Core Concepts
There is an interesting interplay between source distortion (distortion for the probability vector measured via f-divergence) and the subsequent channel encoding/decoding parameters, and a joint design of these parameters is crucial to navigate the latency-distortion tradeoff when communicating classifier decisions over noisy channels.
Abstract
The paper considers the problem of communicating the decisions of a classifier (represented as a probability vector) over a noisy channel. The goal is to study the tradeoff between transmission latency and the distortion between the original probability vector and the reconstructed one at the receiver, where the distortion is measured using f-divergence.
The key highlights and insights are:
The authors analyze this tradeoff using uniform, lattice, and sparse lattice-based quantization techniques to encode the probability vector. They first characterize the bit budgets for each technique given a requirement on the allowed source distortion.
These bounds are then combined with results from finite-blocklength literature to provide a framework for analyzing the effects of both quantization distortion and distortion due to decoding error probability (i.e., channel effects) on the incurred transmission latency.
The results show that there is an interesting interplay between source distortion and the subsequent channel encoding/decoding parameters, and indicate that a joint design of these parameters is crucial to navigate the latency-distortion tradeoff.
The impact of changing different parameters (e.g., number of classes, SNR, source distortion) on the latency-distortion tradeoff is studied, and experiments are performed on AWGN and fading channels.
The authors find that sparse lattice-based quantization is the most effective at minimizing latency for low end-to-end distortion requirements across different parameters and works best for sparse, high-dimensional probability vectors (i.e., high number of classes).
Stats
"In recent years, machine learning (ML) has been increasingly applied to time-sensitive applications, including Vehicle to Vehicle (V2V) and Vehicle to Infrastructure (V2I) communications."
"These applications require reliable and rapid data transmission for tasks such as trajectory prediction [2] and lane change detection [3]."
Quotes
"Semantic communication generally focuses on sending context dependent features/decisions dependent on the data to the receiver (rather than the entire raw message) [7]."
"URLLC is to design protocols in order to transmit low-data rate (short packets) with high reliability (low probability of error) within a small latency [9]."