Analysis and Solution of CNN Accuracy Reduction over Channel Loop Tiling
- Title
- Analysis and Solution of CNN Accuracy Reduction over Channel Loop Tiling
- Authors
- Kang, Seokhyeong; Kang, Yesung; Park, Yoonho; Kim, Sunghoon; Kwon, Eunji; Lim, Taeho; Oh, Sangyun; Woo, Mingyu
- Date Issued
- 2020-03
- Publisher
- Institute of Electrical and Electronics Engineers Inc.
- Abstract
- Owing to the growth of the size of convolutional neural networks (CNNs), quantization and loop tiling (also called loop breaking) are mandatory to implement CNN on an embedded system. However, channel loop tiling of quantized CNNs induces unexpected errors. We explain why channel loop tiling of quantized CNNs induces the unexpected errors, and how the errors affect the accuracy of state-of-the-art CNNs. We also propose a method to recover accuracy under channel tiling by compressing and decompressing the most-significant bits of partial sums. Using the proposed method, we can recover accuracy by 12.3% with only 1% circuit area overhead and an additional 2% of power consumption.
- URI
- https://oasis.postech.ac.kr/handle/2014.oak/106292
- Article Type
- Conference
- Citation
- 2020 Design, Automation and Test in Europe Conference and Exhibition, DATE 2020, page. 1091 - 1096, 2020-03
- Files in This Item:
- There are no files associated with this item.
Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.