# CAT: Compression-Aware Training for bandwidth reduction

arXiv: https://arxiv.org/abs/1909.11481 · HTML: https://mlsanity.com/paper/1909.11481/

## Key facts

- Cited by 8 papers in the map (data through 14 Oct 2026).
- First posted on arXiv 25 Sep 2019.

## Built on by (most cited first, 8 of 8)

- [KV Cache Transform Coding for Compact Storage in LLM Inference](https://mlsanity.com/paper/2511.01815/) (2025)
- [A Generalized Zero-Shot Quantization of Deep Convolutional Neural Networks via Learned Weights Statistics](https://mlsanity.com/paper/2112.02834/) (2021)
- [Bimodal Distributed Binarized Neural Networks](https://mlsanity.com/paper/2204.02004/) (2022)
- [Lossy and Lossless (L2) Post-training Model Size Compression](https://mlsanity.com/paper/2308.04269/) (2023)
- [AMED: Automatic Mixed-Precision Quantization for Edge Devices](https://mlsanity.com/paper/2205.15437/) (2022)
- [Rotation Invariant Quantization for Model Compression](https://mlsanity.com/paper/2303.03106/) (2023)
- [Reducing Storage of Pretrained Neural Networks by Rate-Constrained Quantization and Entropy Coding](https://mlsanity.com/paper/2505.18758/) (2025)
- [Big2Small: A Unifying Neural Network Framework for Model Compression](https://mlsanity.com/paper/2603.29768/) (2026)

## Abstract (arXiv)

Convolutional neural networks (CNNs) have become the dominant neural network architecture for solving visual processing tasks. One of the major obstacles hindering the ubiquitous use of CNNs for inference is their relatively high memory bandwidth requirements, which can be a main energy consumer and throughput bottleneck in hardware accelerators. Accordingly, an efficient feature map compression method can result in substantial performance gains. Inspired by quantization-aware training approaches, we propose a compression-aware training (CAT) method that involves training the model in a way that allows better compression of feature maps during inference. Our method trains the model to achieve low-entropy feature maps, which enables efficient compression at inference time using classical transform coding methods. CAT significantly improves the state-of-the-art results reported for quantization. For example, on ResNet-34 we achieve 73.1% accuracy (0.2% degradation from the baseline) with an average representation of only 1.79 bits per value. Reference implementation accompanies the paper at https://github.com/CAT-teams/CAT