0

I read a lot of math explanations on this here which I understood nothing. Can anyone provide a simple code example?

I am looking for a keras sequential model example.

Thanks for all the answers.

Leon Rai
  • 1,401
  • 3
  • 9
  • 15
  • Keras already has an implementation of Dropout, what's the problem with using it? – Dr. Snoopy Jan 29 '19 at 18:48
  • I was asking about inverse dropout. I couldn't find anything. – Leon Rai Jan 29 '19 at 20:25
  • What I mean is that Keras already implements inverted Dropout. See https://stackoverflow.com/questions/38592943/is-the-keras-implementation-of-dropout-correct – Dr. Snoopy Jan 29 '19 at 20:33
  • I have seen that one, what I understand is that you have to manually reshape the output at every layer? I thought there could be a function as simple as "model.add(InverseDropout(x,y)". I am not sure how to use it with sequential model. – Leon Rai Jan 29 '19 at 21:04
  • I have no idea what you are talking about, there is no reshaping needed, inverted dropout is already the default implementation of Dropout, you just need to do model.add(Dropout(0.5)) – Dr. Snoopy Jan 29 '19 at 21:05
  • #encoder layers: dense() dropout(.5) conv2d #decoder layers: deconv2d dropout(.5) dense() should be like this? – Leon Rai Jan 29 '19 at 22:17
  • 1
    You didn't mention that you were implementing an autoencoder, so you are trying to find an actual inverse to what Dropout does. There is no inverse and you don't actually need one. – Dr. Snoopy Jan 29 '19 at 22:35
  • Let us [continue this discussion in chat](https://chat.stackoverflow.com/rooms/187530/discussion-between-leon-rai-and-matias-valdenegro). – Leon Rai Jan 29 '19 at 22:43

0 Answers0