Autoencoder can be considered a data compression or feature extraction tool, since it can effectively compress the input to lower dimension without losing much information. However it's not generative, in the sense that if you feed the decoder with a Gaussian random vector, you won't get a output image resembling a sample in the dataset. This is because the distribution of the encoded vector is not controlled by the training process, so it's not necessarily Gaussian. If we want a specific distribution, we need to add a loss function that measures the shape of the distribution. In VAE, a KL loss is added on top of a reconstruction loss.
-
Notifications
You must be signed in to change notification settings - Fork 0
spotofleopard/keras_autoencoder
This commit does not belong to any branch on this repository, and may belong to a fork outside of the repository.
Folders and files
Name | Name | Last commit message | Last commit date | |
---|---|---|---|---|
Repository files navigation
About
Keras implementation of autoencoders
Topics
Resources
Stars
Watchers
Forks
Releases
No releases published
Packages 0
No packages published