Searching for Low-Bit Weights in Quantized Neural Networks
This addresses the problem of efficient AI accelerator development through improved quantization, but it is incremental as it builds on existing quantization techniques.
The paper tackles the optimization difficulty in quantized neural networks due to non-differentiable quantization functions by treating discrete weights as searchable variables using a differential method, resulting in higher performance on image classification and super-resolution benchmarks compared to state-of-the-art methods.
Quantized neural networks with low-bit weights and activations are attractive for developing AI accelerators. However, the quantization functions used in most conventional quantization methods are non-differentiable, which increases the optimization difficulty of quantized networks. Compared with full-precision parameters (i.e., 32-bit floating numbers), low-bit values are selected from a much smaller set. For example, there are only 16 possibilities in 4-bit space. Thus, we present to regard the discrete weights in an arbitrary quantized neural network as searchable variables, and utilize a differential method to search them accurately. In particular, each weight is represented as a probability distribution over the discrete value set. The probabilities are optimized during training and the values with the highest probability are selected to establish the desired quantized network. Experimental results on benchmarks demonstrate that the proposed method is able to produce quantized neural networks with higher performance over the state-of-the-art methods on both image classification and super-resolution tasks.