Simple tutorial implementing a gaussian information bottleneck based on the great paper from Gal Chechik, Amir Globerson, Naftali Tishby, and Yair Weiss (JMLR 2005).
Starting from a joint Gaussian (X,Y), the problem is to find a linear projection A that compresses a variable X while preserving as much information as possible about Y.
The tradeoff between compression and preservation is given by a parameter
michnard/GIB_tutorial
Simple tutorial implementing a gaussian information bottleneck
Jupyter NotebookGPL-3.0