ReActXGB: A Hybrid Binary Convolutional Neural Network Architecture for Improved Performance and Computational Efficiency

Po Hsun Chu, Ching Han Chen

Research output: Chapter in Book/Report/Conference proceedingConference contributionpeer-review

Abstract

Binary convolutional neural networks (BCNNs) provide a potential solution to reduce the memory requirements and computational costs associated with deep neural networks (DNNs). However, achieving a trade-off between performance and computational resources remains a significant challenge. Furthermore, the fully connected layer of BCNNs has evolved into a significant computational bottleneck. This is mainly due to the conventional practice of excluding the input layer and fully connected layer from binarization to prevent a substantial loss in accuracy. In this paper, we propose a hybrid model named ReActXGB, where we replace the fully convolutional layer of ReActNet-A with XGBoost. This modification targets to narrow the performance gap between BCNNs and real-valued networks while maintaining lower computational costs. Experimental results on the FashionMNIST benchmark demonstrate that ReActXGB outperforms ReActNet-A by 1.47% in top-1 accuracy, along with a reduction of 7.14% in floating-point operations (FLOPs) and 1.02% in model size.

Original languageEnglish
Title of host publication11th IEEE International Conference on Consumer Electronics - Taiwan, ICCE-Taiwan 2024
PublisherInstitute of Electrical and Electronics Engineers Inc.
Pages327-328
Number of pages2
ISBN (Electronic)9798350386844
DOIs
StatePublished - 2024
Event11th IEEE International Conference on Consumer Electronics - Taiwan, ICCE-Taiwan 2024 - Taichung, Taiwan
Duration: 9 Jul 202411 Jul 2024

Publication series

Name11th IEEE International Conference on Consumer Electronics - Taiwan, ICCE-Taiwan 2024

Conference

Conference11th IEEE International Conference on Consumer Electronics - Taiwan, ICCE-Taiwan 2024
Country/TerritoryTaiwan
CityTaichung
Period9/07/2411/07/24

Keywords

  • binary convolution neural network
  • fully connected layer
  • hybrid model
  • model compression
  • XGBoost

Fingerprint

Dive into the research topics of 'ReActXGB: A Hybrid Binary Convolutional Neural Network Architecture for Improved Performance and Computational Efficiency'. Together they form a unique fingerprint.

Cite this