ARTMAP_Create_Network: ARTMAP_Create_Network

In gbaquer3/fuzzyARTMAP: Implementation of a fuzzy ARTMAP neural network

Description

Creates a new ARTMAP network.

Usage

ARTMAP_Create_Network(numFeatures, numClasses, maxNumCategories = 1000,
  vigilance = 0.75, bias = 1e-06, numEpochs = 100, learningRate = 1)

Arguments

numFeatures	Number of features that the network expects of the input data. Must be a positive integer.
numClasses	Number of classes that exist for the supervisory signal. Must be a positive integer greater than 1.
maxNumCategories	Maximum number of categories that can be activated during the training process. Defaults to 1000.
vigilance	Vigilance parameter that defines the minimum similarity allowed between the input pattern and the weights. Defaults to 0.75.
bias	Constant that is used to differentiate between very similar category activation values. Defaults to 0.000001
numEpochs	Maximum number of training iterations allowed. Defaults to 100.
learningRate	Learning rate. Defaults to 1.

Details

This function creates a new ARTMAP network with the specified number features and classes. The network is created to expand the number of categories as needed. The vigilance parameter defaults to 0.75. The initial number of categories is set to 1. The maximum number of categories defaults to 100. The bias defaults to 0.000001, the number of epochs defaults to 100, and the learning rate defaults to 1.0 (fast-learning).

Value

Structure that holds all of the information for the network. It must be passed into both ARTMAP_LEARN() and ARTMAP_CLASSIFY(). The fields of this structure are numFeatures, numCategories, maxNumCategories, numClasses, weight (an initialized weight vector with the right dimensions), mapField (an initialized map field with the right dimensions), vigilance, bias, numEpochs, neededEpochs (the number of epochs needed in the last training) and learningRate.

gbaquer3/fuzzyARTMAP documentation built on Aug. 2, 2020, 6:33 p.m.