On Optimal Zero-Delay Coding of Vector Markov Sources

TL;DR

This paper investigates the optimal design of zero-delay vector quantizers for Markov sources, establishing existence and structure of optimal policies under various conditions using a stochastic control framework.

Contribution

It provides new existence and structural results for optimal zero-delay quantization policies for vector Markov sources, including finite and infinite horizon cases.

Findings

Optimal policies exist for finite-horizon problems with convex codecells.

Deterministic Markov policies are optimal for stationary infinite-horizon problems.

Shared randomization can be used to achieve optimal stationary policies.

Abstract

Optimal zero-delay coding (quantization) of a vector-valued Markov source driven by a noise process is considered. Using a stochastic control problem formulation, the existence and structure of optimal quantization policies are studied. For a finite-horizon problem with bounded per-stage distortion measure, the existence of an optimal zero-delay quantization policy is shown provided that the quantizers allowed are ones with convex codecells. The bounded distortion assumption is relaxed to cover cases that include the linear quadratic Gaussian problem. For the infinite horizon problem and a stationary Markov source the optimality of deterministic Markov coding policies is shown. The existence of optimal stationary Markov quantization policies is also shown provided randomization that is shared by the encoder and the decoder is allowed.