is it possible to touch the weights & bias in neural network toolbox

Question

Lin Bai 2018-2-14

0
链接

此问题的直接链接

https://ww2.mathworks.cn/matlabcentral/answers/382678-is-it-possible-to-touch-the-weights-bias-in-neural-network-toolbox

评论： Lin Bai 2018-2-14

i am right now using neural network toolbox. and would love to do something on weights during training, like apply the saturation function on weights. but cannot find any helpful solution could anybody give any solution for this?

0 个评论
显示 -2更早的评论隐藏 -2更早的评论

请先登录，再进行评论。

请先登录，再回答此问题。

Answer 1

Greg Heath 2018-2-14

0
链接

此回答的直接链接

https://ww2.mathworks.cn/matlabcentral/answers/382678-is-it-possible-to-touch-the-weights-bias-in-neural-network-toolbox#answer_305160

在 MATLAB Online 中打开

% OMIT THE ENDING SEMICOLONS IN THE FOLLOWING 2 COMMANDS

 net = fitnet
 TRAINPARAM = net.trainParam

% You can directly modify every property listed. For example

net.trainParam.mu = 0.002;

% Now check to se if mu is changed

Hope this helps.

Thank you for formally accepting my answer

Greg

1 个评论
显示 -1更早的评论隐藏 -1更早的评论

Lin Bai 2018-2-14

在 MATLAB Online 中打开

Thank you Greg, but this looks like how to get access to the weights/biases. My problem is to do something during training.

I used the following code to train, seems i have no access to the weights during training.

 layers = [
    imageInputLayer([28 28 1])
    convolution2dLayer(3,16,'Padding',1)
    batchNormalizationLayer
    reluLayer   
    maxPooling2dLayer(2,'Stride',2)
    convolution2dLayer(3,32,'Padding',1)
    batchNormalizationLayer
    *myLayer(change weight when back propagation)*
    reluLayer...]; 
options = trainingOptions('sgdm',...
    'MaxEpochs',6, ...
    'ValidationData',{valImages,valLabels},...
    'ValidationFrequency',30,...
    'Verbose',false,...
    'Plots','training-progress');
net = trainNetwork(trainImages,trainLabels,layers,options);