0703
103 0.3676 1.4755 0.3528 1.4930 0.0705
104 0.3631 1.4784 0.4031 1.4530 0.0686
105 0.3519 1.4879 0.3082 1.4925 0.0676
106 0.3744 1.4633 0.4232 1.4875 0.0684
107 0.3799 1.4376 0.3931 1.4888 0.0673
108 0.3463 1.4716 0.4259 1.4515 0.0679
109 0.3515 1.4730 0.3260 1.5477 0.0979
110 0.3520 1.5001 0.4281 1.4354 0.0680
111 0.3658 1.4565 0.3928 1.4394 0.0675
112 0.3823 1.4440 0.4223 1.4697 0.0688
113 0.3840 1.4308 0.3856 1.4552 0.0671
114 0.3553 1.4711 0.4205 1.4194 0.0677
115 0.3683 1.4532 0.3338 1.5076 0.0673
116 0.3479 1.4780 0.4241 1.4466 0.0680
117 0.3795 1.4317 0.3986 1.4420 0.0775
118 0.3785 1.4313 0.4214 1.4564 0.0675
119 0.3739 1.4317 0.3967 1.4759 0.0677
120 0.3515 1.4548 0.4202 1.4207 0.0687
121 0.3675 1.4305 0.3606 1.4689 0.0677
122 0.3627 1.4608 0.4127 1.4196 0.0664
123 0.3664 1.4445 0.3245 1.5240 0.0682
124 0.3463 1.4743 0.4278 1.4568 0.0681
125 0.3930 1.4255 0.3688 1.4043 0.0682
126 0.3858 1.4151 0.4118 1.4293 0.0709
127 0.3921 1.3874 0.4070 1.4227 0.1473
128 0.3816 1.4383 0.4416 1.4056 0.0762
129 0.3666 1.4338 0.3254 1.5225 0.0709
130 0.3613 1.4656 0.4199 1.4222 0.0721
131 0.3701 1.4175 0.4422 1.4099 0.0697
132 0.3838 1.4098 0.4013 1.4480 0.0704
133 0.3836 1.3838 0.4067 1.4561 0.0702
134 0.3706 1.4279 0.4019 1.4377 0.0694
135 0.4073 1.3657 0.4211 1.3968 0.0694
136 0.3535 1.4681 0.4253 1.3934 0.0688
137 0.3692 1.4468 0.3070 1.5635 0.0686
138 0.3595 1.4641 0.4193 1.4196 0.0681
139 0.3820 1.3985 0.4467 1.3524 0.0720
140 0.3960 1.4016 0.3992 1.4253 0.1147
141 0.3877 1.3705 0.4184 1.4265 0.0798
142 0.4007 1.4042 0.4205 1.4029 0.0709
143 0.3803 1.4144 0.3766 1.4336 0.0710
144 0.3635 1.4435 0.4259 1.4053 0.0699
145 0.3846 1.4013 0.4250 1.3858 0.0698
146 0.3643 1.4081 0.4049 1.4263 0.0715
147 0.3971 1.3558 0.4175 1.3545 0.0697
148 0.3829 1.3855 0.4214 1.4599 0.0701
149 0.4026 1.3782 0.3958 1.4332 0.0703
150 0.3666 1.4325 0.4133 1.4050 0.0693
151 0.3711 1.4550 0.3636 1.4903 0.0684
152 0.3531 1.4747 0.4040 1.4488 0.0691
153 0.4007 1.3743 0.5138 1.3175 0.0673
154 0.4368 1.3496 0.4169 1.4139 0.0686
155 0.4406 1.3593 0.4567 1.3251 0.0679
156 0.4030 1.3586 0.3465 1.4339 0.0676
157 0.3594 1.4298 0.4479 1.3491 0.0685
158 0.3947 1.3666 0.4100 1.3895 0.0677
159 0.3549 1.4573 0.4082 1.3976 0.0742
160 0.3869 1.3672 0.4545 1.2977 0.0677
161 0.4071 1.3711 0.4115 1.3810 0.0681
162 0.4093 1.3234 0.3983 1.3566 0.0684
163 0.3878 1.4095 0.4437 1.4152 0.0682
164 0.3699 1.4731 0.3269 1.5608 0.1417
165 0.3557 1.4567 0.4365 1.4190 0.0679
166 0.4331 1.3133 0.5141 1.3028 0.0682
167 0.4583 1.3309 0.3940 1.4411 0.0680
168 0.4426 1.3636 0.4133 1.3834 0.0689
169 0.4035 1.3761 0.4410 1.3348 0.0677
170 0.3989 1.3963 0.4368 1.3513 0.0682
171 0.4155 1.3574 0.4235 1.4017 0.0679
172 0.3800 1.4160 0.4329 1.3330 0.0681
173 0.4284 1.3154 0.3844 1.4098 0.0674
174 0.3685 1.3952 0.4223 1.3664 0.0680
175 0.4013 1.3476 0.4401 1.3270 0.0678
176 0.3880 1.4197 0.4262 1.4163 0.0667
177 0.3829 1.4239 0.3365 1.5454 0.0689
178 0.3444 1.4796 0.4347 1.4069 0.0681
179 0.4289 1.3429 0.4711 1.2882 0.0742
180 0.4228 1.3481 0.3961 1.4344 0.0672
181 0.4070 1.3701 0.4217 1.3646 0.0679
182 0.4046 1.3305 0.4573 1.2697 0.0676
183 0.3986 1.3502 0.4118 1.3597 0.0680
184 0.4236 1.2872 0.3853 1.3680 0.0680
185 0.3840 1.3895 0.4401 1.3876 0.0672
186 0.3772 1.4543 0.3703 1.5194 0.0669
187 0.3777 1.4337 0.4329 1.3286 0.0687
188 0.4073 1.3253 0.4365 1.3026 0.0681
189 0.3964 1.3524 0.4136 1.3483 0.0749
190 0.4274 1.3052 0.3669 1.4848 0.0716
191 0.3843 1.4045 0.4539 1.3406 0.0673
192 0.4077 1.3443 0.4464 1.3189 0.0679
193 0.3796 1.3824 0.4290 1.3320 0.0677
194 0.4341 1.2930 0.4235 1.3508 0.0686
195 0.4136 1.3531 0.4389 1.3255 0.0682
196 0.4250 1.3295 0.3841 1.4258 0.0684
197 0.3721 1.4316 0.3964 1.3279 0.0678
198 0.3963 1.3590 0.3371 1.4509 0.0669
199 0.3720 1.4062 0.3904 1.4310 0.0734
200 0.4054 1.3172 0.4266 1.3753 0.0682
Out[16]:
<class 'skorch.classifier.NeuralNetClassifier'>[initialized](
module_=VanillaMLP(
(model): Sequential(
(0): Linear(in_features=22, out_features=32, bias=True)
(1): ReLU()
(2): Linear(in_features=32, out_features=7, bias=True)
(3): Softmax(dim=-1)
)
),
)In a Jupyter environment, please rerun this cell to show the HTML representation or trust the notebook. On GitHub, the HTML representation is unable to render, please try loading this page with nbviewer.org.
<class 'skorch.classifier.NeuralNetClassifier'>[initialized](
module_=VanillaMLP(
(model): Sequential(
(0): Linear(in_features=22, out_features=32, bias=True)
(1): ReLU()
(2): Linear(in_features=32, out_features=7, bias=True)
(3): Softmax(dim=-1)
)
),
)In [17]:
y_pred = net.predict(X_test)
f1_score(y_test, y_pred, average="weighted")
Out[17]:
0.3754543665724024
In [18]:
net.criterion, net.optimizer
Out[18]:
(torch.nn.modules.loss.NLLLoss, torch.optim.sgd.SGD)
In [19]:
history = net.history
plt.plot(history[:, 'train_loss'], label='train_loss')
plt.plot(history[:, 'valid_loss'], label='valid_loss')
plt.legend()
plt.xlabel('Epoch')
plt.ylabel('Loss')
plt.show()
plt.plot(history[:, 'train_acc'], label='train_acc')
plt.plot(history[:, 'valid_acc'], label='valid_acc')
plt.legend()
plt.xlabel('Epoch')
plt.ylabel('Accuracy')
plt.show()
In [20]:
from torchviz import make_dot
make_dot(
net.module_(torch.as_tensor(X[:1], dtype=torch.float32).to(net.device)),
params=dict(net.module_.named_parameters()),
show_attrs=False,
# show_saved=True,
)
Out[20]:
Adam¶
In [21]:
net = NeuralNetClassifier(
VanillaMLP,
module__in_features=X_train.shape[1],
module__out_features=len(le.classes_),
max_epochs=500,
device=DEVICE,
optimizer=torch.optim.Adam,
lr=1e-3,
callbacks=[
EpochScoring(
scoring='accuracy',
name='train_acc',
on_train=True,
lower_is_better=False
),
],
batch_size=2048,
iterator_train__shuffle=True,
)
net
Out[21]:
<class 'skorch.classifier.NeuralNetClassifier'>[uninitialized]( module=<class '__main__.VanillaMLP'>, module__in_features=22, module__out_features=7, )In a Jupyter environment, please rerun this cell to show the HTML representation or trust the notebook.
On GitHub, the HTML representation is unable to render, please try loading this page with nbviewer.org.
<class 'skorch.classifier.NeuralNetClassifier'>[uninitialized]( module=<class '__main__.VanillaMLP'>, module__in_features=22, module__out_features=7, )
In [22]:
# Training the network
net.fit(X_train, y_train)
epoch train_acc train_loss valid_acc valid_loss dur
------- ----------- ------------ ----------- ------------ ------
1 0.1055 2.8612 0.1620 2.2308 0.1646
2 0.1413 2.0745 0.2203 1.9541 0.0784
3 0.2041 1.9657 0.2008 1.9668 0.0792
4 0.2097 1.9434 0.2219 1.9088 0.0757
5 0.2224 1.8955 0.2375 1.8621 0.0764
6 0.2520 1.8455 0.2288 1.8194 0.0739
7 0.2314 1.8114 0.2673 1.7938 0.0741
8 0.2886 1.7870 0.2655 1.7698 0.0736
9 0.3005 1.7623 0.2971 1.7436 0.0748
10 0.2797 1.7372 0.2878 1.7211 0.0738
11 0.3106 1.7143 0.3359 1.6989 0.0745
12 0.3443 1.6926 0.3471 1.6771 0.0742
13 0.3371 1.6707 0.3441 1.6566 0.0754
14 0.3685 1.6499 0.3904 1.6358 0.0785
15 0.3984 1.6291 0.4091 1.6159 0.0857
16 0.4128 1.6089 0.4187 1.5962 0.0831
17 0.4253 1.5892 0.4259 1.5769 0.0790
18 0.4429 1.5699 0.4422 1.5578 0.0785
19 0.4446 1.5509 0.4554 1.5393 0.0834
20 0.4712 1.5323 0.4606 1.5211 0.1165
21 0.4771 1.5139 0.4678 1.5031 0.0779
22 0.4753 1.4960 0.4852 1.4856 0.0787
23 0.4962 1.4784 0.4934 1.4682 0.0777
24 0.4944 1.4614 0.4946 1.4515 0.0765
25 0.5078 1.4446 0.5099 1.4346 0.0760
26 0.5106 1.4285 0.5154 1.4181 0.0765
27 0.5105 1.4116 0.5187 1.4023 0.0741
28 0.5307 1.3957 0.5301 1.3861 0.0750
29 0.5341 1.3797 0.5424 1.3707 0.0819
30 0.5483 1.3639 0.5403 1.3549 0.0858
31 0.5482 1.3483 0.5506 1.3396 0.0825
32 0.5614 1.3331 0.5563 1.3247 0.0767
33 0.5574 1.3191 0.5656 1.3098 0.0784
34 0.5770 1.3039 0.5629 1.2952 0.1628
35 0.5711 1.2896 0.5713 1.2804 0.0749
36 0.5742 1.2749 0.5858 1.2663 0.0739
37 0.5869 1.2605 0.5692 1.2527 0.0746
38 0.5795 1.2471 0.5957 1.2385 0.0741
39 0.5934 1.2331 0.5954 1.2257 0.0750
40 0.5961 1.2194 0.5999 1.2121 0.0753
41 0.6001 1.2068 0.6039 1.1989 0.0737
42 0.6070 1.1940 0.5993 1.1854 0.0754
43 0.6043 1.1808 0.6147 1.1727 0.0751
44 0.6107 1.1681 0.6017 1.1588 0.0759
45 0.6092 1.1556 0.6207 1.1461 0.0731
46 0.6197 1.1425 0.6282 1.1337 0.0741
47 0.6303 1.1294 0.6315 1.1206 0.0746
48 0.6312 1.1166 0.6409 1.1072 0.0746
49 0.6402 1.1033 0.6463 1.0945 0.0753
50 0.6466 1.0905 0.6499 1.0815 0.0760
51 0.6454 1.0785 0.6635 1.0697 0.0745
52 0.6564 1.0664 0.6556 1.0565 0.0750
53 0.6561 1.0536 0.6613 1.0443 0.0743
54 0.6589 1.0423 0.6695 1.0333 0.0754
55 0.6592 1.0318 0.6710 1.0214 0.1047
56 0.6643 1.0196 0.6743 1.0102 0.0875
57 0.6713 1.0088 0.6710 1.0000 0.0818
58 0.6686 0.9980 0.6758 0.9890 0.0774
59 0.6733 0.9876 0.6806 0.9788 0.0762
60 0.6740 0.9778 0.6875 0.9686 0.0765
61 0.6796 0.9680 0.6869 0.9596 0.0744
62 0.6793 0.9582 0.6857 0.9494 0.0754
63 0.6832 0.9499 0.6848 0.9407 0.0747
64 0.6838 0.9397 0.6984 0.9319 0.0777
65 0.6874 0.9307 0.6927 0.9228 0.0756
66 0.6926 0.9224 0.6939 0.9138 0.0748
67 0.6885 0.9144 0.7020 0.9052 0.0770
68 0.6960 0.9057 0.7023 0.8990 0.0792
69 0.6950 0.8980 0.7008 0.8892 0.0894
70 0.7005 0.8893 0.7038 0.8814 0.0840
71 0.6999 0.8820 0.7095 0.8739 0.0793
72 0.7045 0.8740 0.7068 0.8664 0.0840
73 0.7057 0.8668 0.7092 0.8599 0.1580
74 0.7078 0.8600 0.7071 0.8534 0.0785
75 0.7106 0.8537 0.7170 0.8461 0.0791
76 0.7062 0.8476 0.7164 0.8400 0.0812
77 0.7132 0.8402 0.7143 0.8338 0.0884
78 0.7122 0.8334 0.7188 0.8263 0.1156
79 0.7124 0.8271 0.7185 0.8201 0.0790
80 0.7150 0.8216 0.7149 0.8144 0.0813
81 0.7172 0.8157 0.7225 0.8084 0.0805
82 0.7193 0.8104 0.7240 0.8045 0.0821
83 0.7189 0.8039 0.7252 0.7969 0.0964
84 0.7214 0.7980 0.7219 0.7911 0.0839
85 0.7199 0.7924 0.7297 0.7867 0.0870
86 0.7222 0.7868 0.7321 0.7808 0.0785
87 0.7240 0.7813 0.7273 0.7758 0.0801
88 0.7255 0.7766 0.7333 0.7714 0.0849
89 0.7267 0.7720 0.7327 0.7680 0.0837
90 0.7289 0.7678 0.7342 0.7617 0.0811
91 0.7282 0.7627 0.7357 0.7569 0.0793
92 0.7308 0.7589 0.7354 0.7552 0.0803
93 0.7318 0.7543 0.7369 0.7482 0.0853
94 0.7322 0.7499 0.7360 0.7446 0.0962
95 0.7335 0.7459 0.7378 0.7407 0.0851
96 0.7319 0.7416 0.7396 0.7389 0.0812
97 0.7349 0.7390 0.7384 0.7328 0.0805
98 0.7339 0.7351 0.7408 0.7298 0.0791
99 0.7372 0.7299 0.7405 0.7282 0.0788
100 0.7386 0.7270 0.7408 0.7230 0.0868
101 0.7354 0.7243 0.7444 0.7198 0.0785
102 0.7395 0.7203 0.7456 0.7169 0.0767
103 0.7362 0.7172 0.7486 0.7114 0.0776
104 0.7397 0.7133 0.7489 0.7085 0.0772
105 0.7398 0.7093 0.7459 0.7054 0.0764
106 0.7414 0.7058 0.7502 0.7013 0.0780
107 0.7409 0.7028 0.7505 0.6969 0.0773
108 0.7448 0.6985 0.7480 0.6944 0.0840
109 0.7419 0.6954 0.7502 0.6909 0.0872
110 0.7451 0.6922 0.7502 0.6884 0.0846
111 0.7450 0.6892 0.7532 0.6855 0.1565
112 0.7477 0.6867 0.7511 0.6833 0.0812
113 0.7456 0.6837 0.7559 0.6813 0.0882
114 0.7480 0.6810 0.7511 0.6789 0.0883
115 0.7463 0.6784 0.7568 0.6757 0.0875
116 0.7507 0.6760 0.7520 0.6741 0.0853
117 0.7453 0.6745 0.7568 0.6710 0.0793
118 0.7517 0.6719 0.7514 0.6699 0.0793
119 0.7472 0.6699 0.7559 0.6665 0.0862
120 0.7516 0.6667 0.7595 0.6652 0.0899
121 0.7505 0.6650 0.7556 0.6628 0.1087
122 0.7516 0.6630 0.7586 0.6620 0.0801
123 0.7516 0.6617 0.7568 0.6625 0.0795
124 0.7520 0.6615 0.7586 0.6569 0.0804
125 0.7517 0.6584 0.7571 0.6562 0.0781
126 0.7556 0.6560 0.7628 0.6534 0.0786
127 0.7553 0.6531 0.7619 0.6514 0.0785
128 0.7554 0.6507 0.7634 0.6496 0.0769
129 0.7558 0.6489 0.7628 0.6472 0.0776
130 0.7573 0.6472 0.7643 0.6464 0.0773
131 0.7603 0.6455 0.7646 0.6442 0.0750
132 0.7557 0.6439 0.7682 0.6422 0.0858
133 0.7608 0.6421 0.7670 0.6413 0.0769
134 0.7560 0.6408 0.7703 0.6397 0.0766
135 0.7613 0.6388 0.7670 0.6376 0.0752
136 0.7596 0.6373 0.7691 0.6355 0.0773
137 0.7613 0.6356 0.7691 0.6345 0.0761
138 0.7612 0.6346 0.7718 0.6340 0.0765
139 0.7617 0.6329 0.7721 0.6322 0.0775
140 0.7614 0.6318 0.7700 0.6308 0.0768
141 0.7628 0.6302 0.7703 0.6285 0.0774
142 0.7611 0.6290 0.7730 0.6274 0.0754
143 0.7624 0.6287 0.7694 0.6315 0.0846
144 0.7651 0.6269 0.7721 0.6234 0.0767
145 0.7660 0.6248 0.7685 0.6228 0.0763
146 0.7652 0.6239 0.7715 0.6211 0.0757
147 0.7651 0.6221 0.7730 0.6208 0.0776
148 0.7659 0.6212 0.7730 0.6205 0.0761
149 0.7649 0.6197 0.7727 0.6173 0.1526
150 0.7678 0.6172 0.7730 0.6164 0.0768
151 0.7681 0.6165 0.7748 0.6162 0.0788
152 0.7681 0.6154 0.7745 0.6140 0.0770
153 0.7690 0.6138 0.7736 0.6127 0.0764
154 0.7698 0.6130 0.7709 0.6117 0.0785
155 0.7658 0.6120 0.7757 0.6111 0.0868
156 0.7711 0.6103 0.7772 0.6096 0.0833
157 0.7693 0.6093 0.7715 0.6097 0.0879
158 0.7696 0.6092 0.7778 0.6085 0.0795
159 0.7714 0.6070 0.7751 0.6059 0.0776
160 0.7715 0.6062 0.7763 0.6050 0.0804
161 0.7715 0.6046 0.7763 0.6036 0.0803
162 0.7727 0.6039 0.7793 0.6031 0.0768
163 0.7715 0.6023 0.7766 0.6009 0.0768
164 0.7726 0.6015 0.7772 0.6008 0.0770
165 0.7727 0.6004 0.7763 0.5996 0.0827
166 0.7736 0.5995 0.7784 0.5986 0.0861
167 0.7748 0.5984 0.7790 0.5974 0.0882
168 0.7743 0.5968 0.7781 0.5964 0.0850
169 0.7730 0.5965 0.7797 0.5953 0.0863
170 0.7751 0.5951 0.7778 0.5955 0.1073
171 0.7748 0.5938 0.7797 0.5938 0.0821
172 0.7749 0.5939 0.7775 0.5923 0.0827
173 0.7733 0.5925 0.7821 0.5919 0.0869
174 0.7773 0.5913 0.7803 0.5918 0.0883
175 0.7768 0.5909 0.7815 0.5896 0.0880
176 0.7751 0.5898 0.7800 0.5892 0.0791
177 0.7782 0.5891 0.7827 0.5898 0.0867
178 0.7776 0.5888 0.7803 0.5898 0.0854
179 0.7779 0.5879 0.7842 0.5867 0.0849
180 0.7790 0.5867 0.7775 0.5873 0.0834
181 0.7766 0.5867 0.7827 0.5840 0.0788
182 0.7797 0.5848 0.7824 0.5836 0.0876
183 0.7784 0.5840 0.7839 0.5848 0.0811
184 0.7800 0.5832 0.7854 0.5828 0.0776
185 0.7776 0.5827 0.7842 0.5812 0.0794
186 0.7775 0.5825 0.7809 0.5820 0.0788
187 0.7801 0.5812 0.7824 0.5803 0.1517
188 0.7811 0.5809 0.7860 0.5810 0.0779
189 0.7770 0.5801 0.7830 0.5818 0.0757
190 0.7826 0.5795 0.7854 0.5777 0.0761
191 0.7803 0.5774 0.7848 0.5767 0.0764
192 0.7835 0.5769 0.7875 0.5774 0.0764
193 0.7793 0.5758 0.7875 0.5756 0.0771
194 0.7833 0.5754 0.7881 0.5748 0.0844
195 0.7817 0.5743 0.7845 0.5747 0.0763
196 0.7815 0.5744 0.7896 0.5736 0.0774
197 0.7805 0.5728 0.7857 0.5731 0.0778
198 0.7826 0.5720 0.7872 0.5733 0.0777
199 0.7827 0.5719 0.7878 0.5712 0.0781
200 0.7834 0.5716 0.7884 0.5716 0.0761
201 0.7804 0.5708 0.7881 0.5707 0.0830
202 0.7864 0.5698 0.7896 0.5705 0.0858
203 0.7831 0.5686 0.7899 0.5691 0.0886
204 0.7852 0.5678 0.7914 0.5681 0.0825
205 0.7842 0.5673 0.7896 0.5673 0.0841
206 0.7849 0.5665 0.7905 0.5666 0.0829
207 0.7861 0.5656 0.7926 0.5668 0.0804
208 0.7863 0.5649 0.7905 0.5649 0.0975
209 0.7852 0.5643 0.7911 0.5660 0.1123
210 0.7867 0.5635 0.7917 0.5635 0.0828
