Combinatorial joint source-channel coding

Abstract

Traditional error correction and source coding has focused on the stochastic setting where separation based schemes are optimal, and current solutions for applications requiring both lossy compression and noise resilience reflect this approach. However, in the adversarial setting, with worst case errors, separation based schemes are far from being even asymptotically optimal. This work investigates fundamental limits, achievability and converse bounds, practical codes, and algorithms for joint source channel coding (JSCC) in the adversarial setting. Particular attention is paid to the cases of flip and erasure errors.