AITIP(Information Theoretic Inequality Prover) is an online service that automatically proves or disproves information theory inequalities in the form of entropy, joint entropy and mutual information. Such problems are not only important in Information Theory, but they also arise in fields like Machine Learning and Cryptography.
For information inequalities involving more than a handful of random variables, it is tedious and sometimes infeasible to construct the proof or disproof by hand, and our web service automates this process. It can also be used as an educational tool for teaching and learning Information Theory.
As described in Proving and Disproving Information Inequalities (ISIT 2014), the proof or disproof of an information inequality can be constructed by solving a linear programming (LP) problem. In this web service, we develop a scalable approach to solve the LP using ADMM, a first-order method that makes the system highly parallelizable to cloud computing implementation. Our follow-up paper describing this new approach will appear in ISIT 2019.
Teletraffic Research Centre, University of Adelaide, Australia
Choh-Ming Li Professor, Department of Information Engineering, Chinese University of Hong Kong
Associate Professor, Department of Computer Science, City University of Hong Kong
PhD Student, Department of Computer Science, City University of Hong Kong