Inference with Communication Constraints: Hypothesis Testing and Delay Estimation

Speaker

Yuval Kochman
Professor at Hebrew University of Jerusalem.

Title

"Inference with Communication Constraints: Hypothesis Testing and Delay Estimation"

Abstract

In many sensing scenarios, nodes observe correlated data but face strict communication limits. This talk explores two canonical problems in distributed inference under such constraints, with an emphasis on the fundamental information-theoretic limits.

First, we consider a binary hypothesis testing problem where two sensors observe sequences of i.i.d. standard Gaussian variables that are jointly Gaussian under both hypotheses, differing only in their correlation. Communication is one-way: a single message is sent from one sensor to the other, which makes the final decision. While the setup is conceptually simple, the optimal error exponent under communication constraints is still unknown. We present new bounds on this exponent and reveal that a surprisingly simple strategy—transmitting the index of the maximum sample—can match the asymptotic performance of vector-quantization schemes.

The second part of the talk considers distributed delay estimation, motivated by direction-of-arrival (DoA) problems. Two sensors again observe correlated Gaussian sequences, now with a fixed correlation structure and an unknown relative shift. We analyze the scaling regime where both the number of possible delays and the communication budget grow. We characterize when consistent estimation is possible and show that, as in the previous problem, maximum-index-based schemes can achieve optimal performance.

Joint work with Ligong Wang (ETH Zurich), Amir Weiss (Bar-Ilan University), and Gregory W. Wornell (MIT).

About Speaker

Yuval Kochman received his B.Sc., M.Sc., and Ph.D. degrees in Electrical Engineering from Tel Aviv University, all with honors. He was a Postdoctoral Associate at MIT’s Signals, Information, and Algorithms Laboratory from 2009 to 2011. Since 2012, he has been with the School of Computer Science and Engineering at the Hebrew University of Jerusalem. His background includes industry work in radar and digital communications, and his research focuses on information theory, communications, and signal processing.