Taming numerical imprecision by adapting the KL divergence to negative probabilities

Simon Pfahler¹, Peter Georg¹, Rudolf Schill², Maren Klever³, Lars Grasedyck³, Rainer Spang⁴, Tilo Wettig¹

¹ Department of Physics, University of Regensburg
² Department of Biosystems Science and Engineering, ETH Zürich
³ Institute for Geometry and Applied Mathematics, RWTH Aachen University
⁴ Department of Informatics and Data Science, University of Regensburg

Correspondence: simon.pfahler@ur.de

Poster

Paper

A paper explaining the method as well as the application in more detail is available here:
https://link.springer.com/article/10.1007/s11222-024-10480-y
https://arxiv.org/abs/2312.13021

Code availability

pRC: C++ library for efficient calculations in the Tensor Train format, open source

cMHN: C++ library to work with Mutual Hazard Networks in the Tensor Train format, soon to be open source

Funding