Lossy Compression for Lossless Prediction

Yann Dubois, Benjamin Bloem-Reddy, Karen Ullrich, Chris J. Maddison

Keywords: Compression, Invariances, Information Theory, Machine Learning, Self-Supervised Learning

TL;DR: We formalize and experimentally evaluate compression that guarantees high downstream predictive performance under invariances.

Abstract: Most data is automatically collected and only ever "seen" by algorithms. Yet, data compressors preserve perceptual fidelity rather than just the information needed by algorithms performing downstream tasks. In this paper, we characterize the bit-rate required to ensure high performance on all predictive tasks that are invariant under a set of transformations, such as data augmentations. Based on our theory, we design unsupervised objectives for training neural compressors. Using these objectives, we train a generic image compressor that achieves substantial rate savings (more than 1000x on ImageNet) compared to JPEG on 8 datasets, without decreasing downstream classification performance.

Code Of Conduct: I certify that all co-authors of this work have read and commit to adhering to the NeurIPS Statement on Ethics, Fairness, Inclusivity, and Code of Conduct.

Code: https://github.com/YannDubs/lossyless

Community Implementations: [ 1 code implementation](https://www.catalyzex.com/paper/arxiv:2106.10800/code)