BitPruning: Learning Bitlengths for Aggressive and Accurate Quantization
This addresses the need for efficient neural network deployment on hardware by enabling aggressive quantization with minimal accuracy loss, though it is incremental as it builds on existing low-bit quantization methods.
The paper tackles the problem of finding the minimum bitlength for neural network quantization to maintain accuracy, introducing a training method that achieves aggressive quantization with per-layer bitlengths as low as 3.76 bits on ResNet18 while staying within 0.5% of the base accuracy on ImageNet.
Neural networks have demonstrably achieved state-of-the art accuracy using low-bitlength integer quantization, yielding both execution time and energy benefits on existing hardware designs that support short bitlengths. However, the question of finding the minimum bitlength for a desired accuracy remains open. We introduce a training method for minimizing inference bitlength at any granularity while maintaining accuracy. Namely, we propose a regularizer that penalizes large bitlength representations throughout the architecture and show how it can be modified to minimize other quantifiable criteria, such as number of operations or memory footprint. We demonstrate that our method learns thrifty representations while maintaining accuracy. With ImageNet, the method produces an average per layer bitlength of 4.13, 3.76 and 4.36 bits on AlexNet, ResNet18 and MobileNet V2 respectively, remaining within 2.0%, 0.5% and 0.5% of the base TOP-1 accuracy.