unfortunately such implementation leads to NaN when Adam is used (many training steps needed).
I added a variational autoencoder example in Keras: https://github.com/fchollet/keras/blob/master/examples/variational_autoencoder.py …pic.twitter.com/hxgf48MUqD
-
-
-
Not sure if other optimizers have same defect.
End of conversation
New conversation -
-
-
any chance there will be a blog post to go with this?
-
sure, I'm actually preparing one.
- Show replies
New conversation -
-
-
Looking forward to the blog post. Not seeing how vae is used after fit. (Also, why encoder, generator don't need a compile.)
-
vae isn't used after fit, only decoder/generator are (same weights). No need to compile to only use "predict".
- Show replies
New conversation -
-
-
any plan for (robust or else) autoencoder LSTMs?
-
what would be a good dataset for that, though?
- Show replies
New conversation -
Loading seems to be taking a while.
Twitter may be over capacity or experiencing a momentary hiccup. Try again or visit Twitter Status for more information.