FaceChannel.SelfAffectiveMemory Model Definition

FaceChannel.SelfAffectiveMemory.SelfAffectiveMemory module

SelfAffectiveMemory.py

Self-Affective memory model.

class FaceChannel.SelfAffectiveMemory.SelfAffectiveMemory.SelfAffectiveMemory(numberOfEpochs=5, insertionThreshold=0.9, learningRateBMU=0.35, learningRateNeighbors=0.76)

Bases: object

buildAffectiveMemory(dataTrain)

Method that activey builds the current affective memory

Parameters

dataTrain – initial training data as an ndarray.

getNodes()

Method that returns all the current nodes of the affective memory :return: a tuple of nodes and ages of each node. :rtype: ndarray tuple

insertionThreshold = 0.9

Activation threshold for node insertion

learningRateBMU = 0.35

Learning rate of the best-matching unit (BMU)

learningRateNeighbors = 0.76

Learning rate of the BMU’s topological neighbors

numberOfEpochs = 5

Number of traning epoches for the GWR

predict(images, preprocess=False)

Method that predicts the current arousal and valence of a given image or set of images.

as the affective memory is an online learning method, every given frame must be temporaly subsequent to the previous ones. It relies on the FaceChannelV1 for feature extraction.

Parameters

• images – The images as one or a list of ndarray.

• preprocess – If the image is already pre-processed or not. a pre-processed image has a format of (64,64,1).

Returns

The prediction of the given image(s) as a ndarray

Return type

ndarray

train(dataPointsTrain)

Method that trains the affective memory online :param dataTrain: initial training data as an ndarray.