How to know what initial weights and biases were used by 'train' function?

Hi!

Is there any way to know what values of initial weights and biases were used by the 'train' function in the MATLAB neural network toolbox, if the net was created using 'patternnet'?

Thanks.

Answer 1

[Greg Heath]

1. Reset the random-number generator and weights (ideally zero) to the state that existed at the call of train

2. Set the training rate to it's lowest allowable value

3. Set the number of training epochs to 1

4. Either use getwb or net.IW, net.b and net.LW

Hope this helps.

Thank you for formally accepting my answer

Greg

[TS Sharma]

Thank you Greg! It seems there is no direct way to know the initial weights using, say, the tr structure returned by the train function.