Freeze specific weights in shallow neural network

4 次查看（过去 30 天）

M. 2022-3-8

0
链接

此问题的直接链接

https://ww2.mathworks.cn/matlabcentral/answers/1666914-freeze-specific-weights-in-shallow-neural-network

编辑： M. 2022-3-8

Dear forum members,

I was wondering if it is possible to initialize specific weights in a custom shallow neural network created with the network function, and not update them during training ?

I have seen these answers :

https://mathworks.com/matlabcentral/answers/311407-freez-specific-weights-in-custom-neural-network

But it seems restricted to deep neural networks.

Is there something similar with the network function ?

I have not seen anything related in this documentation : https://mathworks.com/help/deeplearning/ref/network.html?s_tid=srchtitle_network_2

Let's say I have something like this:

numInputs = nI;
numLayers = nL;
biasConnect = ones(numLayers,1);
inputConnect = ones(numInputs,1);
layerConnect = ones(numLayers,numLayers);
outputConnect = [zeros(1,numLayers-1) 1];
net = network(numInputs,numLayers,biasConnect,inputConnect,layerConnect,outputConnect);