211 0.7857 0.5629 0.7932 0.5640 0.0793
212 0.7873 0.5619 0.7935 0.5627 0.0785
213 0.7873 0.5611 0.7932 0.5617 0.0802
214 0.7881 0.5605 0.7920 0.5644 0.0795
215 0.7877 0.5612 0.7944 0.5610 0.0859
216 0.7888 0.5602 0.7938 0.5611 0.0863
217 0.7870 0.5590 0.7947 0.5598 0.0848
218 0.7897 0.5581 0.7938 0.5593 0.0814
219 0.7882 0.5585 0.7920 0.5582 0.0860
220 0.7882 0.5574 0.7941 0.5580 0.0871
221 0.7884 0.5564 0.7950 0.5577 0.0977
222 0.7901 0.5554 0.7953 0.5571 0.0799
223 0.7891 0.5551 0.7986 0.5562 0.1165
224 0.7885 0.5550 0.7959 0.5569 0.0872
225 0.7897 0.5553 0.7971 0.5565 0.1579
226 0.7893 0.5545 0.7938 0.5546 0.0796
227 0.7915 0.5534 0.7971 0.5537 0.0801
228 0.7897 0.5520 0.7977 0.5530 0.0802
229 0.7914 0.5517 0.7959 0.5551 0.0812
230 0.7917 0.5518 0.7950 0.5535 0.0823
231 0.7928 0.5503 0.7977 0.5514 0.0874
232 0.7925 0.5494 0.7962 0.5523 0.0803
233 0.7919 0.5490 0.7983 0.5504 0.0874
234 0.7926 0.5482 0.7980 0.5497 0.0776
235 0.7942 0.5480 0.7983 0.5490 0.0783
236 0.7931 0.5470 0.7983 0.5484 0.0752
237 0.7945 0.5468 0.7992 0.5488 0.0767
238 0.7937 0.5468 0.7995 0.5483 0.0767
239 0.7931 0.5461 0.7998 0.5470 0.0777
240 0.7923 0.5459 0.7989 0.5485 0.0802
241 0.7923 0.5449 0.7992 0.5459 0.0769
242 0.7955 0.5435 0.7995 0.5468 0.0762
243 0.7960 0.5435 0.7962 0.5464 0.0780
244 0.7947 0.5431 0.7995 0.5447 0.0769
245 0.7967 0.5423 0.8001 0.5445 0.0761
246 0.7947 0.5421 0.7989 0.5477 0.0777
247 0.7953 0.5420 0.8022 0.5439 0.0767
248 0.7946 0.5420 0.7998 0.5423 0.0850
249 0.7963 0.5407 0.8010 0.5425 0.0780
250 0.7960 0.5398 0.8025 0.5420 0.0791
251 0.7970 0.5388 0.8004 0.5418 0.0760
252 0.7964 0.5385 0.8013 0.5400 0.0783
253 0.7963 0.5388 0.7983 0.5408 0.0799
254 0.7963 0.5375 0.8034 0.5395 0.0806
255 0.7970 0.5380 0.8004 0.5419 0.0875
256 0.7977 0.5373 0.8034 0.5381 0.1217
257 0.7985 0.5351 0.8028 0.5375 0.0808
258 0.7987 0.5345 0.7995 0.5401 0.0790
259 0.7989 0.5342 0.8049 0.5361 0.0803
260 0.7990 0.5344 0.8019 0.5370 0.0822
261 0.7964 0.5343 0.8040 0.5354 0.0944
262 0.8004 0.5344 0.8028 0.5384 0.0850
263 0.7964 0.5335 0.8019 0.5358 0.1561
264 0.7998 0.5324 0.8037 0.5347 0.0799
265 0.8006 0.5315 0.8031 0.5353 0.0793
266 0.7986 0.5314 0.8001 0.5376 0.0804
267 0.7998 0.5312 0.8052 0.5319 0.0792
268 0.8013 0.5298 0.8040 0.5322 0.0810
269 0.8013 0.5291 0.8040 0.5330 0.0775
270 0.8023 0.5284 0.8064 0.5307 0.0788
271 0.8010 0.5290 0.8022 0.5332 0.0767
272 0.8007 0.5282 0.8019 0.5327 0.0771
273 0.8004 0.5286 0.8046 0.5310 0.0840
274 0.8021 0.5275 0.8076 0.5290 0.0775
275 0.8007 0.5264 0.8046 0.5315 0.0780
276 0.8022 0.5262 0.8052 0.5285 0.0778
277 0.8020 0.5256 0.8052 0.5288 0.0774
278 0.8028 0.5257 0.8052 0.5283 0.0763
279 0.8015 0.5248 0.8049 0.5292 0.0779
280 0.8029 0.5248 0.8049 0.5269 0.0782
281 0.8025 0.5239 0.8070 0.5258 0.0778
282 0.8041 0.5231 0.8055 0.5273 0.0765
283 0.8028 0.5234 0.8049 0.5269 0.0767
284 0.8034 0.5226 0.8092 0.5251 0.0776
285 0.8056 0.5217 0.8046 0.5244 0.0779
286 0.8053 0.5207 0.8061 0.5274 0.0781
287 0.8025 0.5214 0.8098 0.5229 0.0762
288 0.8034 0.5210 0.8049 0.5252 0.0768
289 0.8040 0.5204 0.8073 0.5223 0.0856
290 0.8056 0.5188 0.8085 0.5223 0.0765
291 0.8035 0.5196 0.8076 0.5222 0.0773
292 0.8053 0.5182 0.8098 0.5212 0.0765
293 0.8068 0.5178 0.8070 0.5219 0.0764
294 0.8056 0.5174 0.8092 0.5209 0.0770
295 0.8037 0.5173 0.8095 0.5207 0.0789
296 0.8084 0.5170 0.8101 0.5196 0.0843
297 0.8044 0.5164 0.8104 0.5185 0.1305
298 0.8076 0.5160 0.8067 0.5235 0.0965
299 0.8054 0.5167 0.8079 0.5193 0.0857
300 0.8070 0.5159 0.8095 0.5185 0.0854
301 0.8062 0.5149 0.8125 0.5168 0.2071
302 0.8095 0.5144 0.8089 0.5198 0.0842
303 0.8057 0.5143 0.8113 0.5174 0.0927
304 0.8086 0.5141 0.8122 0.5166 0.0870
305 0.8073 0.5136 0.8131 0.5154 0.0845
306 0.8081 0.5133 0.8092 0.5170 0.0843
307 0.8068 0.5119 0.8122 0.5151 0.0910
308 0.8077 0.5121 0.8122 0.5147 0.1250
309 0.8080 0.5114 0.8119 0.5154 0.0917
310 0.8083 0.5114 0.8113 0.5137 0.0897
311 0.8081 0.5108 0.8128 0.5137 0.0909
312 0.8092 0.5101 0.8098 0.5146 0.0796
313 0.8074 0.5102 0.8119 0.5145 0.0797
314 0.8083 0.5109 0.8116 0.5132 0.0826
315 0.8087 0.5087 0.8155 0.5119 0.0777
316 0.8092 0.5091 0.8104 0.5127 0.0774
317 0.8077 0.5081 0.8125 0.5125 0.0779
318 0.8104 0.5086 0.8110 0.5124 0.0770
319 0.8098 0.5079 0.8116 0.5108 0.0893
320 0.8101 0.5074 0.8125 0.5118 0.1006
321 0.8114 0.5065 0.8128 0.5102 0.0900
322 0.8110 0.5060 0.8140 0.5099 0.0808
323 0.8107 0.5062 0.8116 0.5093 0.0791
324 0.8096 0.5056 0.8152 0.5089 0.0798
325 0.8120 0.5048 0.8125 0.5093 0.0790
326 0.8107 0.5045 0.8125 0.5098 0.0784
327 0.8109 0.5041 0.8164 0.5080 0.0788
328 0.8117 0.5044 0.8131 0.5090 0.0780
329 0.8105 0.5045 0.8164 0.5071 0.0762
330 0.8111 0.5037 0.8128 0.5102 0.0826
331 0.8104 0.5030 0.8146 0.5076 0.0983
332 0.8129 0.5028 0.8134 0.5077 0.1175
333 0.8107 0.5021 0.8137 0.5064 0.0886
334 0.8117 0.5021 0.8155 0.5067 0.0923
335 0.8126 0.5013 0.8143 0.5055 0.1181
336 0.8131 0.5008 0.8146 0.5062 0.0842
337 0.8132 0.5007 0.8161 0.5049 0.0896
338 0.8118 0.5005 0.8167 0.5045 0.0909
339 0.8126 0.5004 0.8143 0.5058 0.1715
340 0.8121 0.4998 0.8143 0.5043 0.0863
341 0.8135 0.4997 0.8137 0.5066 0.0919
342 0.8090 0.5009 0.8179 0.5029 0.0861
343 0.8142 0.4989 0.8140 0.5057 0.0927
344 0.8124 0.4994 0.8155 0.5045 0.0881
345 0.8138 0.4987 0.8137 0.5043 0.0945
346 0.8144 0.4979 0.8179 0.5028 0.0848
347 0.8128 0.4978 0.8185 0.5019 0.0872
348 0.8143 0.4967 0.8137 0.5034 0.0889
349 0.8146 0.4969 0.8185 0.5006 0.0866
350 0.8133 0.4961 0.8146 0.5022 0.0859
351 0.8147 0.4955 0.8185 0.4998 0.0842
352 0.8144 0.4954 0.8161 0.5030 0.0842
353 0.8169 0.4956 0.8167 0.5004 0.0799
354 0.8141 0.4954 0.8191 0.4989 0.0872
355 0.8145 0.4954 0.8149 0.5001 0.0875
356 0.8132 0.4944 0.8173 0.5003 0.0874
357 0.8161 0.4938 0.8152 0.4995 0.0803
358 0.8142 0.4944 0.8233 0.4978 0.0799
359 0.8153 0.4937 0.8146 0.4986 0.0771
360 0.8154 0.4926 0.8152 0.4990 0.0771
361 0.8149 0.4921 0.8194 0.4977 0.0786
362 0.8166 0.4925 0.8200 0.4969 0.0769
363 0.8153 0.4919 0.8185 0.4974 0.0756
364 0.8164 0.4918 0.8179 0.4977 0.0775
365 0.8148 0.4921 0.8197 0.4972 0.0811
366 0.8144 0.4914 0.8197 0.4951 0.0838
367 0.8139 0.4909 0.8167 0.4982 0.0826
368 0.8174 0.4912 0.8188 0.4966 0.0808
369 0.8163 0.4898 0.8179 0.4966 0.0893
370 0.8177 0.4902 0.8203 0.4962 0.0916
371 0.8168 0.4893 0.8215 0.4944 0.0830
372 0.8162 0.4894 0.8206 0.4962 0.0807
373 0.8166 0.4885 0.8170 0.4958 0.0855
374 0.8160 0.4893 0.8188 0.4954 0.0898
375 0.8163 0.4895 0.8245 0.4927 0.0860
376 0.8180 0.4886 0.8239 0.4933 0.0814
377 0.8182 0.4872 0.8227 0.4920 0.1624
378 0.8171 0.4872 0.8233 0.4915 0.0797
379 0.8196 0.4879 0.8203 0.4928 0.0790
380 0.8174 0.4869 0.8200 0.4920 0.0794
381 0.8204 0.4859 0.8209 0.4916 0.0784
382 0.8173 0.4850 0.8227 0.4916 0.0767
383 0.8195 0.4847 0.8230 0.4904 0.0768
384 0.8191 0.4852 0.8236 0.4910 0.0775
385 0.8202 0.4850 0.8182 0.4939 0.0787
386 0.8165 0.4856 0.8206 0.4920 0.0794
387 0.8199 0.4845 0.8218 0.4916 0.0848
388 0.8188 0.4856 0.8272 0.4897 0.0873
389 0.8198 0.4851 0.8191 0.4932 0.0854
390 0.8201 0.4840 0.8215 0.4910 0.1166
391 0.8199 0.4829 0.8221 0.4890 0.0789
392 0.8209 0.4828 0.8230 0.4887 0.0797
393 0.8203 0.4828 0.8242 0.4885 0.0790
394 0.8223 0.4823 0.8227 0.4880 0.0799
395 0.8206 0.4821 0.8203 0.4928 0.0791
396 0.8207 0.4831 0.8263 0.4870 0.0800
397 0.8205 0.4834 0.8263 0.4872 0.0812
398 0.8190 0.4816 0.8230 0.4879 0.0768
399 0.8229 0.4809 0.8203 0.4925 0.0767
400 0.8196 0.4825 0.8263 0.4857 0.0825
401 0.8220 0.4800 0.8269 0.4870 0.0788
402 0.8201 0.4814 0.8272 0.4845 0.0772
403 0.8208 0.4806 0.8230 0.4877 0.0825
404 0.8216 0.4794 0.8245 0.4865 0.0759
405 0.8217 0.4787 0.8248 0.4843 0.0767
406 0.8233 0.4786 0.8269 0.4865 0.0782
407 0.8230 0.4784 0.8272 0.4838 0.0767
408 0.8217 0.4787 0.8236 0.4843 0.0774
409 0.8213 0.4777 0.8242 0.4843 0.0764
410 0.8220 0.4770 0.8275 0.4829 0.0767
411 0.8217 0.4768 0.8248 0.4846 0.0787
412 0.8231 0.4765 0.8278 0.4836 0.0784
413 0.8237 0.4764 0.8260 0.4836 0.0765
414 0.8229 0.4761 0.8242 0.4856 0.0774
415 0.8230 0.4772 0.8269 0.4822 0.0773
416 0.8233 0.4757 0.8287 0.4819 0.1596
417 0.8215 0.4757 0.8296 0.4806 0.0772
418 0.8245 0.4756 0.8251 0.4826 0.0768
419 0.8222 0.4755 0.8272 0.4825 0.0771
420 0.8229 0.4753 0.8257 0.4833 0.0766
421 0.8237 0.4754 0.8308 0.4810 0.0769
422 0.8230 0.4752 0.8284 0.4802 0.0777
423 0.8247 0.4736 0.8266 0.4818 0.0773
424 0.8238 0.4730 0.8272 0.4805 0.0760
425 0.8247 0.4739 0.8269 0.4825 0.0768
426 0.8232 0.4727 0.8305 0.4784 0.0769
427 0.8232 0.4734 0.8320 0.4782 0.0777
428 0.8243 0.4726 0.8257 0.4812 0.0755
429 0.8247 0.4725 0.8317 0.4779 0.0777
430 0.8227 0.4719 0.8257 0.4816 0.0848
431 0.8242 0.4715 0.8290 0.4781 0.0781
432 0.8246 0.4712 0.8317 0.4769 0.0785
433 0.8241 0.4705 0.8293 0.4781 0.0762
434 0.8241 0.4701 0.8338 0.4764 0.0775
435 0.8254 0.4703 0.8287 0.4782 0.0771
436 0.8248 0.4700 0.8305 0.4776 0.0770
437 0.8255 0.4700 0.8290 0.4767 0.1025
438 0.8244 0.4690 0.8293 0.4769 0.0786
439 0.8266 0.4693 0.8287 0.4761 0.0783
440 0.8269 0.4696 0.8266 0.4791 0.0777
441 0.8245 0.4689 0.8338 0.4754 0.0767
442 0.8275 0.4688 0.8287 0.4773 0.0778
443 0.8221 0.4686 0.8299 0.4769 0.0798
444 0.8289 0.4693 0.8230 0.4789 0.0859
445 0.8226 0.4693 0.8317 0.4751 0.0773
446 0.8269 0.4680 0.8284 0.4755 0.0778
447 0.8254 0.4670 0.8290 0.4749 0.0772
448 0.8259 0.4674 0.8254 0.4752 0.0794
449 0.8259 0.4668 0.8299 0.4744 0.0772
450 0.8267 0.4668 0.8332 0.4727 0.0767
451 0.8274 0.4660 0.8305 0.4735 0.0779
452 0.8254 0.4660 0.8326 0.4734 0.0764
453 0.8268 0.4655 0.8287 0.4762 0.0766
454 0.8260 0.4661 0.8344 0.4727 0.1515
455 0.8278 0.4656 0.8299 0.4739 0.0768
456 0.8254 0.4649 0.8305 0.4727 0.0779
457 0.8289 0.4655 0.8290 0.4730 0.0846
458 0.8260 0.4655 0.8329 0.4722 0.0774
459 0.8282 0.4638 0.8317 0.4720 0.0760
460 0.8244 0.4649 0.8335 0.4726 0.0773
461 0.8289 0.4646 0.8293 0.4716 0.0758
462 0.8275 0.4629 0.8284 0.4733 0.0766
463 0.8302 0.4637 0.8251 0.4731 0.0781
464 0.8272 0.4632 0.8374 0.4700 0.0762
465 0.8303 0.4623 0.8290 0.4717 0.0804
466 0.8264 0.4624 0.8353 0.4694 0.0762
467 0.8302 0.4620 0.8314 0.4702 0.0764
468 0.8281 0.4615 0.8323 0.4703 0.0771
469 0.8275 0.4617 0.8344 0.4687 0.0778
470 0.8292 0.4613 0.8314 0.4699 0.0772
471 0.8295 0.4610 0.8347 0.4691 0.0829
472 0.8297 0.4608 0.8320 0.4694 0.0776
473 0.8293 0.4606 0.8326 0.4695 0.0754
474 0.8318 0.4606 0.8317 0.4700 0.0777
475 0.8299 0.4602 0.8353 0.4684 0.0778
476 0.8290 0.4608 0.8356 0.4676 0.0782
477 0.8284 0.4607 0.8317 0.4717 0.0784
478 0.8283 0.4598 0.8359 0.4672 0.0763
479 0.8311 0.4595 0.8317 0.4678 0.0764
480 0.8302 0.4590 0.8362 0.4680 0.0766
481 0.8307 0.4596 0.8308 0.4689 0.0766
482 0.8308 0.4588 0.8323 0.4691 0.0769
483 0.8290 0.4587 0.8344 0.4656 0.1038
484 0.8330 0.4584 0.8329 0.4672 0.0765
485 0.8313 0.4582 0.8365 0.4652 0.0783
486 0.8324 0.4577 0.8302 0.4693 0.0780
487 0.8292 0.4572 0.8335 0.4677 0.0767
488 0.8321 0.4582 0.8317 0.4700 0.0833
489 0.8293 0.4579 0.8368 0.4643 0.0768
490 0.8314 0.4573 0.8314 0.4667 0.0755
491 0.8327 0.4565 0.8341 0.4655 0.0777
492 0.8326 0.4561 0.8353 0.4662 0.0767
493 0.8305 0.4569 0.8329 0.4663 0.1526
494 0.8319 0.4571 0.8326 0.4667 0.0780
495 0.8305 0.4555 0.8341 0.4640 0.0773
496 0.8318 0.4547 0.8362 0.4635 0.0766
497 0.8332 0.4548 0.8329 0.4644 0.0758
498 0.8326 0.4548 0.8368 0.4633 0.0767
499 0.8308 0.4551 0.8347 0.4643 0.0777
500 0.8331 0.4543 0.8356 0.4641 0.0775
Out[22]:
<class 'skorch.classifier.NeuralNetClassifier'>[initialized](
module_=VanillaMLP(
(model): Sequential(
(0): Linear(in_features=22, out_features=32, bias=True)
(1): ReLU()
(2): Linear(in_features=32, out_features=7, bias=True)
(3): Softmax(dim=-1)
)
),
)In a Jupyter environment, please rerun this cell to show the HTML representation or trust the notebook. On GitHub, the HTML representation is unable to render, please try loading this page with nbviewer.org.
<class 'skorch.classifier.NeuralNetClassifier'>[initialized](
module_=VanillaMLP(
(model): Sequential(
(0): Linear(in_features=22, out_features=32, bias=True)
(1): ReLU()
(2): Linear(in_features=32, out_features=7, bias=True)
(3): Softmax(dim=-1)
)
),
)In [23]:
y_pred = net.predict(X_test)
f1_score(y_test, y_pred, average="weighted")
Out[23]:
0.8323364703729181
In [24]:
history = net.history
plt.plot(history[:, 'train_loss'], label='train_loss')
plt.plot(history[:, 'valid_loss'], label='valid_loss')
plt.legend()
plt.xlabel('Epoch')
plt.ylabel('Loss')
plt.show()
plt.plot(history[:, 'train_acc'], label='train_acc')
plt.plot(history[:, 'valid_acc'], label='valid_acc')
plt.legend()
plt.xlabel('Epoch')
plt.ylabel('Accuracy')
plt.show()
Pipeline with normalization¶
In [25]:
net = NeuralNetClassifier(
VanillaMLP,
module__in_features=X_train.shape[1],
module__out_features=len(le.classes_),
max_epochs=500,
device=DEVICE,
optimizer=torch.optim.Adam,
lr=1e-2,
callbacks=[
EpochScoring(
scoring='accuracy',
name='train_acc',
on_train=True,
lower_is_better=False,
),
],
batch_size=2048,
iterator_train__shuffle=True,
)
net
Out[25]:
<class 'skorch.classifier.NeuralNetClassifier'>[uninitialized]( module=<class '__main__.VanillaMLP'>, module__in_features=22, module__out_features=7, )In a Jupyter environment, please rerun this cell to show the HTML representation or trust the notebook.
On GitHub, the HTML representation is unable to render, please try loading this page with nbviewer.org.
<class 'skorch.classifier.NeuralNetClassifier'>[uninitialized]( module=<class '__main__.VanillaMLP'>, module__in_features=22, module__out_features=7, )
In [26]:
from sklearn.pipeline import make_pipeline
from sklearn.preprocessing import StandardScaler
pipe = make_pipeline(StandardScaler(), net)
pipe
Out[26]:
Pipeline(steps=[('standardscaler', StandardScaler()),
('neuralnetclassifier',
NeuralNetClassifier(_params_to_validate={'iterator_train__shuffle', 'module__out_features', 'module__in_features'}, batch_size=2048, callbacks=[<skorch.callbacks.scoring.EpochScoring object at 0x31b2738c0>], compile=False, dataset=<class 'skorch.dataset.Dataset'>, device='mps', iterator...ataLoader'>, iterator_train__shuffle=True, iterator_valid=<class 'torch.utils.data.dataloader.DataLoader'>, lr=0.01, max_epochs=500, module=<class '__main__.VanillaMLP'>, module__in_features=22, module__out_features=7, optimizer=<class 'torch.optim.adam.Adam'>, predict_nonlinearity='auto', torch_load_kwargs=None, use_caching='auto', verbose=1, warm_start=False))])In a Jupyter environment, please rerun this cell to show the HTML representation or trust the notebook. On GitHub, the HTML representation is unable to render, please try loading this page with nbviewer.org.
