Additive Powers-of-Two Quantization: An Efficient Non-uniform Discretization for Neural Networks
This work addresses the challenge of reducing computational overhead in neural networks for deployment in resource-constrained environments, representing an incremental improvement over existing quantization methods.
The authors tackled the problem of efficiently quantizing neural networks by proposing Additive Powers-of-Two (APoT) quantization, a non-uniform scheme that matches weight distributions and reduces computational costs, achieving 76.6% top-1 accuracy on ImageNet with a 4-bit ResNet-50 and a 22% reduction in computational cost compared to uniform quantization.
We propose Additive Powers-of-Two~(APoT) quantization, an efficient non-uniform quantization scheme for the bell-shaped and long-tailed distribution of weights and activations in neural networks. By constraining all quantization levels as the sum of Powers-of-Two terms, APoT quantization enjoys high computational efficiency and a good match with the distribution of weights. A simple reparameterization of the clipping function is applied to generate a better-defined gradient for learning the clipping threshold. Moreover, weight normalization is presented to refine the distribution of weights to make the training more stable and consistent. Experimental results show that our proposed method outperforms state-of-the-art methods, and is even competitive with the full-precision models, demonstrating the effectiveness of our proposed APoT quantization. For example, our 4-bit quantized ResNet-50 on ImageNet achieves 76.6% top-1 accuracy without bells and whistles; meanwhile, our model reduces 22% computational cost compared with the uniformly quantized counterpart. The code is available at https://github.com/yhhhli/APoT_Quantization.