Weights

Weights is a property of a NeuralNet object. The weights store the information the neural net has obtained during training. It is by default set to random values between -0.1 and 0.1. The RandomizeWeights and InitializeWeights functions can be used to modify the weights to other intervals.
The Weights list contains a list of connections where the source and destination neurons are given by numbers. The numbering of the neurons start with the first input neuron and go sequentially up to the last output neuron. The Bias are included in the numbering in each of the layers where they exist. The weights can be converted to edges in a weighted, directed graph with the NeuralNetGraph function.
The weights can be changed through this property, the connections can not.
The weights are abbreviated to a count in StandardForm when viewed as a property due to the large number that can be present. To view the weights specify InputForm, OutputForm or request only the Weights property as in net[Weights] where they are returned as a list.

Created by freegoldbar  (September 16, 2004)