Pipeline(steps=[('standardscaler', StandardScaler()),
('neuralnetclassifier',
NeuralNetClassifier(_params_to_validate={'iterator_train__shuffle', 'module__out_features', 'module__in_features'}, batch_size=2048, callbacks=[<skorch.callbacks.scoring.EpochScoring object at 0x31b2738c0>], compile=False, dataset=<class 'skorch.dataset.Dataset'>, device='mps', iterator...ataLoader'>, iterator_train__shuffle=True, iterator_valid=<class 'torch.utils.data.dataloader.DataLoader'>, lr=0.01, max_epochs=500, module=<class '__main__.VanillaMLP'>, module__in_features=22, module__out_features=7, optimizer=<class 'torch.optim.adam.Adam'>, predict_nonlinearity='auto', torch_load_kwargs=None, use_caching='auto', verbose=1, warm_start=False))])StandardScaler()
<class 'skorch.classifier.NeuralNetClassifier'>[uninitialized]( module=<class '__main__.VanillaMLP'>, module__in_features=22, module__out_features=7, )
In [27]:
# Training the network
pipe.fit(X_train, y_train)
epoch train_acc train_loss valid_acc valid_loss dur
------- ----------- ------------ ----------- ------------ ------
1 0.3780 1.7503 0.5406 1.4481 0.0785
2 0.5583 1.2945 0.5939 1.0888 0.0761
3 0.6043 1.0283 0.6276 0.9224 0.0770
4 0.6506 0.8943 0.6749 0.8287 0.0760
5 0.6817 0.8096 0.7014 0.7643 0.0801
6 0.7104 0.7471 0.7279 0.7110 0.0961
7 0.7334 0.6930 0.7486 0.6616 0.0852
8 0.7594 0.6437 0.7673 0.6199 0.0786
9 0.7736 0.6025 0.7857 0.5844 0.0784
10 0.7901 0.5679 0.7977 0.5549 0.0763
11 0.8007 0.5389 0.8079 0.5319 0.0736
12 0.8143 0.5128 0.8164 0.5104 0.0741
13 0.8196 0.4914 0.8221 0.4930 0.0741
14 0.8270 0.4737 0.8296 0.4765 0.0746
15 0.8351 0.4576 0.8308 0.4676 0.0739
16 0.8393 0.4443 0.8380 0.4551 0.0761
17 0.8465 0.4314 0.8402 0.4471 0.0829
18 0.8503 0.4223 0.8438 0.4377 0.0830
19 0.8557 0.4116 0.8462 0.4296 0.1117
20 0.8587 0.4043 0.8474 0.4252 0.0792
21 0.8598 0.3966 0.8501 0.4191 0.0783
22 0.8630 0.3908 0.8525 0.4134 0.0788
23 0.8651 0.3857 0.8501 0.4108 0.0809
24 0.8662 0.3812 0.8549 0.4078 0.0856
25 0.8681 0.3768 0.8546 0.4047 0.0882
26 0.8701 0.3734 0.8558 0.4042 0.0779
27 0.8701 0.3696 0.8576 0.3999 0.0789
28 0.8739 0.3657 0.8576 0.3988 0.1562
29 0.8737 0.3632 0.8615 0.3973 0.0772
30 0.8748 0.3608 0.8597 0.3957 0.0763
31 0.8741 0.3595 0.8576 0.3939 0.0831
32 0.8776 0.3550 0.8639 0.3925 0.0856
33 0.8781 0.3530 0.8621 0.3908 0.0846
34 0.8783 0.3514 0.8627 0.3940 0.0796
35 0.8799 0.3499 0.8633 0.3905 0.0769
36 0.8800 0.3483 0.8642 0.3907 0.0759
37 0.8817 0.3473 0.8648 0.3881 0.0756
38 0.8811 0.3470 0.8651 0.3890 0.0767
39 0.8808 0.3468 0.8645 0.3916 0.0743
40 0.8838 0.3426 0.8663 0.3885 0.0750
41 0.8820 0.3434 0.8630 0.3914 0.0758
42 0.8827 0.3422 0.8648 0.3875 0.0744
43 0.8832 0.3411 0.8669 0.3860 0.0758
44 0.8844 0.3403 0.8651 0.3878 0.0792
45 0.8842 0.3390 0.8660 0.3881 0.0846
46 0.8850 0.3377 0.8679 0.3844 0.0780
47 0.8858 0.3373 0.8663 0.3865 0.0781
48 0.8844 0.3366 0.8679 0.3849 0.0753
49 0.8852 0.3366 0.8642 0.3911 0.0750
50 0.8854 0.3375 0.8672 0.3848 0.0749
51 0.8861 0.3342 0.8672 0.3841 0.1094
52 0.8868 0.3350 0.8666 0.3884 0.0775
53 0.8869 0.3333 0.8663 0.3839 0.0752
54 0.8867 0.3325 0.8648 0.3847 0.0758
55 0.8860 0.3327 0.8703 0.3883 0.0772
56 0.8854 0.3326 0.8657 0.3886 0.0773
57 0.8884 0.3313 0.8633 0.3883 0.0823
58 0.8882 0.3300 0.8697 0.3845 0.0837
59 0.8872 0.3294 0.8700 0.3848 0.0847
60 0.8863 0.3307 0.8654 0.3875 0.0796
61 0.8881 0.3292 0.8685 0.3855 0.0789
62 0.8866 0.3294 0.8672 0.3846 0.0787
63 0.8886 0.3282 0.8682 0.3853 0.0759
64 0.8871 0.3266 0.8636 0.3875 0.0755
65 0.8885 0.3269 0.8675 0.3813 0.0739
66 0.8887 0.3260 0.8660 0.3860 0.1527
67 0.8884 0.3250 0.8679 0.3829 0.0761
68 0.8897 0.3255 0.8648 0.3907 0.0737
69 0.8869 0.3266 0.8663 0.3854 0.0766
70 0.8887 0.3253 0.8660 0.3840 0.0747
71 0.8878 0.3250 0.8651 0.3922 0.0756
72 0.8897 0.3248 0.8660 0.3853 0.0740
73 0.8875 0.3247 0.8694 0.3860 0.0750
74 0.8889 0.3232 0.8663 0.3841 0.0743
75 0.8886 0.3232 0.8666 0.3861 0.0757
76 0.8870 0.3233 0.8651 0.3875 0.0759
77 0.8902 0.3232 0.8700 0.3842 0.0758
78 0.8898 0.3216 0.8663 0.3867 0.0787
79 0.8891 0.3226 0.8657 0.3881 0.0839
80 0.8896 0.3229 0.8691 0.3854 0.0936
81 0.8882 0.3233 0.8621 0.3928 0.0891
82 0.8879 0.3230 0.8675 0.3830 0.0807
83 0.8889 0.3218 0.8657 0.3931 0.1043
84 0.8910 0.3196 0.8694 0.3819 0.0803
85 0.8920 0.3185 0.8663 0.3860 0.0780
86 0.8904 0.3196 0.8688 0.3831 0.0769
87 0.8899 0.3191 0.8669 0.3847 0.0860
88 0.8906 0.3184 0.8685 0.3860 0.0773
89 0.8905 0.3180 0.8685 0.3840 0.0773
90 0.8919 0.3193 0.8685 0.3847 0.0771
91 0.8898 0.3204 0.8685 0.3851 0.0758
92 0.8902 0.3207 0.8682 0.3893 0.0769
93 0.8894 0.3193 0.8682 0.3865 0.0770
94 0.8919 0.3186 0.8691 0.3840 0.0770
95 0.8913 0.3170 0.8657 0.3856 0.0810
96 0.8908 0.3168 0.8727 0.3843 0.0848
97 0.8905 0.3177 0.8727 0.3843 0.0867
98 0.8912 0.3189 0.8685 0.3886 0.0816
99 0.8902 0.3183 0.8688 0.3822 0.0795
100 0.8916 0.3167 0.8685 0.3884 0.0790
101 0.8917 0.3172 0.8706 0.3815 0.0773
102 0.8932 0.3157 0.8675 0.3872 0.0775
103 0.8905 0.3157 0.8688 0.3865 0.0916
104 0.8918 0.3148 0.8688 0.3835 0.1666
105 0.8905 0.3160 0.8709 0.3827 0.0801
106 0.8908 0.3171 0.8694 0.3858 0.0801
107 0.8899 0.3156 0.8657 0.3930 0.0787
108 0.8923 0.3163 0.8700 0.3850 0.0804
109 0.8915 0.3158 0.8694 0.3850 0.0797
110 0.8916 0.3167 0.8706 0.3826 0.0802
111 0.8906 0.3148 0.8706 0.3864 0.0775
112 0.8927 0.3143 0.8694 0.3830 0.0775
113 0.8907 0.3137 0.8691 0.3826 0.1010
114 0.8920 0.3138 0.8691 0.3888 0.0782
115 0.8928 0.3131 0.8712 0.3833 0.0769
116 0.8921 0.3133 0.8679 0.3831 0.0766
117 0.8929 0.3128 0.8712 0.3839 0.0774
118 0.8927 0.3132 0.8679 0.3876 0.0775
119 0.8936 0.3128 0.8700 0.3840 0.0774
120 0.8924 0.3124 0.8712 0.3896 0.0849
121 0.8928 0.3144 0.8712 0.3846 0.0768
122 0.8939 0.3118 0.8709 0.3857 0.0792
123 0.8908 0.3136 0.8700 0.3881 0.0816
124 0.8920 0.3137 0.8700 0.3850 0.0845
125 0.8926 0.3135 0.8715 0.3882 0.0823
126 0.8933 0.3113 0.8703 0.3860 0.0818
127 0.8936 0.3112 0.8694 0.3845 0.0821
128 0.8945 0.3112 0.8691 0.3878 0.0803
129 0.8933 0.3107 0.8709 0.3870 0.0794
130 0.8943 0.3105 0.8733 0.3845 0.0905
131 0.8922 0.3127 0.8666 0.3933 0.0865
132 0.8936 0.3111 0.8703 0.3873 0.0858
133 0.8927 0.3113 0.8715 0.3892 0.0861
134 0.8933 0.3114 0.8712 0.3864 0.0879
135 0.8924 0.3110 0.8730 0.3849 0.0865
136 0.8958 0.3089 0.8703 0.3856 0.0816
137 0.8948 0.3093 0.8694 0.3867 0.0844
138 0.8948 0.3089 0.8727 0.3892 0.0817
139 0.8948 0.3089 0.8691 0.3870 0.0819
140 0.8948 0.3080 0.8724 0.3876 0.0818
141 0.8945 0.3086 0.8682 0.3833 0.0798
142 0.8947 0.3077 0.8700 0.3899 0.1540
143 0.8931 0.3089 0.8682 0.3873 0.0772
144 0.8945 0.3091 0.8679 0.3889 0.0764
145 0.8932 0.3085 0.8715 0.3865 0.0768
146 0.8942 0.3080 0.8697 0.3871 0.0766
147 0.8936 0.3078 0.8682 0.3883 0.0766
148 0.8950 0.3081 0.8682 0.3880 0.0777
149 0.8953 0.3077 0.8700 0.3899 0.0785
150 0.8930 0.3089 0.8718 0.3868 0.0862
151 0.8952 0.3063 0.8682 0.3904 0.0769
152 0.8963 0.3069 0.8700 0.3851 0.0775
153 0.8960 0.3066 0.8694 0.3888 0.0763
154 0.8945 0.3067 0.8694 0.3905 0.0771
155 0.8940 0.3071 0.8703 0.3934 0.0768
156 0.8926 0.3091 0.8691 0.3902 0.0778
157 0.8939 0.3081 0.8663 0.3912 0.0766
158 0.8944 0.3073 0.8688 0.3879 0.0814
159 0.8951 0.3063 0.8709 0.3889 0.0851
160 0.8946 0.3059 0.8718 0.3908 0.1195
161 0.8957 0.3063 0.8672 0.3882 0.0818
162 0.8950 0.3056 0.8682 0.3926 0.0835
163 0.8954 0.3051 0.8694 0.3886 0.0869
164 0.8950 0.3055 0.8697 0.3921 0.0782
165 0.8966 0.3053 0.8697 0.3898 0.0780
166 0.8966 0.3061 0.8682 0.3937 0.0797
167 0.8959 0.3077 0.8691 0.3888 0.0789
168 0.8954 0.3054 0.8679 0.3955 0.0800
169 0.8965 0.3041 0.8639 0.3942 0.0833
170 0.8958 0.3060 0.8703 0.3904 0.0887
171 0.8942 0.3057 0.8645 0.4002 0.0878
172 0.8942 0.3065 0.8694 0.3937 0.0820
173 0.8952 0.3062 0.8700 0.3920 0.0798
174 0.8943 0.3065 0.8648 0.3997 0.0803
175 0.8954 0.3052 0.8706 0.3913 0.0796
176 0.8951 0.3045 0.8624 0.3981 0.0815
177 0.8954 0.3051 0.8682 0.3892 0.0863
178 0.8959 0.3033 0.8685 0.3968 0.0774
179 0.8966 0.3035 0.8675 0.3942 0.0763
180 0.8963 0.3032 0.8691 0.3921 0.1556
181 0.8972 0.3038 0.8709 0.3913 0.0859
182 0.8948 0.3042 0.8685 0.3958 0.0774
183 0.8976 0.3040 0.8663 0.3931 0.0781
184 0.8974 0.3038 0.8654 0.3940 0.0769
185 0.8980 0.3039 0.8688 0.3921 0.0759
186 0.8956 0.3051 0.8682 0.3940 0.0770
187 0.8959 0.3032 0.8675 0.3988 0.0774
188 0.8975 0.3029 0.8679 0.3923 0.0775
189 0.8954 0.3038 0.8715 0.3919 0.0774
190 0.8937 0.3055 0.8651 0.4035 0.0759
191 0.8951 0.3056 0.8691 0.3938 0.0860
192 0.8966 0.3030 0.8669 0.3909 0.0782
193 0.8983 0.3017 0.8675 0.3913 0.0764
194 0.8966 0.3024 0.8724 0.3945 0.0948
195 0.8954 0.3027 0.8706 0.3931 0.0764
196 0.8961 0.3017 0.8672 0.3944 0.0754
197 0.8962 0.3023 0.8672 0.3932 0.0786
198 0.8959 0.3038 0.8724 0.3941 0.0773
199 0.8968 0.3016 0.8666 0.3947 0.0773
200 0.8969 0.3012 0.8669 0.3917 0.0772
201 0.8961 0.3005 0.8682 0.3911 0.0779
202 0.8969 0.3013 0.8706 0.3913 0.0765
203 0.8985 0.2999 0.8679 0.3936 0.0772
204 0.8966 0.3005 0.8688 0.3966 0.0779
205 0.8974 0.3005 0.8682 0.3981 0.0782
206 0.8979 0.3013 0.8709 0.3931 0.0770
207 0.8973 0.3008 0.8648 0.3962 0.0852
208 0.8969 0.3017 0.8679 0.3953 0.0773
209 0.8984 0.3004 0.8663 0.3913 0.0779
210 0.8991 0.2992 0.8706 0.3922 0.0772
211 0.8970 0.3001 0.8666 0.3947 0.0766
212 0.8970 0.3002 0.8666 0.3949 0.0751
213 0.8991 0.2993 0.8682 0.3906 0.0776
214 0.8973 0.2995 0.8709 0.3946 0.0776
215 0.8968 0.2999 0.8682 0.3930 0.0777
216 0.8985 0.2991 0.8651 0.3928 0.0766
217 0.8981 0.2995 0.8663 0.3942 0.0773
218 0.8971 0.2996 0.8709 0.3948 0.1538
219 0.8978 0.2984 0.8694 0.3911 0.0773
220 0.9000 0.2993 0.8657 0.3969 0.0775
221 0.8958 0.3010 0.8712 0.3920 0.0761
222 0.8967 0.3008 0.8627 0.3959 0.0786
223 0.8979 0.3001 0.8675 0.3944 0.0780
224 0.8972 0.3005 0.8654 0.3944 0.0852
225 0.8975 0.3013 0.8657 0.3976 0.0773
226 0.8979 0.2988 0.8718 0.3920 0.0768
227 0.8983 0.2998 0.8682 0.3924 0.0772
228 0.8971 0.2989 0.8657 0.3992 0.0769
229 0.8957 0.3006 0.8700 0.3926 0.0773
230 0.8963 0.2996 0.8633 0.3956 0.0776
231 0.8972 0.3001 0.8682 0.3940 0.0775
232 0.8988 0.2980 0.8645 0.3953 0.0768
233 0.8991 0.2971 0.8685 0.3930 0.0778
234 0.8994 0.2980 0.8645 0.3991 0.0772
235 0.8997 0.2971 0.8706 0.3932 0.0782
236 0.8987 0.2982 0.8654 0.3980 0.0768
237 0.8985 0.2983 0.8691 0.3948 0.0768
238 0.8997 0.2981 0.8657 0.3976 0.0758
239 0.8988 0.2988 0.8663 0.3918 0.0780
240 0.8992 0.2970 0.8663 0.3946 0.0774
241 0.8998 0.2970 0.8663 0.3930 0.0855
242 0.8973 0.2977 0.8679 0.3948 0.0761
243 0.8996 0.2962 0.8688 0.3967 0.0767
244 0.8996 0.2974 0.8688 0.3946 0.0781
245 0.8986 0.2975 0.8663 0.3942 0.0792
246 0.8981 0.2977 0.8675 0.3934 0.0770
247 0.8988 0.2976 0.8688 0.3952 0.0770
248 0.9006 0.2965 0.8639 0.4001 0.0764
249 0.8983 0.2973 0.8657 0.3963 0.1016
250 0.8983 0.2963 0.8666 0.3914 0.0783
251 0.9000 0.2977 0.8633 0.3990 0.0768
252 0.8993 0.2973 0.8657 0.3960 0.0775
253 0.8991 0.2957 0.8697 0.3962 0.0773
254 0.8983 0.2982 0.8672 0.3955 0.0758
255 0.8984 0.2959 0.8660 0.3955 0.0767
256 0.9005 0.2962 0.8651 0.4018 0.0772
257 0.9009 0.2952 0.8660 0.3930 0.1543
258 0.8998 0.2957 0.8663 0.3953 0.0828
259 0.9008 0.2953 0.8675 0.3936 0.0763
260 0.9002 0.2965 0.8657 0.3997 0.0772
261 0.8992 0.2961 0.8672 0.3960 0.0771
262 0.9000 0.2952 0.8672 0.3958 0.0769
263 0.9003 0.2946 0.8660 0.3935 0.0773
264 0.9009 0.2944 0.8648 0.3960 0.0764
265 0.9000 0.2940 0.8660 0.3960 0.0779
266 0.9000 0.2942 0.8654 0.3997 0.0777
267 0.9009 0.2968 0.8694 0.3953 0.0768
268 0.8995 0.2948 0.8672 0.3964 0.0760
269 0.8997 0.2967 0.8660 0.3976 0.0760
270 0.8997 0.2951 0.8660 0.4019 0.0776
271 0.8989 0.2969 0.8682 0.3956 0.0771
272 0.8982 0.2964 0.8657 0.4006 0.0765
273 0.8956 0.2981 0.8645 0.4028 0.0839
274 0.8972 0.2989 0.8648 0.3983 0.0773
275 0.8987 0.2967 0.8675 0.3986 0.0763
276 0.8979 0.2948 0.8648 0.3953 0.0777
277 0.8993 0.2938 0.8651 0.3989 0.0756
278 0.9003 0.2948 0.8663 0.3963 0.0772
279 0.9015 0.2929 0.8642 0.3974 0.0768
280 0.9006 0.2928 0.8651 0.3995 0.0777
281 0.9005 0.2946 0.8657 0.3992 0.0779
282 0.9000 0.2947 0.8657 0.3973 0.0772
283 0.8980 0.2950 0.8660 0.3980 0.0767
284 0.9015 0.2928 0.8639 0.3980 0.0764
285 0.9007 0.2934 0.8654 0.3959 0.0768
286 0.9010 0.2934 0.8669 0.3972 0.0767
287 0.9024 0.2934 0.8654 0.4025 0.0757
288 0.8991 0.2958 0.8672 0.3975 0.0779
289 0.8991 0.2942 0.8666 0.3975 0.0772
290 0.9006 0.2924 0.8657 0.3996 0.0856
291 0.9011 0.2933 0.8648 0.3980 0.0780
292 0.9004 0.2928 0.8663 0.3973 0.0771
293 0.9012 0.2929 0.8669 0.3968 0.0773
294 0.9008 0.2929 0.8657 0.3992 0.0774
295 0.8991 0.2938 0.8675 0.3951 0.1534
296 0.9010 0.2918 0.8663 0.3952 0.0777
297 0.9023 0.2920 0.8639 0.3987 0.0767
298 0.9007 0.2942 0.8648 0.4009 0.0764
299 0.9007 0.2934 0.8651 0.3960 0.0779
300 0.9011 0.2927 0.8663 0.3991 0.0775
301 0.9015 0.2939 0.8651 0.3970 0.0768
302 0.9006 0.2944 0.8636 0.3984 0.0774
303 0.8978 0.2936 0.8630 0.4042 0.0776
304 0.8991 0.2944 0.8685 0.3975 0.0776
305 0.9007 0.2934 0.8639 0.3994 0.1031
306 0.8994 0.2946 0.8663 0.3989 0.0787
307 0.9008 0.2919 0.8654 0.3975 0.0854
308 0.9021 0.2921 0.8651 0.3982 0.0764
309 0.9003 0.2922 0.8660 0.3999 0.0759
310 0.8998 0.2922 0.8660 0.3978 0.0766
311 0.9003 0.2940 0.8651 0.3990 0.0770
312 0.9003 0.2930 0.8663 0.4047 0.0777
313 0.8997 0.2934 0.8657 0.3974 0.0763
314 0.9002 0.2930 0.8663 0.3994 0.0761
315 0.9007 0.2923 0.8636 0.4008 0.0771
316 0.8998 0.2928 0.8648 0.3983 0.0766
317 0.9010 0.2919 0.8636 0.4011 0.0767
318 0.9009 0.2908 0.8675 0.3990 0.0769
319 0.9024 0.2909 0.8651 0.4004 0.0777
320 0.9020 0.2905 0.8682 0.3975 0.0768
321 0.9010 0.2915 0.8663 0.3968 0.0765
322 0.9002 0.2918 0.8666 0.3972 0.0762
323 0.9012 0.2901 0.8654 0.3996 0.0861
324 0.9015 0.2904 0.8651 0.3993 0.0760
325 0.9007 0.2922 0.8639 0.3984 0.0758
326 0.9011 0.2924 0.8663 0.3973 0.0765
327 0.9017 0.2913 0.8691 0.3979 0.0789
328 0.9012 0.2913 0.8627 0.4029 0.0776
329 0.9009 0.2916 0.8700 0.3996 0.0788
330 0.9003 0.2927 0.8624 0.4027 0.0759
331 0.9009 0.2910 0.8645 0.3982 0.0776
332 0.9000 0.2920 0.8663 0.3997 0.0765
333 0.9012 0.2896 0.8642 0.3993 0.0768
334 0.9014 0.2902 0.8663 0.3951 0.1542
335 0.9015 0.2907 0.8636 0.3990 0.0766
336 0.9006 0.2915 0.8651 0.3968 0.0779
337 0.9016 0.2905 0.8651 0.4017 0.0873
338 0.9008 0.2909 0.8666 0.3998 0.0780
339 0.9006 0.2919 0.8657 0.3972 0.0769
340 0.9006 0.2930 0.8645 0.4005 0.0763
341 0.8985 0.2913 0.8651 0.4029 0.0776
342 0.9010 0.2903 0.8657 0.3974 0.0778
343 0.9009 0.2901 0.8651 0.3996 0.0772
344 0.9004 0.2912 0.8651 0.4016 0.0756
345 0.9015 0.2922 0.8654 0.3979 0.0766
346 0.9000 0.2915 0.8688 0.3946 0.0768
347 0.9016 0.2901 0.8657 0.4002 0.0774
348 0.9021 0.2907 0.8612 0.4032 0.0781
349 0.8998 0.2914 0.8654 0.3973 0.0776
350 0.9012 0.2900 0.8663 0.3992 0.0848
351 0.9006 0.2920 0.8633 0.4031 0.0779
352 0.9018 0.2930 0.8657 0.4009 0.0774
353 0.9003 0.2921 0.8588 0.4055 0.0771
354 0.9007 0.2924 0.8645 0.3998 0.0771
