Skip to content

Information bottleneck compression using PyTorch. 7th semester Mathematical Engineering (MATTEK) at Aalborg University

License

Notifications You must be signed in to change notification settings

LarsenAndreas/IB_compression

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

9 Commits
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

Information Bottleneck Principle Applied to PyTorch Autoencoder

The main idea is to use the amplitude spectrum to generate the probability distributions used for the cross entropy. Combine this with the MSE and we have a (unique) loss function.

Usage

For training the FMA dataset is used. To load this, a custom dataloader based on Librosa is utilised. To recreate the Python environment, use Miniconda (or Anaconda) with the requirements.txt file.

main.py contains the code to run the script. It is designed to be run from an editor, such as Spyder or Visual Studio Code, in order to change the parameters. The comments in the script should explain what the different parameters does and when to change them.

About

This code was written as part of a 7th semester project in Mathematical Engineering as Aalborg University, Denmark.

Code and associated paper is composed by Alexander F, Andreas L, Gustav Z, Mads J & Magnus L.

About

Information bottleneck compression using PyTorch. 7th semester Mathematical Engineering (MATTEK) at Aalborg University

Topics

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Languages