Lossy Compression for Lossless Prediction
Yann Dubois, Benjamin Bloem-Reddy, Karen Ullrich, Chris J. Maddison
Code Available — Be the first to reproduce this paper.
ReproduceCode
- github.com/YannDubs/lossylessOfficialIn paperpytorch★ 121
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 1000 on ImageNet) compared to JPEG on 8 datasets, without decreasing downstream classification performance.
Tasks
Benchmark Results
| Dataset | Model | Metric | Claimed | Verified | Status |
|---|---|---|---|---|---|
| Caltech101 | Lossyless Compressor | Bit rate | 1,340 | — | Unverified |
| Cars-196 | Lossyless Compressor | Bit rate | 1,470 | — | Unverified |
| CIFAR-10 | Lossyless Compressor | Bit rate | 1,410 | — | Unverified |
| Food-101 | Lossyless Compressor | Bit rate | 1,270 | — | Unverified |
| Oxford-IIIT Pet Dataset | Lossyless Compressor | Bit rate | 1,210 | — | Unverified |
| PCam | Lossyless Compressor | Bit rate | 1,490 | — | Unverified |
| STL-10 | Lossyless Compressor | Bit rate | 1,340 | — | Unverified |