Towards Reproducible Learning-based Compression

TL;DR

This paper addresses the reproducibility issues in deep learning-based compression systems caused by implementation differences, proposing a safeguarding mechanism to ensure bounded mismatches and improve robustness across devices.

Contribution

It introduces a scalable safeguarding mechanism that bounds mismatches in deep learning-based compression, enhancing reproducibility and robustness across different hardware and software environments.

Findings

Effective protection at reconstruction or decoding levels

Reduced overhead when error bounds are suppressed

Improved reproducibility in image and point cloud compression

Abstract

A deep learning system typically suffers from a lack of reproducibility that is partially rooted in hardware or software implementation details. The irreproducibility leads to skepticism in deep learning technologies and it can hinder them from being deployed in many applications. In this work, the irreproducibility issue is analyzed where deep learning is employed in compression systems while the encoding and decoding may be run on devices from different manufacturers. The decoding process can even crash due to a single bit difference, e.g., in a learning-based entropy coder. For a given deep learning-based module with limited resources for protection, we first suggest that reproducibility can only be assured when the mismatches are bounded. Then a safeguarding mechanism is proposed to tackle the challenges. The proposed method may be applied for different levels of protection either at the reconstruction level or at a selected decoding level. Furthermore, the overhead introduced for the protection can be scaled down accordingly when the error bound is being suppressed. Experiments demonstrate the effectiveness of the proposed approach for learning-based compression systems, e.g., in image compression and point cloud compression.