Online Conformal Compression for Zero-Delay Communication with Distortion Guarantees

2025-03-11

AI Summary

This paper addresses zero-delay online lossy source compression, aiming to provide deterministic, per-sequence distortion (outage) upper bounds under the 0–1 distortion metric at any time, without assumptions on the source distribution or reliance on the quality of pretrained predictors. To this end, we introduce online conformal prediction into lossy compression for the first time, proposing a predictor-driven dynamic confidence set construction and symbol subset selection mechanism. Our approach eliminates the dependence of classical rate-distortion theory on statistical models and hindsight knowledge, enabling fully online, prior-free real-time encoding. We theoretically establish that, under strict outage constraints, our method achieves a compression rate approaching that of an ideal hindsight encoder. Experiments demonstrate that our method significantly outperforms classical model-free compression schemes in rate performance.

Technology Category

• Machine Learning: Online Learning & Bandits
• Search and Optimization: Learning to Search
• Cognitive Modeling & Cognitive Systems: Neural Spike Coding

Application Category

• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Graph Algorithms and Modeling for the Web: Algorithms and analysis for incomplete, noisy, or partially observed Web-related graphs
• User Modeling, Personalization and Recommendation: User modeling for targeted and personalized online advertising

Abstract

We investigate a lossy source compression problem in which both the encoder and decoder are equipped with a pre-trained sequence predictor. We propose an online lossy compression scheme that, under a 0-1 loss distortion function, ensures a deterministic, per-sequence upper bound on the distortion (outage) level for any time instant. The outage guarantees apply irrespective of any assumption on the distribution of the sequences to be encoded or on the quality of the predictor at the encoder and decoder. The proposed method, referred to as online conformal compression (OCC), is built upon online conformal prediction--a novel method for constructing confidence intervals for arbitrary predictors. Numerical results show that OCC achieves a compression rate comparable to that of an idealized scheme in which the encoder, with hindsight, selects the optimal subset of symbols to describe to the decoder, while satisfying the overall outage constraint.

Problem

• Ensures deterministic distortion bounds in lossy compression.
• Applies to arbitrary sequence distributions and predictor qualities.
• Achieves compression rates comparable to optimal hindsight schemes.

Innovation

• Uses pre-trained sequence predictors for compression
• Ensures deterministic distortion guarantees per-sequence
• Applies online conformal prediction for confidence intervals