pyrregular.models.borf

BORF pipelines. Bag-Of-Receptive-Fields features (fast_borf) with a LightGBM classifier. iborf_pipeline is the time-aware variant for irregular series (ridge head).

Module Attributes

borf_pipeline	X from to_dense().
iborf_pipeline	This pipeline applies time-aware BORF (I-BORF) → arcsinh → RidgeClassifierCV, as in Spinnato, "A Time-Aware Bag-of-Receptive-Fields for Interpretable Irregular Time Series Classification".

pyrregular.models.borf.borf_pipeline = Pipeline(steps=[('borf', BORF()),                 ('functiontransformer',                  FunctionTransformer(func=<function _to_float>)),                 ('lgbmclassifier', LGBMClassifier(n_jobs=1))])

X from to_dense().

Type:

This pipeline applies BORF → to_float → LGBMClassifier. Input

pyrregular.models.borf.iborf_pipeline = Pipeline(steps=[('borf',                  BORF(min_window_to_signal_std_ratio=0.15, time_channel=True)),                 ('functiontransformer',                  FunctionTransformer(func=<ufunc 'arcsinh'>)),                 ('_ridgeclassifiercvfix', _RidgeClassifierCVFix())])

This pipeline applies time-aware BORF (I-BORF) → arcsinh → RidgeClassifierCV, as in Spinnato, “A Time-Aware Bag-of-Receptive-Fields for Interpretable Irregular Time Series Classification”. Input: X from to_dense(concatenate_time=True, normalize_time=True), where the last channel holds the timestamps.