Dropout Neural Network for Classification

Hello,

I'm studying neural network and i want to know how to do dropout in Matlab for classification. There is something called dropout class, but not much information about that in Matlab. I would like to know if it is possible to apply dropout in neural network in Matlab, and what steps should be done? As far as i know, i have to manipulate the neurons inside hidden layer, however the nntools in Matlab looks like cannot be manipulated. I am trying to create custom neural network, but still i am confused how to create and apply dropout. Unfortunately, i couldn't find any information regarding to this in the internet.

Thank you

[Emma Reeves]

I have the same problem. Have you found any answer yet?

[Greg Heath]

I have been designing classifiers for more than 25 years and have never heard of the term "drop out class". If you can explain to me what it means, I can probably tell you how to deal with it using the MATLAB NN Toolbox.

Hope this helps.

Greg

Answer 1

[Greg Heath]

Drop out is explained in the MATLAB website. Input values are randomly set to 0.0 with a specified probability. This is a mitigation for overtraining an overfit net.

I have never used it. So, cannot give any advice except to first try it using MATLAB NN example data obtained from MATLAB help and documentation

 help nndatasets
 doc nndatasets

If you try it, post results.

Hope this helps.

Thank you for formally accepting my answer

Greg