How to display weight distribution in hidden layers of neural network?

Question

Machine Learning Enthusiast 2017-9-16

0
链接

此问题的直接链接

https://ww2.mathworks.cn/matlabcentral/answers/356966-how-to-display-weight-distribution-in-hidden-layers-of-neural-network

回答： Greg Heath 2017-9-17

image.png

I have 8 inputs in the input layer.Now i want to display weight distribution of these 8 inputs in hidden layer to observe the importance of features.To make it more clear example is shown in figure ( https://pasteboard.co/GKCpA6Q.png ).I used `plotwb` function of Matlab it didn't display the weights of every input.

Actually i want to look at weights(weights connecting inputs to first hidden layer) . Larger the weight is, the more important the input.

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 2017-9-17

0
链接

此回答的直接链接

https://ww2.mathworks.cn/matlabcentral/answers/356966-how-to-display-weight-distribution-in-hidden-layers-of-neural-network#answer_281926

在 MATLAB Online 中打开

That will not work. It does not account for the correlations between inputs.

The best way to rank correlated inputs is

 1. Use NO HIDDEN LAYERS !
 2. Run 10 or more trials each (different random initial weights) 
   using
   a. A single input
   b. All inputs except the one in a.

Hope this helps.

Thank you for formally accepting my answer

Greg