Skip to content

joshua-hill/ConvolutionalSAEs

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

18 Commits
 
 
 
 
 
 

Repository files navigation

Convolutional Sparse Auto-Encoders

We train a convolutional variation of sparse auto-encoders on the activations of AlexNet. This architecture should be able to ~localize the prescence of features in an image (?).

We implement the encoder and decoders of the SAE as conv-nets. Our encoder is of dimension (c x d x 1 x 1), transforming the dimension of the activation to (B, D, H, W). The decoder is of dimension (d x c x 1 x 1).

Inspired by post by Apollo Research:

About

No description, website, or topics provided.

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published