Dropout regularisation using Permutations and Combinations?


I was wondering if we could compare the best results of the Dropout regularisation using Permutations and Combinations instead of randomly turning off certain neurones.

Can this be done and is it effective?


I’m not sure what you mean by applying permutations to implement dropout. Can you give a bit more of a description of your idea?

But it is always the case that if you have a new idea, you can try it and see what happens. Of course the first step there is figuring out how to actually implement the concept that you have in mind. :nerd_face:

By Permutations and Combinations, I mean analysing the output of certain number of activated neurones to choose the most efficient regularisation.

I haven’t tried this yet or haven’t researched a lot about it either. However, I’m very curious to see how this Mathematical concept can help in regularisation.