Reliable Estimation of KL Divergence using a Discriminator in Reproducing Kernel Hilbert Space

Publication
In Neural Information Processing System (NeurIPS) 2021

Summary

This paper answers the fundamental question about estimating KL divergence from samples. We argues that the estimation is unreliable if the complexity of discriminator is not controlled during training. We show how to do that in an RKHS space with theoretical analysis.

Sandesh Ghimire
Postdoctoral Researcher

My research interests include machine learning, computer vision and medical imaging.