Using DropoutLayer in neural network (not only in CNN)

hey guys,

i am using matlab to create a Neural Network for a Regression Problem.

to avoid overfitting i want to add a dropoutLayer after the Input layer, but i see only examples for CNN.

did someone knows, how to add a dropoutLayer in noraml neural Network or Setting the Options of neural Network(not CNN)

thanks

[Greg Heath]

OVERFITTING IS NOT "THE" PROBLEM!

"THE" PROBLEM IS

 "OVERTRAINING" an overfit net!!!

Hope this is clear!

Greg

[Mario Schweda]

I´m looking for the same. A droput layer for simple networks like fitnet. Not only for CNNs.

[Junwei Liu]

have you solve this problem？I have met the seem problem

Answer 1

[Moses Wayne]

The "dropoutLayer" class seems to be able to solve your problem. Calling it with no arguments returns a layer with 0.5 probability of dropping an input element. This layer can then be used in the training of your neural network. Here is a code example of creating a dropout layer with .6 probability of dropping an input element:

myLayer = dropoutLayer(0.6)

I have linked the documentation to the "dropoutLayer" class here.

[Andy_daodao]

hey, thanks for your answer.

i know the dropoutLayer calss

i create a network using net = network. how can i add this dropoutLayer into net. i have tried e.g. net.layers{1} = drpoutLayer(0.5), but it works not.

thanks

[Eman Marzban]

Did you find a solution to this problem?

Thanks

[Lingies Santhirasekaran]

Did anyone found a way to implement dropout in Matlab ?

Thanks

[belen celik]

how can dropoutLayer be implemented in new=newff() type of network?