Base-2 Softmax Function: Suitability for Training and Efficient Hardware Implementation-Reference-Cited by-同舟云学术

Base-2 Softmax Function: Suitability for Training and Efficient Hardware Implementation

Author:

Zhang Yuan¹^ORCID,Zhang Yonggang²^ORCID,Peng Lele¹^ORCID,Quan Lianghua²,Zheng Shubin¹^ORCID,Lu Zhonghai³^ORCID,Chen Hui²^ORCID

Affiliation:

1. School of Urban Railway Transportation, Shanghai University of Engineering Science, Shanghai, China

2. School of Electronic Science and Engineering, Nanjing University, Nanjing, China

3. Department of Electrical Engineering, School of Electrical Engineering and Computer Science, KTH Royal Institute of Technology, Stockholm, Sweden

Funder

National Natural Science Foundation of China

Shanghai Science and Technology Program

Publisher

Institute of Electrical and Electronics Engineers (IEEE)

Subject

Electrical and Electronic Engineering,Hardware and Architecture

Link

Reference35 articles.

Cited by 17 articles. 订阅此论文施引文献订阅此论文施引文献，注册后可以免费订阅5篇论文的施引文献，订阅后可以查看论文全部施引文献

1. Hyft: A Reconfigurable Softmax Accelerator with Hybrid Numeric Format for both Training and Inference;Proceedings of the 29th ACM/IEEE International Symposium on Low Power Electronics and Design;2024-08-05

3. Research on Information Extraction of LCSTS Dataset Based on an Improved BERTSum-LSTM Model;2024 2nd International Conference on Mechatronics, IoT and Industrial Informatics (ICMIII);2024-06-12

5. Reusing Softmax Hardware Unit for GELU Computation in Transformers;2024 IEEE 6th International Conference on AI Circuits and Systems (AICAS);2024-04-22