ACIQ: Analytical Clipping for Integer Quantization of neural networksDownload PDF

Sep 27, 2018 (edited Dec 21, 2018)ICLR 2019 Conference Blind SubmissionReaders: Everyone
  • Abstract: We analyze the trade-off between quantization noise and clipping distortion in low precision networks. We identify the statistics of various tensors, and derive exact expressions for the mean-square-error degradation due to clipping. By optimizing these expressions, we show marked improvements over standard quantization schemes that normally avoid clipping. For example, just by choosing the accurate clipping values, more than 40\% accuracy improvement is obtained for the quantization of VGG-16 to 4-bits of precision. Our results have many applications for the quantization of neural networks at both training and inference time.
  • Keywords: quantization, reduced precision, training, inference, activation
  • TL;DR: We analyze the trade-off between quantization noise and clipping distortion in low precision networks, and show marked improvements over standard quantization schemes that normally avoid clipping
12 Replies

Loading