Learnable Companding Quantization for Accurate Low-Bit Neural Networks

Quantizing deep neural networks is an effective method for reducing memory consumption and improving inference speed, and is thus useful for implementation in resourceconstrained devices. However, it is still hard for extremely low-bit models to achieve accuracy comparable with that of full-precision models. To address this issue, we propose learnable companding quantization (LCQ) as a novel nonuniform quantization method for 2-, 3-, and 4-bit models. LCQ jointly optimizes model weights and learnable companding functions that can flexibly and non-uniformly control the quantization levels of weights and activations. We also present a new weight normalization technique that allows more stable training for quantization. Experimental results show that LCQ outperforms conventional state-ofthe-art methods and narrows the gap between quantized and full-precision models for image classification and object detection tasks. Notably, the 2-bit ResNet-50 model on ImageNet achieves top-1 accuracy of 75.1% and reduces the gap to 1.7%, allowing LCQ to further exploit the potential of non-uniform quantization.