355 0.9011 0.2922 0.8660 0.3996 0.0784
356 0.9000 0.2929 0.8642 0.4025 0.0769
357 0.8996 0.2932 0.8657 0.4018 0.0767
358 0.9009 0.2924 0.8666 0.3981 0.0771
359 0.9006 0.2901 0.8672 0.4050 0.1006
360 0.9000 0.2921 0.8660 0.3977 0.0777
361 0.9008 0.2898 0.8675 0.3982 0.0782
362 0.9011 0.2898 0.8648 0.4046 0.0765
363 0.9018 0.2917 0.8669 0.3988 0.0768
364 0.9014 0.2909 0.8675 0.4015 0.0777
365 0.9024 0.2896 0.8645 0.3987 0.0836
366 0.9027 0.2903 0.8682 0.4014 0.0759
367 0.9014 0.2913 0.8666 0.3996 0.0782
368 0.8997 0.2915 0.8657 0.4010 0.0775
369 0.9007 0.2902 0.8688 0.3960 0.0762
370 0.9016 0.2908 0.8642 0.4018 0.0775
371 0.9021 0.2892 0.8675 0.3951 0.0770
372 0.9013 0.2903 0.8606 0.4080 0.1530
373 0.9024 0.2913 0.8672 0.3999 0.0782
374 0.9015 0.2919 0.8639 0.3983 0.0767
375 0.8994 0.2908 0.8633 0.4039 0.0774
376 0.9001 0.2910 0.8663 0.3988 0.0778
377 0.9014 0.2900 0.8672 0.3991 0.0770
378 0.9018 0.2890 0.8703 0.3982 0.0764
379 0.9017 0.2891 0.8648 0.3997 0.0779
380 0.9032 0.2894 0.8645 0.4004 0.0767
381 0.9015 0.2901 0.8651 0.4007 0.0843
382 0.9021 0.2905 0.8672 0.3988 0.0769
383 0.9018 0.2885 0.8666 0.3967 0.0782
384 0.9014 0.2890 0.8669 0.3997 0.0770
385 0.9024 0.2899 0.8672 0.3998 0.0774
386 0.9025 0.2905 0.8657 0.4056 0.0766
387 0.9017 0.2901 0.8669 0.3958 0.0775
388 0.9023 0.2897 0.8666 0.4025 0.0773
389 0.9019 0.2891 0.8657 0.4026 0.0764
390 0.9000 0.2910 0.8694 0.3962 0.0762
391 0.9023 0.2900 0.8642 0.4018 0.0781
392 0.9021 0.2892 0.8672 0.4000 0.0765
393 0.9014 0.2895 0.8636 0.3991 0.0772
394 0.9024 0.2886 0.8633 0.4047 0.0767
395 0.9027 0.2899 0.8675 0.3982 0.0786
396 0.9020 0.2904 0.8666 0.4030 0.0770
397 0.9015 0.2904 0.8663 0.4000 0.0774
398 0.9012 0.2905 0.8642 0.4057 0.0871
399 0.9020 0.2917 0.8663 0.4000 0.0787
400 0.8998 0.2921 0.8657 0.4001 0.0783
401 0.9028 0.2906 0.8672 0.4014 0.0767
402 0.8997 0.2922 0.8679 0.3994 0.0773
403 0.9008 0.2906 0.8669 0.4009 0.0775
404 0.9013 0.2886 0.8663 0.4032 0.0754
405 0.9006 0.2890 0.8654 0.4027 0.0776
406 0.9020 0.2898 0.8648 0.4043 0.0768
407 0.9021 0.2897 0.8697 0.3979 0.0770
408 0.9019 0.2898 0.8639 0.4040 0.0765
409 0.9020 0.2897 0.8679 0.3981 0.0771
410 0.9010 0.2898 0.8645 0.4024 0.1551
411 0.9008 0.2920 0.8663 0.4029 0.0767
412 0.9003 0.2910 0.8691 0.4005 0.1025
413 0.9027 0.2887 0.8660 0.3998 0.0765
414 0.9031 0.2881 0.8648 0.3999 0.0771
415 0.9027 0.2887 0.8657 0.4001 0.0851
416 0.9019 0.2882 0.8660 0.4013 0.0779
417 0.9012 0.2886 0.8657 0.4041 0.0782
418 0.9033 0.2892 0.8679 0.4007 0.0776
419 0.9027 0.2893 0.8648 0.4004 0.0775
420 0.9030 0.2881 0.8700 0.3972 0.0771
421 0.9026 0.2882 0.8672 0.4006 0.0766
422 0.9030 0.2883 0.8700 0.4014 0.0775
423 0.9020 0.2894 0.8621 0.4061 0.0760
424 0.9017 0.2897 0.8691 0.3979 0.0770
425 0.9012 0.2887 0.8648 0.4062 0.0776
426 0.9016 0.2892 0.8654 0.4009 0.0782
427 0.9015 0.2882 0.8672 0.4009 0.0783
428 0.9009 0.2896 0.8660 0.4033 0.0768
429 0.8994 0.2914 0.8645 0.4003 0.0780
430 0.9016 0.2883 0.8651 0.4047 0.0873
431 0.9012 0.2894 0.8657 0.4018 0.0781
432 0.9031 0.2891 0.8660 0.4040 0.0764
433 0.9012 0.2911 0.8639 0.4039 0.0764
434 0.9004 0.2902 0.8669 0.4011 0.0772
435 0.9018 0.2892 0.8651 0.4043 0.0778
436 0.9018 0.2910 0.8672 0.4011 0.0767
437 0.9003 0.2911 0.8660 0.4010 0.0763
438 0.9018 0.2881 0.8660 0.3996 0.0780
439 0.9030 0.2877 0.8645 0.4045 0.0780
440 0.9020 0.2881 0.8679 0.3990 0.0765
441 0.9015 0.2877 0.8657 0.4045 0.0780
442 0.9030 0.2881 0.8688 0.3994 0.0764
443 0.9027 0.2885 0.8639 0.4039 0.0775
444 0.9009 0.2890 0.8675 0.4010 0.0888
445 0.9027 0.2875 0.8675 0.4006 0.0793
446 0.9025 0.2886 0.8639 0.4068 0.0797
447 0.9014 0.2900 0.8672 0.3998 0.0777
448 0.9021 0.2872 0.8654 0.4029 0.1539
449 0.9024 0.2888 0.8660 0.4011 0.0773
450 0.9023 0.2877 0.8645 0.4055 0.0766
451 0.9018 0.2894 0.8648 0.4006 0.0773
452 0.9016 0.2879 0.8654 0.4004 0.0778
453 0.9030 0.2875 0.8691 0.4021 0.0780
454 0.9009 0.2886 0.8694 0.3993 0.0779
455 0.9024 0.2890 0.8663 0.4020 0.0784
456 0.9024 0.2870 0.8666 0.4020 0.0772
457 0.9030 0.2880 0.8663 0.4011 0.0776
458 0.9032 0.2879 0.8685 0.4008 0.0776
459 0.9018 0.2878 0.8682 0.4010 0.0853
460 0.9021 0.2866 0.8682 0.4028 0.0759
461 0.9025 0.2871 0.8679 0.3985 0.0775
462 0.9022 0.2870 0.8682 0.4059 0.0767
463 0.9017 0.2881 0.8672 0.3995 0.1036
464 0.9016 0.2874 0.8654 0.4007 0.0760
465 0.9032 0.2871 0.8675 0.4003 0.0795
466 0.9017 0.2880 0.8657 0.4068 0.0770
467 0.9037 0.2883 0.8703 0.3984 0.0774
468 0.9033 0.2867 0.8651 0.4036 0.0763
469 0.9032 0.2877 0.8682 0.3991 0.0780
470 0.9015 0.2877 0.8666 0.4053 0.0785
471 0.9024 0.2884 0.8672 0.4041 0.0832
472 0.8991 0.2883 0.8685 0.3991 0.0821
473 0.9016 0.2877 0.8682 0.4047 0.0915
474 0.9028 0.2858 0.8666 0.4006 0.0775
475 0.9025 0.2860 0.8694 0.4015 0.0768
476 0.9037 0.2858 0.8675 0.4028 0.0777
477 0.9024 0.2861 0.8691 0.4020 0.0774
478 0.9030 0.2867 0.8685 0.4022 0.0770
479 0.9020 0.2863 0.8706 0.3996 0.0765
480 0.9017 0.2863 0.8679 0.4016 0.0781
481 0.9012 0.2874 0.8651 0.4038 0.0770
482 0.9031 0.2896 0.8672 0.4011 0.0763
483 0.9024 0.2874 0.8688 0.3997 0.0761
484 0.9022 0.2865 0.8651 0.4052 0.0780
485 0.9029 0.2865 0.8672 0.4020 0.0794
486 0.9019 0.2866 0.8688 0.4021 0.1723
487 0.9028 0.2857 0.8660 0.4028 0.0888
488 0.9024 0.2848 0.8682 0.4005 0.1041
489 0.9036 0.2859 0.8685 0.4008 0.0820
490 0.9033 0.2858 0.8654 0.4033 0.0803
491 0.9025 0.2865 0.8691 0.4015 0.0802
492 0.9017 0.2871 0.8636 0.4050 0.0810
493 0.9003 0.2866 0.8694 0.4032 0.0808
494 0.9027 0.2861 0.8675 0.4014 0.0909
495 0.9026 0.2855 0.8682 0.4017 0.0946
496 0.9024 0.2856 0.8679 0.4020 0.0869
497 0.9021 0.2853 0.8654 0.4040 0.0813
498 0.9036 0.2857 0.8679 0.4008 0.0895
499 0.9019 0.2850 0.8682 0.3992 0.1055
500 0.9017 0.2860 0.8660 0.4048 0.0791
Out[27]:
Pipeline(steps=[('standardscaler', StandardScaler()),
('neuralnetclassifier',
NeuralNetClassifier(_params_to_validate={'iterator_train__shuffle', 'module__out_features', 'module__in_features'}, batch_size=2048, callbacks=[<skorch.callbacks.scoring.EpochScoring object at 0x31b2738c0>], compile=False, dataset=<class 'skorch.dataset.Dataset'>, device='mps', iterator...ataLoader'>, iterator_train__shuffle=True, iterator_valid=<class 'torch.utils.data.dataloader.DataLoader'>, lr=0.01, max_epochs=500, module=<class '__main__.VanillaMLP'>, module__in_features=22, module__out_features=7, optimizer=<class 'torch.optim.adam.Adam'>, predict_nonlinearity='auto', torch_load_kwargs=None, use_caching='auto', verbose=1, warm_start=False))])In a Jupyter environment, please rerun this cell to show the HTML representation or trust the notebook. On GitHub, the HTML representation is unable to render, please try loading this page with nbviewer.org.
Pipeline(steps=[('standardscaler', StandardScaler()),
('neuralnetclassifier',
NeuralNetClassifier(_params_to_validate={'iterator_train__shuffle', 'module__out_features', 'module__in_features'}, batch_size=2048, callbacks=[<skorch.callbacks.scoring.EpochScoring object at 0x31b2738c0>], compile=False, dataset=<class 'skorch.dataset.Dataset'>, device='mps', iterator...ataLoader'>, iterator_train__shuffle=True, iterator_valid=<class 'torch.utils.data.dataloader.DataLoader'>, lr=0.01, max_epochs=500, module=<class '__main__.VanillaMLP'>, module__in_features=22, module__out_features=7, optimizer=<class 'torch.optim.adam.Adam'>, predict_nonlinearity='auto', torch_load_kwargs=None, use_caching='auto', verbose=1, warm_start=False))])StandardScaler()
<class 'skorch.classifier.NeuralNetClassifier'>[initialized](
module_=VanillaMLP(
(model): Sequential(
(0): Linear(in_features=22, out_features=32, bias=True)
(1): ReLU()
(2): Linear(in_features=32, out_features=7, bias=True)
(3): Softmax(dim=-1)
)
),
)In [28]:
y_pred = pipe.predict(X_test)
f1_score(y_test, y_pred, average="weighted")
Out[28]:
0.8655852188076297
In [29]:
history = pipe[1].history
plt.plot(history[:, 'train_loss'], label='train_loss')
plt.plot(history[:, 'valid_loss'], label='valid_loss')
plt.legend()
plt.xlabel('Epoch')
plt.ylabel('Loss')
plt.show()
plt.plot(history[:, 'train_acc'], label='train_acc')
plt.plot(history[:, 'valid_acc'], label='valid_acc')
plt.legend()
plt.xlabel('Epoch')
plt.ylabel('Accuracy')
plt.show()
Improving the model¶
In [30]:
class BetterMLP(nn.Module):
def __init__(self, in_features, out_features):
super().__init__()
self.model = nn.Sequential(
nn.Linear(in_features, 32),
nn.BatchNorm1d(32), # batch normalization layer
nn.ReLU(),
nn.Dropout(p=0.2), # dropout layer with 20% dropout rate
nn.Linear(32, out_features),
nn.Softmax(dim=-1),
)
def forward(self, X, **kwargs):
return self.model(X)
The better model, monitoring F1 besides Acc¶
In [31]:
net = NeuralNetClassifier(
BetterMLP,
module__in_features=X_train.shape[1],
module__out_features=len(le.classes_),
max_epochs=500,
lr = 1e-3,
device=DEVICE,
optimizer=torch.optim.Adam,
callbacks=[
EpochScoring(
scoring='f1_weighted',
name='train_f1',
on_train=True,
lower_is_better=False,
),
EpochScoring(
scoring='f1_weighted',
name='valid_f1',
on_train=False,
lower_is_better=False,
),
],
batch_size=2048,
iterator_train__shuffle=True,
)
net
Out[31]:
<class 'skorch.classifier.NeuralNetClassifier'>[uninitialized]( module=<class '__main__.BetterMLP'>, module__in_features=22, module__out_features=7, )In a Jupyter environment, please rerun this cell to show the HTML representation or trust the notebook.
On GitHub, the HTML representation is unable to render, please try loading this page with nbviewer.org.
<class 'skorch.classifier.NeuralNetClassifier'>[uninitialized]( module=<class '__main__.BetterMLP'>, module__in_features=22, module__out_features=7, )
In [32]:
# Training the network
net.fit(X_train, y_train)
epoch train_f1 train_loss valid_acc valid_f1 valid_loss dur
------- ---------- ------------ ----------- ---------- ------------ ------
1 0.0712 1.9186 0.1469 0.0382 2.5864 1.0286
2 0.0943 1.8163 0.1328 0.0501 1.9488 0.0847
3 0.1528 1.7282 0.2098 0.1403 1.7196 0.0836
4 0.2173 1.6527 0.3281 0.2249 1.6047 0.0888
5 0.2647 1.5878 0.3326 0.2299 1.5355 0.0932
6 0.3042 1.5298 0.3064 0.2185 1.4855 0.0900
7 0.3294 1.4790 0.3200 0.2436 1.4443 0.0861
8 0.3486 1.4336 0.3323 0.2344 1.4102 0.0828
9 0.3643 1.3945 0.3320 0.2237 1.3842 0.0830
10 0.3936 1.3611 0.3347 0.2252 1.3619 0.1050
11 0.4126 1.3320 0.4747 0.4102 1.3504 0.0813
12 0.4315 1.3053 0.5009 0.4346 1.3345 0.0813
13 0.4557 1.2771 0.4985 0.4444 1.3171 0.0826
14 0.4682 1.2542 0.5220 0.4866 1.3050 0.0809
15 0.4786 1.2305 0.5385 0.5085 1.2910 0.0823
16 0.5030 1.2034 0.5590 0.5370 1.2734 0.0812
17 0.5225 1.1722 0.5695 0.5692 1.2896 0.0809
18 0.5392 1.1466 0.5713 0.5582 1.2856 0.0806
19 0.5631 1.1149 0.5641 0.5459 1.2802 0.0806
20 0.5821 1.0827 0.4603 0.4243 1.2996 0.0809
21 0.5977 1.0529 0.5500 0.5442 1.2446 0.1603
22 0.6119 1.0215 0.6219 0.6173 1.1837 0.0887
23 0.6215 0.9950 0.5834 0.5702 1.1501 0.0928
24 0.6209 0.9686 0.6285 0.6208 1.0946 0.0912
25 0.6227 0.9507 0.6343 0.6233 1.0672 0.0852
26 0.6410 0.9262 0.6445 0.6343 1.0064 0.0811
27 0.6531 0.9077 0.6731 0.6671 0.9535 0.0806
28 0.6574 0.8922 0.6728 0.6698 0.9340 0.0814
29 0.6613 0.8756 0.6833 0.6767 0.8976 0.0850
30 0.6690 0.8610 0.6824 0.6762 0.8837 0.0925
31 0.6753 0.8452 0.7056 0.6972 0.8494 0.0929
32 0.6807 0.8303 0.7065 0.6973 0.8240 0.0859
33 0.6906 0.8160 0.7125 0.7033 0.8228 0.1072
34 0.6949 0.8090 0.7291 0.7209 0.7874 0.0848
35 0.6975 0.7970 0.7101 0.7005 0.8011 0.0811
36 0.7001 0.7859 0.7267 0.7205 0.7730 0.0810
37 0.7076 0.7743 0.7363 0.7281 0.7564 0.0815
38 0.7054 0.7666 0.7423 0.7346 0.7403 0.0812
39 0.7072 0.7589 0.7134 0.7037 0.7646 0.0822
40 0.7181 0.7498 0.7523 0.7462 0.7148 0.0805
41 0.7224 0.7402 0.7574 0.7523 0.7191 0.0804
42 0.7207 0.7321 0.7502 0.7438 0.7136 0.0831
43 0.7233 0.7233 0.7622 0.7570 0.6978 0.0812
44 0.7292 0.7164 0.7417 0.7323 0.7011 0.0802
45 0.7350 0.7074 0.7577 0.7521 0.6789 0.0840
46 0.7339 0.6997 0.7294 0.7201 0.7124 0.0813
47 0.7367 0.6935 0.7514 0.7423 0.6874 0.0805
48 0.7361 0.6888 0.7538 0.7457 0.6762 0.0802
49 0.7415 0.6860 0.7724 0.7673 0.6590 0.0814
50 0.7394 0.6784 0.7778 0.7731 0.6470 0.0811
51 0.7474 0.6681 0.7426 0.7334 0.6808 0.0819
52 0.7487 0.6629 0.7902 0.7855 0.6277 0.0800
53 0.7498 0.6603 0.7658 0.7582 0.6508 0.0807
54 0.7537 0.6546 0.7769 0.7713 0.6302 0.0808
55 0.7523 0.6544 0.7763 0.7707 0.6365 0.0820
56 0.7578 0.6446 0.7914 0.7869 0.6220 0.0806
57 0.7573 0.6378 0.7938 0.7903 0.6125 0.0806
58 0.7612 0.6376 0.7652 0.7595 0.6545 0.0816
59 0.7668 0.6323 0.8001 0.7958 0.5906 0.1588
60 0.7625 0.6313 0.7414 0.7335 0.6735 0.0809
61 0.7674 0.6218 0.7995 0.7951 0.5917 0.0812
62 0.7721 0.6172 0.7872 0.7839 0.6191 0.0824
63 0.7724 0.6131 0.7929 0.7895 0.6042 0.0814
64 0.7751 0.6081 0.7878 0.7826 0.6015 0.0820
65 0.7760 0.6032 0.8061 0.8021 0.5815 0.0821
66 0.7756 0.6032 0.7923 0.7853 0.6069 0.0821
67 0.7751 0.5963 0.7812 0.7761 0.6047 0.0816
68 0.7829 0.5932 0.8019 0.8000 0.5889 0.0808
69 0.7790 0.5911 0.7727 0.7691 0.6193 0.0812
70 0.7862 0.5850 0.8230 0.8203 0.5487 0.0809
71 0.7822 0.5836 0.8082 0.8052 0.5704 0.0803
72 0.7911 0.5785 0.7842 0.7797 0.6094 0.0825
73 0.7899 0.5758 0.8245 0.8214 0.5420 0.0816
74 0.7868 0.5770 0.8185 0.8159 0.5604 0.0789
75 0.7895 0.5697 0.8206 0.8181 0.5480 0.0860
76 0.7923 0.5706 0.8128 0.8102 0.5634 0.0814
77 0.7936 0.5661 0.8311 0.8297 0.5346 0.0904
78 0.7995 0.5589 0.8341 0.8328 0.5186 0.0938
79 0.7920 0.5597 0.7793 0.7728 0.5891 0.1328
80 0.7912 0.5604 0.8278 0.8272 0.5256 0.0863
81 0.7964 0.5566 0.8061 0.8044 0.5661 0.0939
82 0.7974 0.5536 0.8197 0.8163 0.5275 0.0861
83 0.7997 0.5518 0.8287 0.8292 0.5318 0.0851
84 0.8004 0.5459 0.8137 0.8102 0.5372 0.0883
85 0.8013 0.5461 0.8263 0.8226 0.5266 0.0942
86 0.7997 0.5411 0.8368 0.8357 0.5000 0.0946
87 0.8027 0.5379 0.8007 0.7972 0.5602 0.0929
88 0.8015 0.5430 0.8396 0.8393 0.4954 0.0832
89 0.8018 0.5373 0.8293 0.8263 0.5091 0.0857
90 0.8044 0.5366 0.8380 0.8373 0.5040 0.0967
91 0.8041 0.5349 0.8359 0.8367 0.5184 0.0924
92 0.8047 0.5367 0.8149 0.8097 0.5239 0.0866
93 0.8050 0.5318 0.8365 0.8342 0.5006 0.0878
94 0.8014 0.5276 0.8356 0.8365 0.4984 0.0847
95 0.8082 0.5313 0.8248 0.8217 0.5127 0.0855
96 0.8070 0.5255 0.8365 0.8358 0.4949 0.0861
97 0.8036 0.5283 0.8417 0.8411 0.4951 0.0940
98 0.8035 0.5278 0.8155 0.8097 0.5187 0.1626
99 0.8076 0.5202 0.8377 0.8375 0.5017 0.0831
100 0.8118 0.5163 0.8182 0.8150 0.5329 0.0839
101 0.8060 0.5221 0.8317 0.8285 0.4920 0.0834
102 0.8117 0.5167 0.8377 0.8364 0.4969 0.0870
103 0.8117 0.5134 0.8438 0.8422 0.4739 0.0906
104 0.8080 0.5142 0.8444 0.8430 0.4788 0.0867
105 0.8122 0.5132 0.8206 0.8173 0.5046 0.0861
106 0.8138 0.5119 0.8414 0.8397 0.4830 0.0847
107 0.8175 0.5063 0.8417 0.8401 0.4845 0.0834
108 0.8136 0.5093 0.8245 0.8198 0.4978 0.1054
109 0.8174 0.5054 0.8486 0.8482 0.4658 0.0846
110 0.8157 0.5019 0.8450 0.8432 0.4791 0.0832
111 0.8168 0.5035 0.8453 0.8453 0.4771 0.0926
112 0.8181 0.5059 0.8489 0.8479 0.4683 0.0928
