avalanche.evaluation.metrics.MultiStreamAMCA

class avalanche.evaluation.metrics.MultiStreamAMCA(classes=None, streams=None)[source]

An extension of the Average Mean Class Accuracy (AMCA) metric (class:AverageMeanClassAccuracy) able to separate the computation of the AMCA based on the current stream.

__init__(classes=None, streams=None)[source]

Creates an instance of a MultiStream AMCA.

Parameters
  • classes – The list of classes to track. This has the same semantic of the classes parameter of class AverageMeanClassAccuracy.

  • streams – The list of streams to track. Defaults to None, which means that all stream will be tracked. This is not recommended, as you usually will want to track the “test” stream only.

Methods

__init__([classes, streams])

Creates an instance of a MultiStream AMCA.

finish_phase()

Moves to the next phase.

reset()

Resets the metric.

result()

Retrieves the running AMCA for each stream.

set_stream(stream_name)

Switches to a specific stream.

update(predicted_y, true_y, task_labels)

Update the running accuracy given the true and predicted labels for each class.