Reference to "adaptwb" function in Neural Network toolbox.

I applied the "adaptwb" function to the implementation of my algorithm. Thus, I want to describe the mechanism of this function in my paper, but the detail of this function is not provided in the technical document of MATLAB. So, I wonder if there are any references to this function.

Answer 1

[Nick Hobbs]

I understand you want to determine how 'adaptwb' is updating the weights for the neural network. If you make a quick neural network with

    net = feedforwardnet(20)

you will get an output of information for the network. From there, if you click on 'layerWeights' a help page will appear. From there you can click 'learnFcn', which describes that the parameter is used to define the learning function for 'adaptwb'. You can also see the list of learning functions by clicking on 'learning function', which will provide you a list of learning functions for networks. To determine which learning function your network uses you should be able to use

    net.layerWeights{i,j}.learnFcn

where i and j are for the values provided by net.layerWeights.

This same procedure can be used for 'inputWeights' and 'biases' to determine how these learn.

I hope this information helps to guide you on how to determine how your network is adapting with adaptwb.

Answer 2

[Greg Heath]

Have you tried the 3 command line documentation sources?

 help adapt

 doc adapt

 type adapt

Hope this helps.

Thank you for formally accepting my answer

Greg

[Alvaro Macias]

Hi, I have tried those commands and tried to go deeper on the Neural Network Adapt Functions.

In the nnproperty.net_adaptFcn said that the nnadapt just contains adaptwb as the only option. Inside the doc said that adaptwb is a "Sequential order incremental training w/learning functions" if we assume that it is similar to trains, that's mean that the algoritms will be similiar "Each weight and bias is updated according to its learning function after each time step in the input sequence".

I want to know if adaptwb uses the same learning function as trians. In that case the adaptation can be done with all the function listed in nnlearn. Is it that correct?

Best regards,

Alvaro

[Alvaro Macias]

Finally I found that adapt uses the learn function learnFcn and train uses the train function trainFcn.

net.biases{i}.learnFcn

This property defines which of the learning functions is used to update the th layer's bias vector (net.b{i}) during training, if the network training function is trainb, trainc, or trainr, or during adapt, if the network adapt function is trains. For a list of functions, type help nnlearn [Matlab].

Hope this helps others