113 0.8161 0.5023 0.8317 0.8286 0.4819 0.0955
114 0.8156 0.5005 0.8387 0.8365 0.4771 0.0872
115 0.8200 0.5004 0.8489 0.8484 0.4759 0.0870
116 0.8188 0.4971 0.8329 0.8302 0.4866 0.0846
117 0.8174 0.4985 0.8393 0.8363 0.4831 0.0863
118 0.8187 0.4941 0.8480 0.8482 0.4801 0.0884
119 0.8177 0.4959 0.8480 0.8467 0.4670 0.0919
120 0.8202 0.4925 0.8347 0.8313 0.4888 0.0923
121 0.8183 0.4944 0.8519 0.8515 0.4618 0.0894
122 0.8223 0.4914 0.8332 0.8301 0.4882 0.0868
123 0.8249 0.4910 0.8302 0.8313 0.4865 0.0935
124 0.8226 0.4923 0.8380 0.8346 0.4672 0.0855
125 0.8251 0.4897 0.8414 0.8391 0.4798 0.0857
126 0.8262 0.4874 0.8483 0.8485 0.4606 0.0858
127 0.8232 0.4853 0.8356 0.8320 0.4805 0.0845
128 0.8259 0.4853 0.8462 0.8447 0.4610 0.0835
129 0.8225 0.4858 0.8447 0.8438 0.4669 0.0875
130 0.8235 0.4828 0.8501 0.8489 0.4543 0.1232
131 0.8241 0.4809 0.8531 0.8526 0.4560 0.1016
132 0.8253 0.4813 0.8540 0.8534 0.4696 0.0936
133 0.8237 0.4817 0.8438 0.8408 0.4528 0.0982
134 0.8294 0.4823 0.8498 0.8493 0.4511 0.0981
135 0.8272 0.4770 0.8417 0.8397 0.4671 0.1164
136 0.8228 0.4830 0.8504 0.8507 0.4600 0.1731
137 0.8263 0.4797 0.8426 0.8394 0.4648 0.0913
138 0.8267 0.4766 0.8417 0.8396 0.4617 0.0946
139 0.8273 0.4762 0.8510 0.8495 0.4587 0.0958
140 0.8296 0.4730 0.8465 0.8447 0.4582 0.0874
141 0.8306 0.4731 0.8558 0.8548 0.4440 0.0846
142 0.8245 0.4775 0.8438 0.8421 0.4639 0.0852
143 0.8287 0.4718 0.8516 0.8509 0.4571 0.0856
144 0.8280 0.4699 0.8531 0.8525 0.4371 0.0809
145 0.8299 0.4703 0.8405 0.8377 0.4629 0.0815
146 0.8301 0.4706 0.8519 0.8500 0.4408 0.0826
147 0.8271 0.4689 0.8432 0.8403 0.4527 0.0804
148 0.8334 0.4673 0.8549 0.8543 0.4486 0.0873
149 0.8235 0.4742 0.8540 0.8529 0.4395 0.0930
150 0.8278 0.4715 0.8567 0.8563 0.4288 0.1243
151 0.8300 0.4690 0.8507 0.8500 0.4365 0.0918
152 0.8327 0.4676 0.8561 0.8548 0.4364 0.0945
153 0.8315 0.4667 0.8567 0.8563 0.4367 0.0952
154 0.8320 0.4622 0.8474 0.8443 0.4511 0.0927
155 0.8324 0.4636 0.8402 0.8419 0.4629 0.1425
156 0.8250 0.4656 0.8314 0.8270 0.4809 0.0965
157 0.8368 0.4593 0.8609 0.8602 0.4368 0.1002
158 0.8354 0.4606 0.8582 0.8568 0.4381 0.0908
159 0.8342 0.4615 0.8516 0.8512 0.4258 0.0892
160 0.8326 0.4587 0.8588 0.8582 0.4376 0.0847
161 0.8311 0.4607 0.8564 0.8561 0.4283 0.0828
162 0.8350 0.4585 0.8534 0.8525 0.4414 0.0868
163 0.8348 0.4584 0.8555 0.8535 0.4432 0.0884
164 0.8379 0.4555 0.8627 0.8627 0.4345 0.0932
165 0.8366 0.4542 0.8600 0.8602 0.4302 0.0936
166 0.8389 0.4550 0.8519 0.8495 0.4420 0.1078
167 0.8339 0.4590 0.8519 0.8511 0.4261 0.0941
168 0.8348 0.4588 0.8513 0.8494 0.4443 0.0920
169 0.8362 0.4534 0.8477 0.8466 0.4382 0.0854
170 0.8380 0.4545 0.8573 0.8561 0.4319 0.0891
171 0.8379 0.4501 0.8522 0.8518 0.4267 0.0911
172 0.8353 0.4550 0.8546 0.8531 0.4286 0.0968
173 0.8363 0.4524 0.8546 0.8526 0.4418 0.1333
174 0.8364 0.4525 0.8531 0.8536 0.4327 0.1647
175 0.8339 0.4542 0.8546 0.8526 0.4372 0.0878
176 0.8361 0.4494 0.8456 0.8425 0.4509 0.0855
177 0.8354 0.4598 0.8486 0.8486 0.4366 0.0848
178 0.8359 0.4512 0.8594 0.8590 0.4147 0.0909
179 0.8369 0.4532 0.8534 0.8510 0.4288 0.0835
180 0.8386 0.4489 0.8540 0.8539 0.4257 0.0830
181 0.8375 0.4494 0.8558 0.8544 0.4346 0.0807
182 0.8380 0.4530 0.8320 0.8322 0.4577 0.0833
183 0.8401 0.4497 0.8588 0.8573 0.4325 0.0827
184 0.8396 0.4449 0.8552 0.8529 0.4309 0.0828
185 0.8378 0.4492 0.8540 0.8534 0.4327 0.0872
186 0.8398 0.4467 0.8615 0.8598 0.4356 0.0830
187 0.8402 0.4449 0.8528 0.8509 0.4375 0.0841
188 0.8370 0.4523 0.8299 0.8329 0.4643 0.0848
189 0.8395 0.4458 0.8377 0.8336 0.4762 0.0846
190 0.8426 0.4440 0.8540 0.8517 0.4324 0.0903
191 0.8426 0.4403 0.8558 0.8570 0.4270 0.0862
192 0.8371 0.4423 0.8507 0.8476 0.4372 0.0858
193 0.8405 0.4468 0.8468 0.8441 0.4496 0.0901
194 0.8422 0.4439 0.8543 0.8544 0.4308 0.0823
195 0.8416 0.4447 0.8483 0.8457 0.4384 0.0847
196 0.8392 0.4471 0.8564 0.8578 0.4288 0.0841
197 0.8399 0.4408 0.8555 0.8545 0.4257 0.0833
198 0.8410 0.4439 0.8597 0.8578 0.4215 0.0831
199 0.8410 0.4442 0.8582 0.8571 0.4176 0.0842
200 0.8428 0.4417 0.8546 0.8546 0.4225 0.0838
201 0.8404 0.4414 0.8540 0.8518 0.4275 0.0834
202 0.8433 0.4432 0.8543 0.8551 0.4235 0.0823
203 0.8428 0.4372 0.8561 0.8541 0.4314 0.0847
204 0.8405 0.4399 0.8570 0.8551 0.4253 0.0869
205 0.8427 0.4375 0.8543 0.8516 0.4245 0.0832
206 0.8426 0.4345 0.8642 0.8645 0.4138 0.0839
207 0.8448 0.4354 0.8615 0.8609 0.4102 0.0832
208 0.8423 0.4372 0.8642 0.8634 0.4215 0.0924
209 0.8455 0.4377 0.8549 0.8538 0.4230 0.0825
210 0.8425 0.4388 0.8540 0.8530 0.4152 0.0842
211 0.8409 0.4392 0.8576 0.8579 0.4167 0.0827
212 0.8401 0.4370 0.8540 0.8521 0.4218 0.1600
213 0.8422 0.4368 0.8399 0.8393 0.4339 0.0826
214 0.8427 0.4343 0.8600 0.8588 0.4182 0.0824
215 0.8447 0.4346 0.8576 0.8581 0.4142 0.0839
216 0.8437 0.4338 0.8353 0.8284 0.4727 0.1080
217 0.8400 0.4366 0.8362 0.8373 0.4496 0.0829
218 0.8460 0.4340 0.8507 0.8468 0.4464 0.0828
219 0.8446 0.4331 0.8621 0.8621 0.4116 0.0831
220 0.8427 0.4372 0.8573 0.8547 0.4336 0.0836
221 0.8476 0.4333 0.8441 0.8446 0.4298 0.0873
222 0.8459 0.4333 0.8597 0.8605 0.4328 0.0915
223 0.8445 0.4334 0.8423 0.8390 0.4524 0.0982
224 0.8482 0.4346 0.8618 0.8611 0.4089 0.0835
225 0.8466 0.4352 0.8567 0.8548 0.4342 0.0852
226 0.8477 0.4307 0.8522 0.8499 0.4281 0.0829
227 0.8444 0.4306 0.8600 0.8604 0.4097 0.0829
228 0.8444 0.4297 0.8320 0.8288 0.4804 0.0831
229 0.8422 0.4368 0.8534 0.8525 0.4203 0.0830
230 0.8437 0.4352 0.8597 0.8577 0.4290 0.0843
231 0.8452 0.4306 0.8522 0.8538 0.4243 0.0824
232 0.8442 0.4301 0.7965 0.7865 0.5295 0.0868
233 0.8487 0.4319 0.8621 0.8626 0.4179 0.0909
234 0.8443 0.4307 0.8522 0.8524 0.4129 0.0933
235 0.8434 0.4321 0.8591 0.8570 0.4147 0.0845
236 0.8452 0.4280 0.8594 0.8592 0.4146 0.0870
237 0.8460 0.4290 0.8525 0.8535 0.4169 0.0847
238 0.8467 0.4311 0.8528 0.8498 0.4311 0.0817
239 0.8430 0.4335 0.8621 0.8620 0.4134 0.0817
240 0.8467 0.4271 0.8621 0.8608 0.4109 0.0820
241 0.8451 0.4252 0.8579 0.8585 0.4223 0.0831
242 0.8470 0.4266 0.8672 0.8660 0.4098 0.0839
243 0.8440 0.4309 0.8621 0.8608 0.4073 0.0908
244 0.8505 0.4262 0.8612 0.8605 0.4064 0.1305
245 0.8473 0.4288 0.8579 0.8577 0.4202 0.0964
246 0.8460 0.4277 0.8615 0.8599 0.4105 0.0840
247 0.8524 0.4263 0.8492 0.8466 0.4271 0.0821
248 0.8495 0.4241 0.8639 0.8637 0.4014 0.0837
249 0.8488 0.4277 0.8639 0.8635 0.4043 0.0826
250 0.8496 0.4269 0.8522 0.8526 0.4141 0.0818
251 0.8494 0.4251 0.8447 0.8403 0.4503 0.1661
252 0.8466 0.4294 0.8654 0.8652 0.4065 0.0886
253 0.8528 0.4225 0.8507 0.8482 0.4374 0.0846
254 0.8447 0.4283 0.8525 0.8509 0.4184 0.0828
255 0.8497 0.4228 0.8483 0.8492 0.4177 0.0830
256 0.8485 0.4264 0.8669 0.8661 0.4065 0.0845
257 0.8485 0.4272 0.8561 0.8547 0.4089 0.0879
258 0.8494 0.4215 0.8633 0.8631 0.4051 0.0856
259 0.8470 0.4233 0.8414 0.8376 0.4590 0.0814
260 0.8458 0.4301 0.8423 0.8434 0.4251 0.0821
261 0.8523 0.4190 0.8564 0.8532 0.4227 0.0830
262 0.8496 0.4204 0.8666 0.8655 0.4030 0.0840
263 0.8484 0.4211 0.8636 0.8638 0.3999 0.0828
264 0.8519 0.4205 0.8642 0.8627 0.4245 0.0828
265 0.8510 0.4254 0.8504 0.8492 0.4244 0.0824
266 0.8521 0.4253 0.8633 0.8618 0.4188 0.0821
267 0.8537 0.4240 0.8483 0.8491 0.4225 0.0829
268 0.8528 0.4226 0.8534 0.8503 0.4358 0.0825
269 0.8521 0.4228 0.8624 0.8637 0.4096 0.0882
270 0.8515 0.4225 0.8609 0.8603 0.4087 0.0820
271 0.8504 0.4249 0.8570 0.8542 0.4244 0.0847
272 0.8457 0.4219 0.8576 0.8568 0.4111 0.0830
273 0.8517 0.4186 0.8682 0.8686 0.4000 0.0832
274 0.8509 0.4206 0.8615 0.8610 0.4061 0.0815
275 0.8498 0.4182 0.8519 0.8488 0.4221 0.0829
276 0.8568 0.4169 0.8627 0.8625 0.4078 0.0821
277 0.8531 0.4180 0.8600 0.8593 0.4075 0.0808
278 0.8513 0.4191 0.8549 0.8535 0.4148 0.0830
279 0.8506 0.4216 0.8630 0.8618 0.4007 0.0816
280 0.8525 0.4185 0.8636 0.8635 0.4002 0.0821
281 0.8532 0.4193 0.8648 0.8639 0.4046 0.0829
282 0.8533 0.4145 0.8639 0.8639 0.4119 0.0905
283 0.8515 0.4212 0.8600 0.8576 0.4202 0.0831
284 0.8505 0.4195 0.8645 0.8645 0.4058 0.1092
285 0.8519 0.4198 0.8645 0.8639 0.3999 0.0820
286 0.8525 0.4189 0.8546 0.8542 0.4075 0.0807
287 0.8553 0.4172 0.8498 0.8459 0.4434 0.0829
288 0.8544 0.4164 0.8525 0.8525 0.4226 0.0819
289 0.8481 0.4233 0.8170 0.8140 0.5024 0.1625
290 0.8495 0.4188 0.8459 0.8464 0.4227 0.0823
291 0.8526 0.4202 0.8639 0.8629 0.4045 0.0819
292 0.8491 0.4235 0.8651 0.8642 0.4037 0.0834
293 0.8556 0.4177 0.8627 0.8617 0.4079 0.0827
294 0.8500 0.4216 0.8564 0.8541 0.4245 0.0823
295 0.8507 0.4144 0.8423 0.8419 0.4247 0.0911
296 0.8515 0.4222 0.8672 0.8675 0.4100 0.0845
297 0.8553 0.4112 0.8492 0.8483 0.4199 0.0837
298 0.8519 0.4158 0.8582 0.8570 0.4089 0.0840
299 0.8537 0.4117 0.8591 0.8579 0.4344 0.0820
300 0.8517 0.4190 0.8576 0.8552 0.4331 0.0826
301 0.8539 0.4154 0.8555 0.8555 0.4076 0.0827
302 0.8534 0.4144 0.8549 0.8527 0.4271 0.0850
303 0.8563 0.4136 0.8438 0.8456 0.4192 0.0834
304 0.8553 0.4147 0.8591 0.8572 0.4143 0.0836
305 0.8523 0.4138 0.8609 0.8606 0.4092 0.0821
306 0.8526 0.4158 0.8582 0.8577 0.4056 0.0827
307 0.8540 0.4145 0.8660 0.8652 0.4075 0.0909
308 0.8559 0.4110 0.8651 0.8646 0.4031 0.0826
309 0.8555 0.4117 0.8582 0.8578 0.4025 0.0829
310 0.8582 0.4117 0.8660 0.8656 0.3964 0.0828
311 0.8566 0.4121 0.8549 0.8533 0.4177 0.0837
312 0.8539 0.4138 0.8612 0.8608 0.4041 0.0817
313 0.8541 0.4100 0.8558 0.8563 0.4162 0.0837
314 0.8522 0.4132 0.8606 0.8591 0.4138 0.0834
315 0.8532 0.4123 0.8591 0.8583 0.4036 0.0826
316 0.8533 0.4166 0.8498 0.8507 0.4123 0.0838
317 0.8513 0.4159 0.8651 0.8637 0.4058 0.0886
318 0.8526 0.4146 0.8645 0.8639 0.3942 0.0857
319 0.8564 0.4111 0.8591 0.8576 0.4082 0.0833
320 0.8569 0.4094 0.8546 0.8558 0.4061 0.0839
321 0.8555 0.4105 0.8612 0.8574 0.4259 0.0856
322 0.8544 0.4133 0.8453 0.8431 0.4383 0.0839
323 0.8567 0.4149 0.8675 0.8673 0.4034 0.0924
324 0.8525 0.4129 0.8642 0.8626 0.4079 0.0822
325 0.8565 0.4081 0.8630 0.8626 0.4035 0.0845
326 0.8567 0.4068 0.8654 0.8633 0.4154 0.0840
327 0.8535 0.4162 0.8540 0.8540 0.4129 0.0831
328 0.8589 0.4089 0.8660 0.8659 0.4039 0.1603
329 0.8578 0.4115 0.8609 0.8595 0.4071 0.0837
330 0.8551 0.4095 0.8633 0.8624 0.3972 0.0820
331 0.8554 0.4092 0.8585 0.8572 0.4092 0.0833
332 0.8553 0.4141 0.8688 0.8678 0.4013 0.0850
333 0.8546 0.4099 0.8648 0.8626 0.4108 0.0875
334 0.8512 0.4165 0.8456 0.8445 0.4257 0.1248
335 0.8542 0.4112 0.8633 0.8640 0.4142 0.0915
336 0.8575 0.4091 0.8576 0.8579 0.4091 0.0856
337 0.8572 0.4066 0.8474 0.8452 0.4287 0.0849
338 0.8550 0.4130 0.8618 0.8607 0.4003 0.0877
339 0.8563 0.4098 0.8600 0.8596 0.3976 0.0873
340 0.8565 0.4098 0.8633 0.8625 0.4002 0.0872
341 0.8552 0.4107 0.8621 0.8612 0.4018 0.0874
342 0.8532 0.4142 0.8600 0.8599 0.4157 0.0844
343 0.8540 0.4155 0.8636 0.8618 0.4073 0.0848
344 0.8562 0.4135 0.8459 0.8454 0.4246 0.0860
345 0.8548 0.4140 0.8543 0.8512 0.4246 0.0849
346 0.8550 0.4122 0.8570 0.8570 0.4107 0.0834
347 0.8598 0.4085 0.8609 0.8588 0.4146 0.0904
348 0.8541 0.4089 0.8636 0.8635 0.3911 0.0883
349 0.8586 0.4073 0.8567 0.8547 0.4175 0.0885
350 0.8573 0.4078 0.8675 0.8670 0.3998 0.0878
351 0.8566 0.4086 0.8597 0.8600 0.4011 0.0845
352 0.8554 0.4104 0.8486 0.8474 0.4186 0.0828
353 0.8564 0.4084 0.8645 0.8635 0.4156 0.0847
354 0.8602 0.4034 0.8675 0.8663 0.3965 0.0820
355 0.8588 0.4027 0.8528 0.8514 0.4131 0.0820
356 0.8574 0.4106 0.8576 0.8586 0.4101 0.0838
357 0.8554 0.4075 0.8603 0.8590 0.4106 0.0846
358 0.8572 0.4091 0.8552 0.8540 0.4259 0.0859
359 0.8551 0.4103 0.8543 0.8508 0.4234 0.0829
360 0.8549 0.4102 0.8546 0.8550 0.4153 0.0935
361 0.8575 0.4058 0.8597 0.8576 0.4268 0.0871
362 0.8585 0.4050 0.8615 0.8620 0.4032 0.0846
363 0.8564 0.4055 0.8582 0.8574 0.4039 0.0845
364 0.8608 0.4062 0.8648 0.8636 0.4003 0.0855
365 0.8547 0.4130 0.8606 0.8585 0.4135 0.0829
366 0.8571 0.4069 0.8645 0.8634 0.4030 0.1621
367 0.8568 0.4102 0.8618 0.8599 0.4051 0.0837
368 0.8576 0.4080 0.8567 0.8572 0.4062 0.0834
369 0.8560 0.4065 0.8573 0.8582 0.4015 0.0829
370 0.8605 0.4037 0.8645 0.8630 0.4108 0.0824
371 0.8611 0.4040 0.8660 0.8665 0.4002 0.0843
372 0.8569 0.4065 0.8660 0.8647 0.4127 0.0908
373 0.8593 0.4030 0.8450 0.8438 0.4191 0.0832
374 0.8574 0.4105 0.8609 0.8594 0.4158 0.0823
375 0.8581 0.4036 0.8642 0.8632 0.3964 0.0812
376 0.8583 0.4041 0.8510 0.8515 0.4161 0.0834
377 0.8594 0.4069 0.8564 0.8557 0.4048 0.0837
378 0.8566 0.4111 0.8615 0.8600 0.4077 0.0836
379 0.8610 0.3995 0.8688 0.8690 0.3972 0.0825
380 0.8578 0.4060 0.8621 0.8600 0.4130 0.0815
381 0.8593 0.4052 0.8606 0.8611 0.3950 0.1083
382 0.8574 0.4043 0.8639 0.8632 0.3952 0.0820
383 0.8583 0.4061 0.8588 0.8577 0.4104 0.0831
384 0.8597 0.4065 0.8663 0.8667 0.3927 0.0833
385 0.8604 0.4037 0.8480 0.8466 0.4144 0.0822
386 0.8598 0.4013 0.8633 0.8623 0.4139 0.0898
387 0.8586 0.4058 0.8618 0.8613 0.4112 0.0827
388 0.8596 0.4054 0.8645 0.8639 0.3947 0.0829
389 0.8577 0.4025 0.8624 0.8608 0.4041 0.0824
390 0.8582 0.4024 0.8534 0.8535 0.4011 0.0835
391 0.8568 0.4031 0.8576 0.8586 0.3973 0.0829
392 0.8596 0.4020 0.8639 0.8632 0.4002 0.0817
393 0.8575 0.4063 0.8633 0.8617 0.4064 0.0839
394 0.8595 0.4007 0.8591 0.8591 0.4075 0.0823
395 0.8591 0.4018 0.8612 0.8591 0.4157 0.0824
396 0.8543 0.4099 0.8639 0.8636 0.3914 0.0839
397 0.8564 0.4102 0.8636 0.8635 0.4001 0.0819
398 0.8628 0.4023 0.8576 0.8574 0.3985 0.0862
399 0.8615 0.3993 0.8615 0.8605 0.4022 0.0971
400 0.8606 0.3990 0.8570 0.8567 0.4049 0.0856
401 0.8637 0.3972 0.8633 0.8614 0.4075 0.0847
402 0.8591 0.4022 0.8642 0.8636 0.4089 0.0870
403 0.8610 0.3986 0.8555 0.8548 0.4068 0.0856
404 0.8610 0.3983 0.8573 0.8577 0.4037 0.0827
405 0.8605 0.4017 0.8624 0.8602 0.4178 0.1566
406 0.8592 0.4028 0.8588 0.8581 0.4165 0.0835
407 0.8555 0.4088 0.8591 0.8583 0.3936 0.0833
408 0.8599 0.4029 0.8591 0.8580 0.4033 0.0832
409 0.8566 0.4016 0.8603 0.8613 0.4040 0.0839
410 0.8652 0.4007 0.8639 0.8636 0.4064 0.0838
411 0.8612 0.4008 0.8627 0.8617 0.4084 0.0828
412 0.8586 0.4024 0.8585 0.8570 0.4028 0.0884
413 0.8578 0.4067 0.8669 0.8659 0.3970 0.0880
414 0.8608 0.3994 0.8639 0.8645 0.3920 0.0963
415 0.8588 0.4010 0.8666 0.8656 0.4027 0.0830
416 0.8592 0.4033 0.8564 0.8561 0.3994 0.0844
417 0.8623 0.4041 0.8579 0.8577 0.3965 0.0833
418 0.8597 0.4014 0.8679 0.8666 0.4004 0.0826
419 0.8607 0.4025 0.8627 0.8624 0.3901 0.0835
420 0.8613 0.4011 0.8606 0.8609 0.3959 0.0812
421 0.8591 0.4017 0.8648 0.8640 0.3888 0.0831
422 0.8628 0.4006 0.8549 0.8541 0.4095 0.0836
423 0.8592 0.3988 0.8621 0.8613 0.4089 0.0816
424 0.8592 0.3989 0.8639 0.8632 0.3931 0.0831
425 0.8547 0.4058 0.8495 0.8483 0.4174 0.0837
426 0.8570 0.4062 0.8660 0.8652 0.4018 0.0818
427 0.8613 0.4059 0.8597 0.8599 0.3979 0.0834
428 0.8628 0.3975 0.8660 0.8649 0.3945 0.0823
429 0.8581 0.3995 0.8618 0.8607 0.4012 0.0893
430 0.8601 0.4056 0.8672 0.8674 0.3944 0.1127
431 0.8617 0.4005 0.8555 0.8542 0.4125 0.0826
432 0.8589 0.4011 0.8615 0.8616 0.3981 0.0819
433 0.8599 0.4006 0.8660 0.8658 0.3882 0.0827
434 0.8610 0.4012 0.8648 0.8647 0.3926 0.0818
