ltbio.ml.supervised.SupervisingTrainer

Overview

Classes

SupervisingTrainer

A Single Pipeline Unit is any agent that can act (use, process or make changes) to a collection (usually of Timeseries).

Contents

class ltbio.ml.supervised.SupervisingTrainer.SupervisingTrainer(model: ltbio.ml.supervised.models.SupervisedModel.SupervisedModel, train_conditions: Collection[ltbio.ml.supervised.SupervisedTrainConditions.SupervisedTrainConditions], evaluation_metrics: Collection = None, name: str = None, save_report_to: str = None)

Bases: ltbio.pipeline.PipelineUnit.SinglePipelineUnit

A Single Pipeline Unit is any agent that can act (use, process or make changes) to a collection (usually of Timeseries). Following the Command design pattern, a SingleUnit is the abstract ‘Command’, so that Pipeline can execute various kinds of processing by calling the ‘apply’ method of each concrete unit.

E.g. Filter, Segmenter, FeatureExtractor, FeatureSelector, SupervisingTrainer, DecisionMaker Any subclass that implements ‘apply’.

Every subclass must define ‘apply’ and implement a concrete behaviour. To map the parameters’ names of ‘apply’ to the labels inside any arriving Packet, PIPELINE_INPUT_LABELS should be defined. To map the outputs to the labels of the resulting Packet, PIPELINE_OUTPUT_LABELS should be defined.

PIPELINE_INPUT_LABELS Maps every label of a needed input inside a Packet to the name of the corresponding ‘apply’ parameter. PIPELINE_OUTPUT_LABELS Maps every output name of ‘apply’ to a label to be saved inside a Packet.

ART_PATH = 'resources/pipeline_media/ml.png'

PIPELINE_INPUT_LABELS

PIPELINE_OUTPUT_LABELS

apply(dataset: ltbio.ml.datasets.BiosignalDataset.BiosignalDataset, test_dataset: ltbio.ml.datasets.BiosignalDataset.BiosignalDataset = None)