Dynamic binary neural network by learning channel-wise thresholds

Zhang, Jiehua; Su, Zhuo; Feng, Yanghe; Lu, Xin; Pietikäinen, Matti; Liu, Li (2022-04-27)
 
Avaa tiedosto
Lataukset: 

URL:
https://doi.org/10.1109/icassp43922.2022.9747328

Zhang, Jiehua
Su, Zhuo
Feng, Yanghe
Lu, Xin
Pietikäinen, Matti
Liu, Li
Institute of Electrical and Electronics Engineers
27.04.2022

J. Zhang, Z. Su, Y. Feng, X. Lu, M. Pietikäinen and L. Liu, "Dynamic Binary Neural Network by Learning Channel-Wise Thresholds," ICASSP 2022 - 2022 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP), Singapore, Singapore, 2022, pp. 1885-1889, doi: 10.1109/ICASSP43922.2022.9747328.

https://rightsstatements.org/vocab/InC/1.0/
© 2022 IEEE. Personal use of this material is permitted. Permission from IEEE must be obtained for all other uses, in any current or future media, including reprinting/republishing this material for advertising or promotional purposes, creating new collective works, for resale or redistribution to servers or lists, or reuse of any copyrighted component of this work in other works.
https://rightsstatements.org/vocab/InC/1.0/
doi:https://doi.org/10.1109/icassp43922.2022.9747328
Näytä kaikki kuvailutiedot
Julkaisun pysyvä osoite on
https://urn.fi/URN:NBN:fi-fe2023040334614
Tiivistelmä

Abstract

Binary neural networks (BNNs) constrain weights and activations to +1 or -1 with limited storage and computational cost, which is hardware-friendly for portable devices. Recently, BNNs have achieved remarkable progress and been adopted into various fields. However, the performance of BNNs is sensitive to activation distribution. The existing BNNs utilized the Sign function with predefined or learned static thresholds to binarize activations. This process limits representation capacity of BNNs since different samples may adapt to unequal thresholds. To address this problem, we propose a dynamic BNN (DyBNN) incorporating dynamic learnable channel-wise thresholds of Sign function and shift parameters of PReLU. The method aggregates the global information into the hyper function and effectively increases the feature expression ability. The experimental results prove that our method is an effective and straightforward way to reduce information loss and enhance performance of BNNs. The DyBNN based on two backbones of ReActNet (MobileNetV1 and ResNet18) achieve 71.2% and 67.4% top1-accuracy on ImageNet dataset, outperforming baselines by a large margin (i.e., 1.8% and 1.5% respectively).

Kokoelmat