435 0.8611 0.3999 0.8549 0.8545 0.3989 0.0821
436 0.8632 0.3910 0.8633 0.8634 0.4096 0.0840
437 0.8588 0.4012 0.8709 0.8697 0.3945 0.0902
438 0.8598 0.4027 0.8549 0.8540 0.4049 0.0912
439 0.8606 0.4001 0.8570 0.8549 0.4179 0.0881
440 0.8619 0.3943 0.8657 0.8664 0.3995 0.0858
441 0.8614 0.3976 0.8591 0.8578 0.4091 0.0864
442 0.8597 0.4032 0.8669 0.8662 0.4058 0.0873
443 0.8558 0.4044 0.8579 0.8571 0.4033 0.1765
444 0.8601 0.3971 0.8329 0.8317 0.4373 0.0829
445 0.8626 0.3983 0.8688 0.8682 0.4061 0.0830
446 0.8602 0.4001 0.8477 0.8463 0.4385 0.0834
447 0.8590 0.3973 0.8477 0.8467 0.4211 0.0830
448 0.8643 0.3960 0.8630 0.8622 0.4104 0.0817
449 0.8577 0.3992 0.8675 0.8675 0.3981 0.0824
450 0.8595 0.4043 0.8600 0.8586 0.3984 0.0833
451 0.8619 0.3963 0.8564 0.8554 0.4036 0.0827
452 0.8650 0.3961 0.8609 0.8601 0.3999 0.0824
453 0.8613 0.3947 0.8679 0.8684 0.3945 0.0890
454 0.8573 0.4024 0.8688 0.8676 0.3983 0.0921
455 0.8586 0.4037 0.8525 0.8518 0.4110 0.0848
456 0.8621 0.4030 0.8588 0.8593 0.4021 0.0905
457 0.8629 0.3964 0.8552 0.8533 0.4331 0.0844
458 0.8627 0.4058 0.8564 0.8563 0.4103 0.0840
459 0.8581 0.4029 0.8633 0.8619 0.3994 0.0833
460 0.8601 0.3951 0.8669 0.8662 0.4063 0.0831
461 0.8607 0.3957 0.8585 0.8590 0.4062 0.0884
462 0.8600 0.3974 0.8564 0.8536 0.4181 0.0854
463 0.8587 0.3996 0.8645 0.8638 0.3979 0.0818
464 0.8623 0.3950 0.8639 0.8632 0.4005 0.0844
465 0.8605 0.4026 0.8675 0.8663 0.4015 0.0909
466 0.8625 0.3945 0.8513 0.8504 0.4088 0.0923
467 0.8603 0.3958 0.8621 0.8619 0.4090 0.0844
468 0.8603 0.3933 0.8651 0.8639 0.3982 0.0837
469 0.8615 0.4005 0.8648 0.8639 0.4035 0.0834
470 0.8580 0.3994 0.8507 0.8499 0.4084 0.0842
471 0.8629 0.3992 0.8624 0.8617 0.4036 0.0828
472 0.8634 0.3968 0.8663 0.8659 0.4034 0.0908
473 0.8616 0.3934 0.8633 0.8627 0.3934 0.0830
474 0.8609 0.4007 0.8609 0.8593 0.4002 0.0821
475 0.8637 0.3946 0.8519 0.8518 0.4162 0.0835
476 0.8591 0.3968 0.8579 0.8566 0.4150 0.0830
477 0.8610 0.3968 0.8597 0.8601 0.3942 0.0828
478 0.8620 0.3961 0.8576 0.8571 0.4016 0.1083
479 0.8618 0.3967 0.8657 0.8641 0.3966 0.0840
480 0.8593 0.3966 0.8615 0.8599 0.4043 0.0818
481 0.8621 0.3968 0.8618 0.8624 0.3978 0.1608
482 0.8624 0.3962 0.8519 0.8482 0.4484 0.0829
483 0.8590 0.3996 0.8438 0.8448 0.4273 0.0835
484 0.8601 0.4056 0.8570 0.8551 0.4066 0.0829
485 0.8593 0.3985 0.8621 0.8622 0.4066 0.0823
486 0.8597 0.3949 0.8546 0.8542 0.4085 0.0827
487 0.8635 0.3957 0.8555 0.8560 0.3968 0.0826
488 0.8620 0.3967 0.8597 0.8581 0.4069 0.0837
489 0.8614 0.4029 0.8582 0.8574 0.4126 0.0901
490 0.8619 0.3925 0.8492 0.8496 0.4094 0.0821
491 0.8611 0.3962 0.8474 0.8421 0.4516 0.0850
492 0.8661 0.3927 0.8600 0.8585 0.4130 0.0818
493 0.8651 0.3951 0.8694 0.8687 0.3893 0.0818
494 0.8575 0.4083 0.8663 0.8649 0.4131 0.0822
495 0.8600 0.3959 0.8663 0.8653 0.3989 0.0834
496 0.8619 0.3941 0.8603 0.8596 0.3972 0.0825
497 0.8611 0.3939 0.8621 0.8628 0.3981 0.0840
498 0.8626 0.3950 0.8645 0.8635 0.3916 0.0831
499 0.8596 0.3948 0.8642 0.8628 0.4055 0.0821
500 0.8614 0.3976 0.8597 0.8594 0.3929 0.0828
Out[32]:
<class 'skorch.classifier.NeuralNetClassifier'>[initialized](
module_=BetterMLP(
(model): Sequential(
(0): Linear(in_features=22, out_features=32, bias=True)
(1): BatchNorm1d(32, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
(2): ReLU()
(3): Dropout(p=0.2, inplace=False)
(4): Linear(in_features=32, out_features=7, bias=True)
(5): Softmax(dim=-1)
)
),
)In a Jupyter environment, please rerun this cell to show the HTML representation or trust the notebook. On GitHub, the HTML representation is unable to render, please try loading this page with nbviewer.org.
<class 'skorch.classifier.NeuralNetClassifier'>[initialized](
module_=BetterMLP(
(model): Sequential(
(0): Linear(in_features=22, out_features=32, bias=True)
(1): BatchNorm1d(32, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
(2): ReLU()
(3): Dropout(p=0.2, inplace=False)
(4): Linear(in_features=32, out_features=7, bias=True)
(5): Softmax(dim=-1)
)
),
)In [33]:
y_pred = net.predict(X_test)
f1_score(y_test, y_pred, average="weighted")
Out[33]:
0.8731897654334311
In [34]:
history = net.history
plt.plot(history[:, 'train_loss'], label='train_loss')
plt.plot(history[:, 'valid_loss'], label='valid_loss')
plt.legend()
plt.xlabel('Epoch')
plt.ylabel('Loss')
plt.show()
plt.plot(history[:, 'train_f1'], label='train_f1')
plt.plot(history[:, 'valid_f1'], label='valid_f1')
plt.legend()
plt.xlabel('Epoch')
plt.ylabel('F1')
plt.show()
Better Training¶
In [35]:
from skorch.callbacks import LRScheduler, EarlyStopping, Checkpoint
from torch.optim.lr_scheduler import ReduceLROnPlateau
In [36]:
class BetterMLP(nn.Module):
def __init__(self, in_features, out_features):
super().__init__()
self.model = nn.Sequential(
nn.Linear(in_features, 32),
nn.BatchNorm1d(32), # batch normalization layer
nn.ReLU(),
nn.Dropout(p=0.2), # dropout layer with 20% dropout rate
nn.Linear(32, 16),
nn.ReLU(),
nn.Linear(16, out_features),
nn.Softmax(dim=-1),
)
def forward(self, X, **kwargs):
return self.model(X)
In [37]:
from skorch.dataset import ValidSplit
net = NeuralNetClassifier(
BetterMLP,
module__in_features=X_train.shape[1],
module__out_features=len(le.classes_),
max_epochs=500,
lr = 1e-3,
device=DEVICE,
optimizer=torch.optim.AdamW, # using AdamW instead of Adam for better regularization
callbacks=[
EpochScoring(
scoring='f1_weighted',
name='train_f1',
on_train=True,
lower_is_better=False,
),
EpochScoring(
scoring='f1_weighted',
name='valid_f1',
on_train=False,
lower_is_better=False,
),
EarlyStopping(
monitor="valid_f1",
patience=50,
lower_is_better=False,
),
LRScheduler(
policy=ReduceLROnPlateau,
monitor="valid_f1",
mode="max",
factor=0.5,
patience=25,
threshold=1e-4,
min_lr=1e-6,
verbose=True,
),
Checkpoint(
monitor="valid_acc_best",
load_best=True,
f_history=None,
f_optimizer=None,
)
],
train_split=ValidSplit(cv=5, stratified=True),
batch_size=2048,
iterator_train__shuffle=True,
)
net
Out[37]:
<class 'skorch.classifier.NeuralNetClassifier'>[uninitialized]( module=<class '__main__.BetterMLP'>, module__in_features=22, module__out_features=7, )In a Jupyter environment, please rerun this cell to show the HTML representation or trust the notebook.
On GitHub, the HTML representation is unable to render, please try loading this page with nbviewer.org.
<class 'skorch.classifier.NeuralNetClassifier'>[uninitialized]( module=<class '__main__.BetterMLP'>, module__in_features=22, module__out_features=7, )
In [38]:
# Training the network
net.fit(X_train, y_train)
/Users/francesco/miniforge3/envs/timeseries_dl/lib/python3.12/site-packages/torch/optim/lr_scheduler.py:28: UserWarning: The verbose parameter is deprecated. Please use get_last_lr() to access the learning rate.
warnings.warn("The verbose parameter is deprecated. Please use get_last_lr() "
epoch train_f1 train_loss valid_acc valid_f1 valid_loss cp lr dur
------- ---------- ------------ ----------- ---------- ------------ ---- ------ ------
1 0.0830 1.9260 0.2914 0.1600 1.9539 + 0.0010 0.4120
2 0.1850 1.8699 0.2718 0.1739 1.8814 0.0010 0.1013
3 0.2015 1.8139 0.3104 0.1848 1.8047 + 0.0010 0.1000
4 0.1937 1.7535 0.3152 0.1703 1.7270 + 0.0010 0.0916
5 0.1846 1.6877 0.3170 0.1677 1.6488 + 0.0010 0.0918
6 0.1801 1.6216 0.3209 0.1745 1.5796 + 0.0010 0.0918
7 0.1797 1.5583 0.3200 0.1732 1.5111 0.0010 0.0880
8 0.1845 1.5001 0.3104 0.1737 1.4591 0.0010 0.0891
9 0.1917 1.4470 0.2914 0.1951 1.4184 0.0010 0.0893
10 0.2063 1.4011 0.2974 0.2017 1.3791 0.0010 0.0994
11 0.2258 1.3611 0.3046 0.2167 1.3560 0.0010 0.1279
12 0.2540 1.3213 0.3206 0.2416 1.3406 0.0010 0.0967
13 0.2821 1.2876 0.3582 0.2753 1.3242 + 0.0010 0.0929
14 0.3195 1.2537 0.3700 0.2814 1.2863 + 0.0010 0.0903
15 0.3676 1.2217 0.3934 0.3179 1.2419 + 0.0010 0.0903
16 0.4058 1.1885 0.4181 0.3584 1.2412 + 0.0010 0.0963
17 0.4588 1.1557 0.4314 0.3873 1.1864 + 0.0010 0.1811
18 0.4995 1.1236 0.4561 0.4032 1.1542 + 0.0010 0.0940
19 0.5262 1.0911 0.5298 0.4898 1.1012 + 0.0010 0.0934
20 0.5531 1.0560 0.5187 0.4770 1.1129 0.0010 0.0892
21 0.5907 1.0218 0.5704 0.5430 1.0544 + 0.0010 0.0952
22 0.6102 0.9857 0.5771 0.5503 1.0365 + 0.0010 0.1190
23 0.6326 0.9500 0.5825 0.5526 1.0288 + 0.0010 0.0949
24 0.6500 0.9163 0.6303 0.6050 0.9144 + 0.0010 0.0923
25 0.6669 0.8851 0.6132 0.5903 0.9737 0.0010 0.0885
26 0.6807 0.8516 0.6565 0.6349 0.8852 + 0.0010 0.0883
27 0.6928 0.8224 0.6493 0.6264 0.8810 0.0010 0.0874
28 0.6984 0.7977 0.7002 0.6863 0.8057 + 0.0010 0.0890
29 0.7174 0.7704 0.7146 0.7060 0.8089 + 0.0010 0.0906
30 0.7223 0.7452 0.7306 0.7199 0.7801 + 0.0010 0.0895
31 0.7318 0.7258 0.7246 0.7122 0.7462 0.0010 0.0878
32 0.7375 0.7058 0.7547 0.7439 0.6959 + 0.0010 0.0952
33 0.7403 0.6903 0.7297 0.7183 0.7159 0.0010 0.1047
34 0.7494 0.6718 0.7787 0.7716 0.6406 + 0.0010 0.0989
35 0.7526 0.6622 0.7634 0.7556 0.6515 0.0010 0.0908
36 0.7569 0.6457 0.7896 0.7844 0.6323 + 0.0010 0.0925
37 0.7607 0.6353 0.7917 0.7888 0.6161 + 0.0010 0.0997
38 0.7657 0.6256 0.7974 0.7919 0.5893 + 0.0010 0.1053
39 0.7633 0.6148 0.7926 0.7881 0.5938 0.0010 0.0948
40 0.7706 0.6056 0.8019 0.7976 0.5914 + 0.0010 0.1021
41 0.7766 0.5986 0.8131 0.8109 0.5580 + 0.0010 0.0927
42 0.7727 0.5923 0.7793 0.7747 0.5991 0.0010 0.0928
43 0.7800 0.5825 0.7778 0.7733 0.5917 0.0010 0.0925
44 0.7773 0.5797 0.7845 0.7810 0.5832 0.0010 0.0941
45 0.7840 0.5685 0.7715 0.7664 0.5952 0.0010 0.1024
46 0.7858 0.5627 0.7848 0.7803 0.5730 0.0010 0.0993
47 0.7914 0.5573 0.8061 0.8017 0.5371 0.0010 0.0942
48 0.7882 0.5559 0.7520 0.7470 0.6484 0.0010 0.1010
49 0.7871 0.5546 0.8082 0.8075 0.5486 0.0010 0.1012
50 0.7934 0.5453 0.8311 0.8322 0.5053 + 0.0010 0.1323
51 0.7927 0.5386 0.7920 0.7884 0.5633 0.0010 0.0994
52 0.7933 0.5359 0.8046 0.8014 0.5390 0.0010 0.1120
53 0.7969 0.5309 0.8441 0.8426 0.4813 + 0.0010 0.0946
54 0.7960 0.5315 0.7875 0.7841 0.5811 0.0010 0.0950
55 0.7992 0.5264 0.8290 0.8270 0.5009 0.0010 0.1714
56 0.8052 0.5188 0.8396 0.8391 0.4993 0.0010 0.0932
57 0.8055 0.5106 0.8016 0.7976 0.5277 0.0010 0.0932
58 0.8041 0.5129 0.7122 0.7087 0.6774 0.0010 0.0923
59 0.8099 0.5078 0.8402 0.8381 0.4814 0.0010 0.0927
60 0.8034 0.5089 0.8441 0.8459 0.4929 0.0010 0.0894
61 0.8103 0.5022 0.8290 0.8272 0.4952 0.0010 0.0902
62 0.8079 0.5012 0.7565 0.7534 0.6229 0.0010 0.0897
63 0.8089 0.4989 0.8299 0.8278 0.4925 0.0010 0.0906
64 0.8132 0.4968 0.8522 0.8518 0.4572 + 0.0010 0.0901
65 0.8157 0.4932 0.7875 0.7873 0.5667 0.0010 0.0897
66 0.8143 0.4896 0.8263 0.8228 0.4857 0.0010 0.0904
67 0.8189 0.4867 0.8152 0.8132 0.5100 0.0010 0.0900
68 0.8184 0.4797 0.8257 0.8244 0.4988 0.0010 0.0986
69 0.8185 0.4813 0.8474 0.8477 0.4474 0.0010 0.0915
70 0.8224 0.4802 0.8441 0.8431 0.4610 0.0010 0.0896
71 0.8225 0.4768 0.7890 0.7852 0.5392 0.0010 0.0902
72 0.8194 0.4781 0.8495 0.8483 0.4389 0.0010 0.0910
73 0.8198 0.4762 0.8558 0.8553 0.4287 + 0.0010 0.0911
74 0.8202 0.4749 0.8215 0.8181 0.4901 0.0010 0.0899
75 0.8247 0.4709 0.8396 0.8380 0.4660 0.0010 0.0897
76 0.8218 0.4716 0.8471 0.8457 0.4425 0.0010 0.0897
77 0.8236 0.4738 0.8588 0.8590 0.4275 + 0.0010 0.0918
78 0.8251 0.4711 0.8468 0.8462 0.4365 0.0010 0.1202
79 0.8193 0.4714 0.7766 0.7731 0.5634 0.0010 0.0926
80 0.8205 0.4713 0.8480 0.8480 0.4291 0.0010 0.0892
81 0.8265 0.4646 0.8603 0.8599 0.4206 + 0.0010 0.0900
82 0.8278 0.4629 0.8453 0.8426 0.4403 0.0010 0.0912
83 0.8277 0.4628 0.8585 0.8583 0.4342 0.0010 0.0984
84 0.8237 0.4612 0.8314 0.8272 0.4556 0.0010 0.0898
85 0.8331 0.4527 0.8362 0.8368 0.4588 0.0010 0.0897
86 0.8261 0.4591 0.8435 0.8420 0.4496 0.0010 0.0904
87 0.8260 0.4583 0.8585 0.8570 0.4249 0.0010 0.0911
88 0.8311 0.4530 0.8555 0.8543 0.4267 0.0010 0.0902
89 0.8296 0.4567 0.8573 0.8570 0.4319 0.0010 0.0890
90 0.8297 0.4509 0.8576 0.8571 0.4201 0.0010 0.0878
91 0.8299 0.4472 0.8570 0.8566 0.4315 0.0010 0.0900
92 0.8311 0.4556 0.8585 0.8586 0.4141 0.0010 0.0893
93 0.8306 0.4510 0.8522 0.8505 0.4311 0.0010 0.0901
94 0.8309 0.4549 0.8374 0.8370 0.4566 0.0010 0.1686
95 0.8325 0.4499 0.8531 0.8529 0.4177 0.0010 0.0896
96 0.8320 0.4484 0.8561 0.8553 0.4169 0.0010 0.0885
97 0.8339 0.4454 0.8543 0.8534 0.4337 0.0010 0.0989
98 0.8321 0.4483 0.8558 0.8546 0.4236 0.0010 0.0899
99 0.8366 0.4455 0.8522 0.8494 0.4298 0.0010 0.0882
100 0.8350 0.4412 0.8444 0.8437 0.4455 0.0010 0.0889
101 0.8345 0.4431 0.8561 0.8538 0.4197 0.0010 0.0904
102 0.8397 0.4409 0.8459 0.8457 0.4309 0.0010 0.0896
103 0.8363 0.4408 0.8561 0.8561 0.4098 0.0010 0.0892
104 0.8371 0.4391 0.8594 0.8575 0.4159 0.0010 0.0905
105 0.8357 0.4423 0.8299 0.8278 0.4690 0.0010 0.0922
106 0.8324 0.4456 0.8525 0.8540 0.4273 0.0010 0.0892
107 0.8393 0.4440 0.8411 0.8387 0.4411 0.0010 0.0912
108 0.8416 0.4322 0.8582 0.8577 0.4031 0.0005 0.0895
109 0.8393 0.4345 0.8591 0.8589 0.4177 0.0005 0.0908
110 0.8388 0.4356 0.8600 0.8603 0.4036 0.0005 0.0982
111 0.8365 0.4356 0.8624 0.8618 0.4018 + 0.0005 0.0882
112 0.8421 0.4347 0.8528 0.8512 0.4127 0.0005 0.0888
113 0.8385 0.4365 0.8657 0.8650 0.4014 + 0.0005 0.0903
114 0.8409 0.4335 0.8609 0.8602 0.4042 0.0005 0.0897
115 0.8379 0.4378 0.8630 0.8626 0.4030 0.0005 0.0915
116 0.8398 0.4387 0.8642 0.8636 0.4019 0.0005 0.0907
117 0.8408 0.4303 0.8615 0.8605 0.4056 0.0005 0.0893
118 0.8430 0.4327 0.8633 0.8626 0.4031 0.0005 0.0907
119 0.8409 0.4381 0.8627 0.8623 0.4068 0.0005 0.0901
120 0.8364 0.4372 0.8630 0.8631 0.4007 0.0005 0.0902
121 0.8443 0.4275 0.8669 0.8659 0.4011 + 0.0005 0.0914
122 0.8444 0.4261 0.8504 0.8493 0.4140 0.0005 0.0900
123 0.8398 0.4330 0.8639 0.8634 0.3996 0.0005 0.0903
124 0.8403 0.4252 0.8639 0.8629 0.4059 0.0005 0.0901
125 0.8456 0.4279 0.8624 0.8613 0.4053 0.0005 0.0890
126 0.8398 0.4333 0.8636 0.8632 0.4020 0.0005 0.0984
127 0.8453 0.4278 0.8624 0.8615 0.4023 0.0005 0.0894
128 0.8415 0.4291 0.8621 0.8619 0.3996 0.0005 0.0877
129 0.8429 0.4274 0.8672 0.8669 0.4000 + 0.0005 0.0890
130 0.8383 0.4308 0.8519 0.8512 0.4082 0.0005 0.0878
131 0.8391 0.4327 0.8648 0.8646 0.3964 0.0005 0.0895
132 0.8476 0.4202 0.8639 0.8630 0.4024 0.0005 0.1677
133 0.8425 0.4304 0.8510 0.8509 0.4122 0.0005 0.0886
134 0.8423 0.4302 0.8600 0.8582 0.4060 0.0005 0.1195
135 0.8472 0.4246 0.8540 0.8539 0.4039 0.0005 0.0907
136 0.8400 0.4318 0.8573 0.8557 0.4089 0.0005 0.0894
137 0.8444 0.4241 0.8588 0.8585 0.4031 0.0005 0.0905
138 0.8436 0.4280 0.8633 0.8627 0.4064 0.0005 0.0915
139 0.8439 0.4249 0.8648 0.8647 0.4040 0.0005 0.0897
140 0.8427 0.4252 0.8549 0.8550 0.4043 0.0005 0.0970
141 0.8413 0.4277 0.8685 0.8676 0.3972 + 0.0005 0.0897
142 0.8441 0.4255 0.8603 0.8594 0.3996 0.0005 0.0896
143 0.8450 0.4259 0.8600 0.8586 0.4060 0.0005 0.0889
144 0.8471 0.4225 0.8630 0.8632 0.4018 0.0005 0.0895
145 0.8437 0.4248 0.8645 0.8639 0.4011 0.0005 0.0890
146 0.8446 0.4311 0.8564 0.8555 0.4023 0.0005 0.0891
147 0.8442 0.4215 0.8618 0.8606 0.4000 0.0005 0.0899
148 0.8434 0.4224 0.8615 0.8602 0.4031 0.0005 0.0895
149 0.8479 0.4223 0.8633 0.8635 0.3998 0.0005 0.0889
