Reading Group: Information-Theoretic & PAC-Bayes Generalization

[Grayson York]

Hi folks, 

Mohammad Tinati is starting a reading group which may be of interest to several members of this group. The first meeting will be next Tuesday 11/4 at 10:00AM in EE403.  The description of the information which will be covered is as follows:

Goal: study generalization bounds for learning algorithms using information theoretic and PAC-Bayesian tools. We will read and reconstruct proofs for mutual/conditional mutual information bounds, change-of-measure techniques, and PAC-Bayes bounds, starting from the references below and extending as needed.

A more detailed description is attached.

Thanks,

Grayson