Technical Program

Paper Detail

Paper IDG.5.3
Paper Title Structure of Optimal Quantizer for Binary-Input Continuous-Output Channels with Output Constraints
Authors Thuan Nguyen, Thinh Nguyen, Oregon State University, United States
Session G.5: Signal Processing
Presentation Lecture
Track Graphs, Games, Sparsity, and Signal Processing
Manuscript  Click here to download the manuscript
Virtual Presentation  Click here to watch in the Virtual Symposium
Abstract In this paper, we consider a channel whose the input is a binary random source $X \in \{x_1,x_2\}$ with the probability mass function (pmf) $\textbf{p}_X = [p_{x_1}, p_{x_2}]$ and the output is a continuous random variable $Y \in \mathbb{R}$ as a result of a continuous noise, characterized by the channel conditional densities $p_{y|x_1}=\phi_1(y)$ and $p_{y|x_2}=\phi_2(y)$. A quantizer $Q$ is used to map $Y$ back to a discrete set $Z \in \{z_1, z_2, \dots, z_N\}$. To retain most amount of information about $X$, an optimal $Q$ is one that maximizes $I(X;Z)$. On the other hand, our goal is not only to recover $X$ but also ensure that $\textbf{p}_Z = [p_{z_1}, p_{z_2}, \dots, p_{z_N}]$ satisfies a certain constraint. In particular, we are interested in designing a quantizer that maximizes $\beta I(X;Z) - C(\textbf{p}_Z)$ where $\beta$ is a trade-off parameter and $C(\textbf{p}_Z)$ is an arbitrary cost function of $\textbf{p}_Z$. Let the posterior probability $p_{x_1|y} = r_y=\dfrac{p_{x_1} \phi_1(y)}{p_{x_1} \phi_1(y) + p_{x_2} \phi_2(y)}$, our result shows that the structure of the optimal quantizer separates $r_y$ into convex cells. In other words, the optimal quantizer has the form: $Q^*(r_y)= z_i, \text{ if } a^*_{i-1} \leq r_y < a^*_i,$ for some optimal thresholds $a^*_0=0 < a^*_1 < a^*_2< \dots < a^*_{N-1} < a^*_N = 1.$ Based on this optimal structure, we describe some fast algorithms for determining the optimal quantizers.

Plan Ahead


2021 IEEE International Symposium on Information Theory

11-16 July 2021 | Melbourne, Victoria, Australia

Visit Website!