150 0.8445 0.4247 0.8579 0.8572 0.4042 0.0005 0.0897
151 0.8474 0.4176 0.8666 0.8665 0.3968 0.0005 0.0896
152 0.8429 0.4246 0.8651 0.8638 0.4012 0.0005 0.0895
153 0.8455 0.4218 0.8636 0.8629 0.3969 0.0005 0.0890
154 0.8474 0.4197 0.8609 0.8608 0.3976 0.0005 0.0900
155 0.8467 0.4200 0.8585 0.8580 0.3978 0.0005 0.0982
156 0.8426 0.4227 0.8645 0.8639 0.4025 0.0005 0.0882
157 0.8452 0.4204 0.8669 0.8664 0.3968 0.0005 0.0905
158 0.8444 0.4203 0.8648 0.8642 0.3954 0.0005 0.0887
159 0.8472 0.4176 0.8639 0.8629 0.4029 0.0005 0.0882
160 0.8449 0.4207 0.8537 0.8542 0.4067 0.0005 0.0895
161 0.8424 0.4233 0.8609 0.8596 0.4021 0.0005 0.0896
162 0.8460 0.4202 0.8573 0.8575 0.4034 0.0005 0.0894
163 0.8463 0.4206 0.8648 0.8636 0.3980 0.0005 0.0898
164 0.8464 0.4178 0.8642 0.8635 0.3953 0.0005 0.0878
165 0.8452 0.4212 0.8597 0.8579 0.4102 0.0005 0.0886
166 0.8448 0.4229 0.8561 0.8563 0.4064 0.0005 0.0903
167 0.8452 0.4237 0.8645 0.8634 0.3955 0.0005 0.0882
168 0.8504 0.4173 0.8666 0.8661 0.3953 0.0003 0.0902
169 0.8449 0.4223 0.8700 0.8692 0.3942 + 0.0003 0.0885
170 0.8502 0.4193 0.8669 0.8657 0.3999 0.0003 0.0891
171 0.8455 0.4188 0.8594 0.8589 0.3959 0.0003 0.1734
172 0.8462 0.4180 0.8694 0.8688 0.3924 0.0003 0.0899
173 0.8468 0.4198 0.8648 0.8645 0.3937 0.0003 0.0897
174 0.8475 0.4159 0.8588 0.8586 0.3965 0.0003 0.0896
175 0.8480 0.4169 0.8588 0.8581 0.3975 0.0003 0.0900
176 0.8446 0.4200 0.8672 0.8668 0.3921 0.0003 0.0907
177 0.8493 0.4131 0.8675 0.8670 0.3947 0.0003 0.0931
178 0.8469 0.4203 0.8657 0.8653 0.3933 0.0003 0.0923
179 0.8474 0.4154 0.8645 0.8641 0.3937 0.0003 0.0897
180 0.8502 0.4152 0.8669 0.8660 0.3951 0.0003 0.0909
181 0.8450 0.4201 0.8639 0.8635 0.3936 0.0003 0.0908
182 0.8513 0.4137 0.8639 0.8632 0.3928 0.0003 0.0922
183 0.8483 0.4181 0.8642 0.8636 0.3924 0.0003 0.0913
184 0.8464 0.4197 0.8682 0.8671 0.3922 0.0003 0.0914
185 0.8463 0.4178 0.8675 0.8664 0.3966 0.0003 0.0903
186 0.8463 0.4171 0.8621 0.8615 0.3939 0.0003 0.0898
187 0.8506 0.4158 0.8669 0.8667 0.3905 0.0003 0.0996
188 0.8491 0.4147 0.8672 0.8667 0.3914 0.0003 0.1182
189 0.8458 0.4181 0.8648 0.8649 0.3930 0.0003 0.0896
190 0.8453 0.4245 0.8663 0.8660 0.3905 0.0003 0.0888
191 0.8471 0.4208 0.8669 0.8664 0.3924 0.0003 0.0902
192 0.8531 0.4168 0.8648 0.8646 0.3930 0.0003 0.0902
193 0.8473 0.4175 0.8642 0.8640 0.3911 0.0003 0.0911
194 0.8510 0.4142 0.8660 0.8650 0.3933 0.0003 0.0902
195 0.8469 0.4221 0.8639 0.8628 0.3951 0.0003 0.0881
196 0.8498 0.4110 0.8642 0.8635 0.3919 0.0001 0.0891
197 0.8479 0.4191 0.8654 0.8647 0.3916 0.0001 0.0910
198 0.8472 0.4206 0.8654 0.8646 0.3910 0.0001 0.0918
199 0.8509 0.4122 0.8688 0.8679 0.3925 0.0001 0.0899
200 0.8488 0.4134 0.8691 0.8684 0.3914 0.0001 0.0903
201 0.8479 0.4140 0.8660 0.8653 0.3915 0.0001 0.0982
202 0.8483 0.4155 0.8624 0.8617 0.3935 0.0001 0.0894
203 0.8492 0.4140 0.8642 0.8634 0.3917 0.0001 0.0893
204 0.8526 0.4140 0.8660 0.8650 0.3920 0.0001 0.0914
205 0.8487 0.4143 0.8648 0.8641 0.3917 0.0001 0.0914
206 0.8480 0.4141 0.8669 0.8663 0.3902 0.0001 0.0902
207 0.8524 0.4115 0.8697 0.8690 0.3899 0.0001 0.0904
208 0.8481 0.4210 0.8685 0.8679 0.3903 0.0001 0.0892
209 0.8515 0.4084 0.8666 0.8660 0.3913 0.0001 0.1671
210 0.8504 0.4148 0.8675 0.8669 0.3906 0.0001 0.0909
211 0.8472 0.4135 0.8682 0.8673 0.3915 0.0001 0.0909
212 0.8481 0.4152 0.8694 0.8687 0.3904 0.0001 0.0907
213 0.8484 0.4141 0.8666 0.8659 0.3905 0.0001 0.0892
214 0.8495 0.4171 0.8657 0.8651 0.3916 0.0001 0.0903
215 0.8487 0.4166 0.8654 0.8648 0.3917 0.0001 0.0893
216 0.8476 0.4145 0.8657 0.8650 0.3914 0.0001 0.0973
217 0.8508 0.4133 0.8682 0.8674 0.3915 0.0001 0.0910
218 0.8487 0.4156 0.8688 0.8682 0.3900 0.0001 0.0900
Stopping since valid_f1 has not improved in the last 50 epochs.
Out[38]:
<class 'skorch.classifier.NeuralNetClassifier'>[initialized](
module_=BetterMLP(
(model): Sequential(
(0): Linear(in_features=22, out_features=32, bias=True)
(1): BatchNorm1d(32, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
(2): ReLU()
(3): Dropout(p=0.2, inplace=False)
(4): Linear(in_features=32, out_features=16, bias=True)
(5): ReLU()
(6): Linear(in_features=16, out_features=7, bias=True)
(7): Softmax(dim=-1)
)
),
)In a Jupyter environment, please rerun this cell to show the HTML representation or trust the notebook. On GitHub, the HTML representation is unable to render, please try loading this page with nbviewer.org.
<class 'skorch.classifier.NeuralNetClassifier'>[initialized](
module_=BetterMLP(
(model): Sequential(
(0): Linear(in_features=22, out_features=32, bias=True)
(1): BatchNorm1d(32, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
(2): ReLU()
(3): Dropout(p=0.2, inplace=False)
(4): Linear(in_features=32, out_features=16, bias=True)
(5): ReLU()
(6): Linear(in_features=16, out_features=7, bias=True)
(7): Softmax(dim=-1)
)
),
)In [39]:
y_pred = net.predict(X_test)
f1_score(y_test, y_pred, average="weighted")
Out[39]:
0.8739429187585216
In [40]:
history = net.history
plt.plot(history[:, 'train_loss'], label='train_loss')
plt.plot(history[:, 'valid_loss'], label='valid_loss')
ax1 = plt.gca() # get current axes
ax2 = plt.gca().twinx() # create a second y-axis
plt.plot(history[:, "event_lr"], label='learning_rate', color='gray', linestyle='--')
ax2.set_ylabel('Learning Rate')
ax1.set_ylabel('Loss')
for i, checkpoint in enumerate(history[:, "event_cp"][::-1]):
if checkpoint:
ax1.axvline(x=len(history[:, "event_cp"])-i-1, color='red',
linestyle='-.', label='Checkpoint', alpha=0.5)
break
ax1.legend(loc='lower left')
plt.xlabel('Epoch')
plt.show()
plt.plot(history[:, 'train_f1'], label='train_f1')
plt.plot(history[:, 'valid_f1'], label='valid_f1')
plt.xlabel('Epoch')
plt.ylabel('F1')
plt.legend()
plt.show()
Regression¶
In [41]:
import numpy as np
import pandas as pd
import matplotlib.pyplot as plt
from sklearn.model_selection import train_test_split, cross_val_score
import torch
from torch import nn
import torch.nn.functional as F
from skorch import NeuralNetClassifier
from skorch.callbacks import EpochScoring
from sklearn.metrics import mean_absolute_error
from skorch import NeuralNetRegressor
In [42]:
target = "Age"
y = df[target].astype(np.float32).values
X = df.drop(columns=[target, "id"])
In [43]:
# Convert categorical variables to numeric using one-hot encoding
X[X.select_dtypes('object').columns] = X.select_dtypes('object').apply(pd.Categorical)
X = pd.get_dummies(X, drop_first=True).astype(np.float32).values
In [44]:
# Split the data into training and testing sets
X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=0.2, random_state=42)
In [45]:
from sklearn.dummy import DummyRegressor
model = DummyRegressor(strategy="mean")
model.fit(X_train, y_train)
y_pred = model.predict(X_test)
mean_absolute_error(y_test, y_pred)
Out[45]:
4.175730228424072
In [46]:
from sklearn.neural_network import MLPRegressor
model = MLPRegressor(
hidden_layer_sizes=(32,),
solver="adam",
alpha=0.001,
max_iter=500,
random_state=42,
batch_size=2048,
shuffle=True,
)
model.fit(X_train, y_train)
y_pred = model.predict(X_test)
mean_absolute_error(y_test, y_pred)
/Users/francesco/miniforge3/envs/timeseries_dl/lib/python3.12/site-packages/sklearn/neural_network/_multilayer_perceptron.py:691: ConvergenceWarning: Stochastic Optimizer: Maximum iterations (500) reached and the optimization hasn't converged yet. warnings.warn(
Out[46]:
2.272099018096924
In [47]:
class BetterMLP(nn.Module):
def __init__(self, in_features, out_features):
super().__init__()
self.model = nn.Sequential(
nn.Linear(in_features, 32),
nn.BatchNorm1d(32), # batch normalization layer
nn.ReLU(),
nn.Linear(32, out_features),
# no activation at the end since this is regression
)
def forward(self, X, **kwargs):
return self.model(X)
In [48]:
from skorch.dataset import ValidSplit
from skorch.callbacks import LRScheduler, EarlyStopping, Checkpoint
from torch.optim.lr_scheduler import ReduceLROnPlateau
net = NeuralNetRegressor(
BetterMLP,
module__in_features=X_train.shape[1],
module__out_features=1,
max_epochs=500,
lr = 1e-3,
device=DEVICE,
optimizer=torch.optim.AdamW, # using AdamW instead of Adam for better regularization
criterion=nn.MSELoss, # using MSE loss for regression
callbacks=[
EpochScoring(
mean_absolute_error,
name="valid_mae",
lower_is_better=True,
on_train=False,
),
EpochScoring(
mean_absolute_error,
name="train_mae",
lower_is_better=True,
on_train=True,
),
LRScheduler(
policy=ReduceLROnPlateau,
monitor="valid_mae",
mode="min",
factor=0.5,
patience=25,
threshold=1e-4,
min_lr=1e-6,
verbose=True,
),
EarlyStopping(
monitor="valid_mae",
lower_is_better=True,
patience=40,
load_best=True,
),
Checkpoint(
monitor="valid_mae_best",
load_best=True,
f_history=None,
f_optimizer=None,
),
],
batch_size=2048,
iterator_train__shuffle=True,
)
net
Out[48]:
<class 'skorch.regressor.NeuralNetRegressor'>[uninitialized]( module=<class '__main__.BetterMLP'>, module__in_features=27, module__out_features=1, )In a Jupyter environment, please rerun this cell to show the HTML representation or trust the notebook.
On GitHub, the HTML representation is unable to render, please try loading this page with nbviewer.org.
<class 'skorch.regressor.NeuralNetRegressor'>[uninitialized]( module=<class '__main__.BetterMLP'>, module__in_features=27, module__out_features=1, )
In [49]:
# Training the network
net.fit(X_train, y_train[:, None])
/Users/francesco/miniforge3/envs/timeseries_dl/lib/python3.12/site-packages/torch/optim/lr_scheduler.py:28: UserWarning: The verbose parameter is deprecated. Please use get_last_lr() to access the learning rate.
warnings.warn("The verbose parameter is deprecated. Please use get_last_lr() "
epoch train_loss train_mae valid_loss valid_mae cp lr dur
------- ------------ ----------- ------------ ----------- ---- ------ ------
1 615.6085 24.1479 620.8583 24.2237 + 0.0010 0.7635
2 609.6344 24.0249 594.3533 23.6981 + 0.0010 0.0959
3 603.7377 23.9016 584.2501 23.4937 + 0.0010 0.0958
4 597.9963 23.7810 582.1346 23.4469 + 0.0010 0.0967
5 592.3411 23.6611 583.7179 23.4732 0.0010 0.1025
6 586.7383 23.5401 585.3101 23.4981 0.0010 0.1052
7 581.1118 23.4179 581.9609 23.4199 + 0.0010 0.1372
8 575.3965 23.2951 577.8797 23.3286 + 0.0010 0.0998
9 569.5366 23.1720 566.8957 23.0928 + 0.0010 0.1016
10 563.5585 23.0449 554.6788 22.8320 + 0.0010 0.1034
11 557.5785 22.9171 557.9930 22.8983 0.0010 0.1123
12 551.2931 22.7883 555.6525 22.8413 0.0010 0.1007
13 544.8574 22.6553 544.7441 22.6148 + 0.0010 0.0988
14 538.0173 22.5182 528.9284 22.2732 + 0.0010 0.1278
15 531.0705 22.3894 515.9805 21.9958 + 0.0010 0.1047
16 523.9074 22.2379 502.9861 21.7242 + 0.0010 0.1934
17 516.4339 22.0936 476.4068 21.1286 + 0.0010 0.1012
18 508.8763 21.9524 492.5345 21.5311 0.0010 0.1002
19 501.1751 21.7883 503.3328 21.7922 0.0010 0.1001
20 493.2712 21.6231 492.8367 21.5734 0.0010 0.1050
21 485.1305 21.4574 490.8459 21.5350 0.0010 0.1486
22 476.7567 21.2818 490.0673 21.5339 0.0010 0.1028
23 468.1747 21.0929 477.2153 21.2477 0.0010 0.1028
24 459.3939 20.9044 453.6153 20.7284 + 0.0010 0.1011
25 450.3256 20.7047 440.3734 20.4285 + 0.0010 0.0989
26 441.1153 20.5007 435.4073 20.3271 + 0.0010 0.0982
27 431.8324 20.2846 438.4441 20.4027 0.0010 0.0972
28 422.4637 20.0646 419.2206 19.9451 + 0.0010 0.1118
29 413.0077 19.8392 418.4272 19.9355 + 0.0010 0.1058
30 403.4338 19.6067 416.7641 19.8898 + 0.0010 0.1002
31 393.6827 19.3667 400.5614 19.5050 + 0.0010 0.1018
32 383.8090 19.1158 394.2751 19.3527 + 0.0010 0.0994
33 373.7780 18.8614 368.6050 18.6858 + 0.0010 0.1022
34 363.7001 18.6007 356.5817 18.3663 + 0.0010 0.1127
35 353.6262 18.3346 357.5906 18.3751 0.0010 0.1061
36 343.5071 18.0691 347.4042 18.1160 + 0.0010 0.0995
37 333.4587 17.7955 343.3687 18.0017 + 0.0010 0.1271
38 323.3497 17.5192 332.8711 17.7251 + 0.0010 0.1073
39 313.2658 17.2363 338.4815 17.8486 0.0010 0.1129
40 303.2852 16.9536 311.5698 17.1403 + 0.0010 0.1156
41 293.3487 16.6645 290.5961 16.5457 + 0.0010 0.0986
42 283.4980 16.3709 289.1552 16.4901 + 0.0010 0.1816
43 273.7227 16.0756 278.2753 16.1927 + 0.0010 0.1102
44 264.0995 15.7783 247.7530 15.2263 + 0.0010 0.1061
45 254.5279 15.4759 242.6402 15.0534 + 0.0010 0.1025
46 245.0704 15.1733 249.7249 15.3108 0.0010 0.1076
47 235.6129 14.8630 221.6600 14.3668 + 0.0010 0.1025
48 226.3223 14.5489 203.6025 13.7316 + 0.0010 0.0997
49 217.1373 14.2352 200.6701 13.6246 + 0.0010 0.1244
50 208.1794 13.9182 205.3880 13.8112 0.0010 0.1063
51 199.3254 13.6014 193.3978 13.3722 + 0.0010 0.1097
52 190.5778 13.2789 179.9962 12.8689 + 0.0010 0.1088
53 182.0794 12.9580 183.8069 13.0041 0.0010 0.1020
54 173.7490 12.6366 175.0394 12.6347 + 0.0010 0.1019
55 165.7041 12.3157 173.6547 12.6299 + 0.0010 0.0989
56 157.8427 11.9969 157.0681 11.9408 + 0.0010 0.0978
57 150.2190 11.6757 172.7212 12.5955 0.0010 0.0986
58 142.8233 11.3602 159.1385 12.0206 0.0010 0.0987
59 135.6361 11.0460 142.8611 11.3420 + 0.0010 0.0980
60 128.7858 10.7330 140.0285 11.2383 + 0.0010 0.0980
61 122.0661 10.4216 130.1695 10.7868 + 0.0010 0.0982
62 115.6664 10.1191 114.6827 10.0530 + 0.0010 0.0971
63 109.5891 9.8123 161.0770 12.0380 0.0010 0.0997
64 103.6504 9.5141 132.7724 10.8983 0.0010 0.0983
65 97.8957 9.2143 112.9393 9.9443 + 0.0010 0.1067
66 92.4600 8.9230 87.7266 8.6260 + 0.0010 0.0977
67 87.3025 8.6312 92.8610 8.9661 0.0010 0.0960
68 82.3165 8.3483 72.5451 7.7481 + 0.0010 0.1762
69 77.4450 8.0623 63.0638 7.1091 + 0.0010 0.0964
70 72.9954 7.7904 47.7499 6.0034 + 0.0010 0.0985
71 68.7289 7.5203 46.2908 5.8458 + 0.0010 0.0980
72 64.7365 7.2526 48.5953 6.0177 0.0010 0.0979
73 60.9653 6.9960 71.1719 7.7048 0.0010 0.0989
74 57.2131 6.7405 74.8591 7.8285 0.0010 0.0968
75 53.9756 6.5011 83.8327 8.3164 0.0010 0.0988
76 50.7223 6.2654 53.1948 6.4641 0.0010 0.0986
77 47.5298 6.0291 42.4083 5.6492 + 0.0010 0.0999
78 44.4640 5.8015 34.7861 4.9674 + 0.0010 0.1568
79 41.8125 5.5790 31.2314 4.5917 + 0.0010 0.1015
80 39.2421 5.3651 37.4587 5.1288 0.0010 0.1008
81 36.9390 5.1664 21.9634 3.6475 + 0.0010 0.0986
82 34.6786 4.9686 18.3454 3.1985 + 0.0010 0.0993
83 32.8448 4.7937 18.3207 3.1803 + 0.0010 0.0974
84 30.9007 4.6095 20.1405 3.3134 0.0010 0.0968
85 29.2056 4.4354 21.7719 3.5636 0.0010 0.0972
86 27.5839 4.2816 20.9368 3.4854 0.0010 0.0968
87 26.0905 4.1299 28.9042 4.4373 0.0010 0.1009
88 24.6701 3.9742 27.1625 4.2657 0.0010 0.1055
89 23.3916 3.8412 39.4685 5.3296 0.0010 0.1012
90 22.3035 3.7193 27.7205 4.2721 0.0010 0.0972
91 21.2619 3.6063 29.0469 4.4439 0.0010 0.0979
92 20.1868 3.4866 24.3693 3.9239 0.0010 0.1057
93 19.3472 3.3668 15.9236 2.9404 + 0.0010 0.0982
94 18.5393 3.2793 26.9245 4.1964 0.0010 0.1781
95 17.8448 3.1980 26.7227 4.1912 0.0010 0.0982
96 17.1511 3.1040 13.7885 2.6586 + 0.0010 0.0953
97 16.5009 3.0160 16.5534 3.0281 0.0010 0.0972
98 15.9473 2.9467 21.0993 3.5798 0.0010 0.0980
99 15.5190 2.8937 26.9774 4.1806 0.0010 0.0967
100 15.0258 2.8135 12.6611 2.4495 + 0.0010 0.0972
101 14.5719 2.7478 20.8927 3.5404 0.0010 0.0961
102 14.3742 2.7155 29.4341 4.4394 0.0010 0.0969
103 13.9830 2.6591 47.4778 5.8598 0.0010 0.0975
104 13.8089 2.6272 22.7402 3.6499 0.0010 0.0981
105 13.5517 2.6086 15.6046 2.7635 0.0010 0.0981
106 13.0961 2.5259 12.8392 2.4594 0.0010 0.0984
107 12.8343 2.4917 11.4291 2.2354 + 0.0010 0.1044
108 12.4338 2.4204 13.0088 2.5335 0.0010 0.0982
109 12.2874 2.3946 19.3264 3.3196 0.0010 0.0971
110 12.1138 2.3691 11.6018 2.2473 0.0010 0.0974
111 12.0251 2.3638 17.5865 3.1926 0.0010 0.0972
112 11.8808 2.3140 11.9010 2.3256 0.0010 0.0970
113 11.7689 2.3023 20.0774 3.4022 0.0010 0.1008
114 11.6270 2.2998 16.3630 2.8137 0.0010 0.0967
115 11.4558 2.2411 22.8836 3.7181 0.0010 0.0985
116 11.4614 2.2522 11.5658 2.3253 0.0010 0.0974
117 11.3422 2.2258 11.2492 2.1832 + 0.0010 0.0974
118 11.2104 2.2154 11.0497 2.1661 + 0.0010 0.0986
119 11.1275 2.1805 11.4183 2.2102 0.0010 0.1220
120 11.0619 2.1749 10.9135 2.1287 + 0.0010 0.0969
121 11.0731 2.1914 10.9170 2.1291 0.0010 0.1768
122 11.0093 2.1494 14.0311 2.7625 0.0010 0.0979
123 10.9927 2.1789 11.1363 2.2367 0.0010 0.1071
124 10.9019 2.1455 13.7135 2.4961 0.0010 0.0992
125 10.8556 2.1473 11.6349 2.2285 0.0010 0.0972
126 10.8413 2.1387 14.6112 2.7877 0.0010 0.0972
127 10.8254 2.1354 13.2510 2.6168 0.0010 0.0970
128 10.9000 2.1733 12.9842 2.3663 0.0010 0.0985
129 10.8152 2.1253 18.7111 3.1951 0.0010 0.0976
130 10.7504 2.1261 18.5230 3.2343 0.0010 0.0979
131 10.7674 2.1355 15.9319 2.9066 0.0010 0.0960
132 10.7451 2.1299 11.9354 2.2964 0.0010 0.0986
133 10.6533 2.1341 12.3813 2.3897 0.0010 0.1002
134 10.9910 2.1734 16.5868 3.1522 0.0010 0.0975
135 10.7916 2.1382 11.0472 2.2028 0.0010 0.0983
136 10.5851 2.1182 11.0833 2.1458 0.0010 0.0983
137 10.6278 2.1097 10.6915 2.1014 + 0.0010 0.1050
138 10.6092 2.1031 13.5398 2.6567 0.0010 0.0985
139 10.5511 2.0842 11.3438 2.2877 0.0010 0.0964
140 10.5511 2.0964 11.7257 2.2922 0.0010 0.0969
141 10.5219 2.0968 15.4311 2.8124 0.0010 0.0977
142 10.5396 2.0941 13.7698 2.5947 0.0010 0.0977
143 10.5058 2.0898 17.9566 3.1381 0.0010 0.0976
144 10.4883 2.0809 11.2515 2.2494 0.0010 0.0994
145 10.5158 2.1023 14.7135 2.6793 0.0010 0.0967
146 10.4078 2.0806 12.2493 2.3467 0.0010 0.0970
147 10.3714 2.0606 11.9672 2.3063 0.0010 0.1768
148 10.4436 2.0819 10.8371 2.1445 0.0010 0.0984
149 10.4166 2.0812 12.1619 2.4191 0.0010 0.0965
150 10.5072 2.0944 12.3933 2.2661 0.0010 0.0975
151 10.4302 2.0820 10.6376 2.0709 + 0.0010 0.0978
152 10.3692 2.0560 13.0043 2.4951 0.0010 0.0960
153 10.3992 2.0768 11.6122 2.2322 0.0010 0.0979
154 10.5164 2.0875 15.8421 2.8443 0.0010 0.0965
155 10.5096 2.0917 16.9917 3.1570 0.0010 0.1072
156 10.4363 2.0992 11.5760 2.3513 0.0010 0.0976
157 10.4430 2.0747 13.0481 2.3995 0.0010 0.0967
158 10.4114 2.0881 11.4282 2.3227 0.0010 0.0972
159 10.3397 2.0652 13.9916 2.6246 0.0010 0.0968
160 10.3568 2.0654 10.7009 2.0894 0.0010 0.0972
161 10.3551 2.0874 11.2556 2.2864 0.0010 0.0962
162 10.3279 2.0604 11.4774 2.3197 0.0010 0.0978
163 10.2556 2.0597 10.6614 2.0671 + 0.0010 0.0977
164 10.3278 2.0691 17.9811 3.1131 0.0010 0.0968
165 10.3651 2.0652 13.1818 2.6077 0.0010 0.0978
166 10.2216 2.0542 11.2597 2.2906 0.0010 0.0960
167 10.3857 2.0945 24.9576 4.0110 0.0010 0.0976
168 10.4270 2.0713 17.0359 3.1897 0.0010 0.0979
169 10.3028 2.0664 11.6389 2.3148 0.0010 0.0963
170 10.2629 2.0591 11.5156 2.3158 0.0010 0.0969
171 10.2682 2.0553 10.5681 2.0708 0.0010 0.1053
172 10.2060 2.0289 12.4935 2.5328 0.0010 0.1235
173 10.2312 2.0472 10.6320 2.0809 0.0010 0.1755
174 10.2038 2.0410 12.7901 2.3138 0.0010 0.0970
175 10.2220 2.0375 16.7133 2.9997 0.0010 0.0967
176 10.2629 2.0649 12.2742 2.4031 0.0010 0.0962
177 10.2818 2.0660 11.1812 2.2602 0.0010 0.0957
178 10.1738 2.0388 13.2913 2.6674 0.0010 0.0972
179 10.1454 2.0332 10.5103 2.0506 + 0.0010 0.0967
180 10.1075 2.0229 10.9831 2.1108 0.0010 0.0985
181 10.1103 2.0286 12.9929 2.5500 0.0010 0.0987
182 10.2273 2.0563 14.4476 2.7819 0.0010 0.0984
183 10.3006 2.0489 11.1331 2.1658 0.0010 0.0982
184 10.2019 2.0501 13.2543 2.6810 0.0010 0.0967
185 10.2433 2.0600 19.0089 3.3647 0.0010 0.0972
186 10.2751 2.0513 24.7641 3.9840 0.0010 0.0960
187 10.3142 2.0698 12.5326 2.4930 0.0010 0.1051
188 10.1317 2.0252 13.8861 2.7852 0.0010 0.0981
189 10.2103 2.0570 10.9316 2.1724 0.0010 0.0964
190 10.1055 2.0246 10.9014 2.1548 0.0010 0.0992
191 10.1341 2.0368 12.3572 2.5084 0.0010 0.0969
192 10.0761 2.0291 11.1435 2.2622 0.0010 0.0985
193 10.0840 2.0187 10.7761 2.1510 0.0010 0.0962
194 10.1373 2.0428 15.5093 2.9806 0.0010 0.0974
195 10.1754 2.0451 14.3928 2.8143 0.0010 0.0983
196 10.1540 2.0443 11.6802 2.3649 0.0010 0.0983
197 10.3802 2.0922 10.7879 2.1506 0.0010 0.0954
198 10.3887 2.0830 14.1898 2.5277 0.0010 0.0974
199 10.3347 2.0836 25.6714 3.9037 0.0010 0.1752
200 10.3367 2.0789 10.5787 2.1010 0.0010 0.0965
201 10.0463 2.0189 11.2732 2.1099 0.0010 0.0963
202 10.0429 2.0123 12.6029 2.5336 0.0010 0.0974
203 10.0345 2.0180 12.3066 2.3730 0.0010 0.1035
204 10.0725 2.0198 18.0473 3.1458 0.0010 0.0959
205 10.2650 2.0609 15.9492 2.9051 0.0010 0.0970
206 10.0590 2.0343 12.7779 2.5979 0.0005 0.0971
207 10.1405 2.0518 11.1314 2.1864 0.0005 0.0968
208 10.0398 2.0120 12.6483 2.4137 0.0005 0.0969
209 10.0545 2.0259 14.7432 2.9243 0.0005 0.0967
210 10.0426 2.0122 16.5693 2.9820 0.0005 0.1022
211 10.0309 2.0212 12.4197 2.5236 0.0005 0.1066
212 10.0322 2.0247 10.3571 2.0170 + 0.0005 0.1052
213 10.0299 2.0145 10.4687 2.0372 0.0005 0.1016
214 10.0551 2.0213 11.4687 2.2574 0.0005 0.1071
215 9.9964 2.0116 10.9590 2.2061 0.0005 0.1101
216 10.0055 2.0206 10.5415 2.0987 0.0005 0.1383
217 10.0039 2.0101 10.5187 2.0707 0.0005 0.1073
218 9.9858 2.0041 10.7047 2.0782 0.0005 0.1079
219 9.9855 2.0065 10.5367 2.0725 0.0005 0.1008
220 10.0492 2.0273 13.8437 2.7778 0.0005 0.1014
221 10.0196 2.0120 16.6357 2.9922 0.0005 0.1132
222 10.0316 2.0194 11.8387 2.4227 0.0005 0.1078
223 10.0487 2.0218 10.9232 2.1662 0.0005 0.1083
224 9.9952 2.0089 11.3090 2.1821 0.0005 0.1089
225 10.0200 2.0206 11.8820 2.4094 0.0005 0.1841
226 9.9997 2.0097 11.0449 2.1328 0.0005 0.0993
227 9.9902 2.0025 10.9042 2.1621 0.0005 0.1005
228 9.9940 2.0135 10.5860 2.0957 0.0005 0.0993
229 9.9460 1.9948 13.8859 2.6220 0.0005 0.1006
230 9.9939 2.0063 11.1663 2.2734 0.0005 0.0989
231 9.9522 2.0027 10.6127 2.1194 0.0005 0.0959
232 10.0346 2.0286 10.4554 2.0537 0.0005 0.0973
233 9.9554 2.0006 10.8173 2.1557 0.0005 0.0972
234 9.9768 2.0126 13.3576 2.6772 0.0005 0.0959
235 10.0369 2.0209 11.3771 2.2567 0.0005 0.1049
236 9.9747 2.0062 10.3786 2.0097 + 0.0005 0.0980
237 9.9411 1.9977 10.3196 2.0196 0.0005 0.0997
238 9.9379 2.0035 10.4509 2.0644 0.0005 0.0985
239 9.9418 1.9980 10.6072 2.1213 0.0005 0.1211
240 9.9556 1.9989 11.2521 2.1921 0.0005 0.0986
241 9.9709 2.0029 10.8331 2.1209 0.0005 0.0979
242 9.9977 2.0206 11.6271 2.3703 0.0005 0.0961
243 9.9816 2.0051 11.6028 2.2476 0.0005 0.0959
244 9.9832 2.0114 10.6162 2.0818 0.0005 0.0969
245 9.9652 2.0129 10.4408 2.0606 0.0005 0.0969
246 9.9912 2.0222 14.7667 2.9274 0.0005 0.0983
247 9.9915 2.0144 12.7221 2.4374 0.0005 0.0969
248 10.0175 2.0363 10.8693 2.1885 0.0005 0.0985
249 9.9691 2.0083 11.6561 2.3021 0.0005 0.0990
250 10.0141 2.0202 10.4279 2.0265 0.0005 0.1063
251 9.9495 2.0056 10.3374 2.0439 0.0005 0.1774
252 10.0045 2.0258 13.8486 2.7867 0.0005 0.0976
253 10.0667 2.0253 15.0957 2.7715 0.0005 0.0982
254 10.0643 2.0397 11.3925 2.3188 0.0005 0.0974
255 9.9743 2.0147 13.3847 2.5553 0.0005 0.0979
256 9.9358 1.9938 10.3613 2.0266 0.0005 0.0974
257 9.9704 2.0093 10.9700 2.2140 0.0005 0.0972
258 9.9994 2.0231 10.8503 2.1709 0.0005 0.0973
259 9.9250 2.0016 10.9251 2.0872 0.0005 0.0969
260 9.9697 2.0112 11.6003 2.3092 0.0005 0.0968
261 10.0126 2.0142 12.0197 2.2593 0.0005 0.0977
262 9.9637 2.0096 10.6802 2.1333 0.0005 0.0973
263 9.9887 2.0218 10.4781 2.0542 0.0003 0.0979
264 9.9253 2.0017 10.6417 2.1334 0.0003 0.0967
265 9.9038 1.9868 10.5920 2.1144 0.0003 0.0967
266 9.9451 2.0056 10.3308 2.0236 0.0003 0.1053
267 9.8958 1.9990 10.3029 2.0108 0.0003 0.0988
268 9.9277 2.0042 10.3527 2.0301 0.0003 0.0974
269 9.9342 2.0036 10.2913 2.0003 + 0.0003 0.0972
270 9.9623 2.0051 10.4179 2.0402 0.0003 0.0967
271 9.9495 2.0155 10.3165 2.0165 0.0003 0.0976
272 9.9215 2.0046 10.2776 2.0001 + 0.0003 0.0975
273 9.8989 1.9960 10.4992 2.0451 0.0003 0.0980
274 9.9241 2.0078 10.4035 2.0469 0.0003 0.0975
275 9.9012 1.9974 10.3583 2.0218 0.0003 0.0978
276 9.9218 2.0024 10.4039 2.0556 0.0003 0.0977
277 9.9308 2.0046 10.8659 2.1386 0.0003 0.0954
278 9.8990 1.9961 10.2940 1.9984 + 0.0003 0.1760
279 9.9578 2.0150 10.2932 2.0050 0.0003 0.0990
280 9.8998 1.9930 10.3219 2.0323 0.0003 0.0969
281 9.8907 1.9893 10.2723 2.0066 0.0003 0.0965
282 9.8944 1.9931 10.3314 2.0269 0.0003 0.1031
283 9.9090 2.0011 10.2808 2.0005 0.0003 0.0976
284 9.9638 2.0212 10.7474 2.1588 0.0003 0.0984
285 9.9082 2.0054 10.3627 2.0594 0.0003 0.0988
286 9.8999 1.9983 10.4395 2.0342 0.0003 0.0999
287 9.9051 1.9996 10.4167 2.0270 0.0003 0.0972
288 9.9367 2.0029 11.0638 2.2175 0.0003 0.0974
289 9.9510 2.0201 10.4041 2.0410 0.0003 0.0984
290 9.8862 1.9973 10.2883 1.9912 + 0.0003 0.0972
291 9.9335 2.0058 10.4414 2.0597 0.0003 0.0988
292 9.8801 1.9896 11.2034 2.1817 0.0003 0.1210
293 9.9612 2.0084 10.2912 1.9978 0.0003 0.0974
294 9.9149 2.0004 10.4739 2.0721 0.0003 0.0975
295 9.9201 2.0009 10.3878 2.0552 0.0003 0.0971
296 9.8749 1.9893 10.3235 2.0346 0.0003 0.0975
297 9.8962 2.0044 10.3851 2.0348 0.0003 0.1053
298 9.8853 1.9896 10.9170 2.1318 0.0003 0.0979
299 9.9470 2.0191 10.6505 2.1118 0.0003 0.0961
300 9.9054 2.0016 10.3083 2.0103 0.0003 0.0967
301 9.8884 1.9979 10.3231 2.0380 0.0003 0.0975
302 9.8951 2.0063 10.2794 2.0001 0.0003 0.0958
303 9.9217 2.0057 10.3672 2.0450 0.0003 0.0958
304 9.8987 2.0062 10.4386 2.0577 0.0003 0.1781
305 9.8900 1.9948 10.4546 2.0307 0.0003 0.0991
306 9.9696 2.0328 11.2850 2.3061 0.0003 0.0983
307 9.9196 1.9999 10.8129 2.1385 0.0003 0.0971
308 9.9653 2.0233 11.3015 2.2968 0.0003 0.0972
309 9.9826 2.0090 11.2653 2.1935 0.0003 0.0970
310 9.9207 2.0054 10.4656 2.0680 0.0003 0.0964
311 9.9065 1.9921 10.6278 2.0744 0.0003 0.0986
312 9.9461 2.0164 10.7961 2.1606 0.0003 0.0984
313 9.8878 1.9958 10.4509 2.0730 0.0003 0.0968
314 9.8795 1.9931 10.3534 2.0164 0.0003 0.0976
315 9.8701 1.9904 10.2612 2.0008 0.0003 0.0953
316 9.8853 1.9932 10.2719 2.0080 0.0003 0.1041
317 9.8927 1.9913 10.2841 2.0177 0.0001 0.0977
318 9.8730 1.9928 10.2693 2.0031 0.0001 0.0966
319 9.8797 1.9944 10.2654 2.0090 0.0001 0.0977
320 9.8604 1.9903 10.2589 2.0022 0.0001 0.0963
321 9.8712 1.9901 10.2800 2.0177 0.0001 0.0971
322 9.8859 1.9965 10.3494 2.0452 0.0001 0.0969
323 9.9518 2.0256 10.2760 2.0057 0.0001 0.0966
324 9.8958 1.9982 10.3993 2.0551 0.0001 0.0990
325 9.8887 1.9944 10.3061 2.0065 0.0001 0.1006
326 9.8660 1.9876 10.3734 2.0388 0.0001 0.1000
327 9.8825 1.9938 10.2654 2.0101 0.0001 0.0977
328 9.8930 1.9934 10.2830 1.9940 0.0001 0.0987
329 9.8861 2.0004 10.4599 2.0542 0.0001 0.0968
Stopping since valid_mae has not improved in the last 40 epochs.
Restoring best model from epoch 290.
Out[49]:
<class 'skorch.regressor.NeuralNetRegressor'>[initialized](
module_=BetterMLP(
(model): Sequential(
(0): Linear(in_features=27, out_features=32, bias=True)
(1): BatchNorm1d(32, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
(2): ReLU()
(3): Linear(in_features=32, out_features=1, bias=True)
)
),
)In a Jupyter environment, please rerun this cell to show the HTML representation or trust the notebook. On GitHub, the HTML representation is unable to render, please try loading this page with nbviewer.org.
<class 'skorch.regressor.NeuralNetRegressor'>[initialized](
module_=BetterMLP(
(model): Sequential(
(0): Linear(in_features=27, out_features=32, bias=True)
(1): BatchNorm1d(32, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
(2): ReLU()
(3): Linear(in_features=32, out_features=1, bias=True)
)
),
)In [50]:
y_pred = net.predict(X_test)
mean_absolute_error(y_test, y_pred)
Out[50]:
2.0315136909484863
In [51]:
history = net.history
plt.plot(history[:, 'train_mae'], label='train_mae')
plt.plot(history[:, 'valid_mae'], label='valid_mae')
ax1 = plt.gca() # get current axes
ax2 = plt.gca().twinx() # create a second y-axis
plt.plot(history[:, "event_lr"], label='learning_rate', color='gray', linestyle='--')
ax2.set_ylabel('Learning Rate')
ax1.set_ylabel('MAE')
for i, checkpoint in enumerate(history[:, "event_cp"][::-1]):
if checkpoint:
ax1.axvline(x=len(history[:, "event_cp"])-i-1, color='red',
linestyle='-.', label='Checkpoint', alpha=0.5)
break
ax1.legend(loc='lower left')
plt.xlabel('Epoch')
plt.show()
Exercise¶
Build an MLP to predict whether a patient has an in-hospital complication, based on features like BMI, age, etc.
- use a dummy baseline that always predicts 1
- use an sklearn MLP
- use a torch MLP
optimize for F1-score w.r.t. the positive class:
f1_score(y_pred, y_test) # no "weighted"
Plot the loss curves for the torch model
In [52]:
df = pd.read_csv("./Data/surgical_complications.csv")
df
Out[52]:
| bmi | Age | asa_status | baseline_cancer | baseline_charlson | baseline_cvd | baseline_dementia | baseline_diabetes | baseline_digestive | baseline_osteoart | ... | complication_rsi | dow | gender | hour | month | moonphase | mort30 | mortality_rsi | race | complication | |
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| 0 | 19.31 | 59.2 | 1 | 1 | 0 | 0 | 0 | 0 | 0 | 0 | ... | -0.57 | 3 | 0 | 7.63 | 6 | 1 | 0 | -0.43 | 1 | 0 |
| 1 | 18.73 | 59.1 | 0 | 0 | 0 | 0 | 0 | 0 | 0 | 0 | ... | 0.21 | 0 | 0 | 12.93 | 0 | 1 | 0 | -0.41 | 1 | 0 |
| 2 | 21.85 | 59.0 | 0 | 0 | 0 | 0 | 0 | 0 | 0 | 0 | ... | 0.00 | 2 | 0 | 7.68 | 5 | 3 | 0 | 0.08 | 1 | 0 |
| 3 | 18.49 | 59.0 | 1 | 0 | 1 | 0 | 0 | 1 | 1 | 0 | ... | -0.65 | 2 | 1 | 7.58 | 4 | 3 | 0 | -0.32 | 1 | 0 |
| 4 | 19.70 | 59.0 | 1 | 0 | 0 | 0 | 0 | 0 | 0 | 0 | ... | 0.00 | 0 | 0 | 7.88 | 11 | 0 | 0 | 0.00 | 1 | 0 |
| ... | ... | ... | ... | ... | ... | ... | ... | ... | ... | ... | ... | ... | ... | ... | ... | ... | ... | ... | ... | ... | ... |
| 14630 | 18.79 | 14.1 | 1 | 0 | 1 | 0 | 0 | 0 | 0 | 0 | ... | -0.54 | 1 | 0 | 7.78 | 0 | 1 | 0 | -0.16 | 1 | 1 |
| 14631 | 19.65 | 12.6 | 0 | 0 | 0 | 0 | 0 | 0 | 0 | 0 | ... | -1.42 | 4 | 0 | 8.40 | 6 | 1 | 0 | -0.77 | 1 | 1 |
| 14632 | 14.84 | 12.6 | 1 | 0 | 0 | 0 | 0 | 0 | 0 | 0 | ... | 0.65 | 0 | 0 | 13.25 | 3 | 3 | 0 | 0.99 | 1 | 1 |
| 14633 | 17.75 | 8.9 | 0 | 0 | 1 | 0 | 0 | 0 | 1 | 0 | ... | -0.50 | 0 | 1 | 8.30 | 5 | 0 | 0 | 0.17 | 1 | 1 |
| 14634 | 14.40 | 6.1 | 1 | 0 | 1 | 0 | 0 | 0 | 1 | 0 | ... | 0.78 | 2 | 0 | 7.65 | 4 | 1 | 0 | 1.06 | 0 | 1 |
14635 rows × 25 columns
In [53]:
target = "complication"
X = df.drop(columns=[target]).values.astype(np.float32)
y = df[target].values
X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=0.2, random_state=42)