Training Binarized Neural Networks Using Ternary Multipliers

Amir Ardakani,Arash Ardakani,Warren J. Gross

IEEE Design & Test（2021）

引用 4|浏览4

暂无评分

摘要

Editor’s note: This article considers the under-investigated problem of training neural networks based on stochastic computing. A new dynamic sign magnitude representation for symbols in ternary format {-1, 0, 1} facilitates learning while retaining SC’s benefits. —John Hayes, University of Michigan

查看译文

关键词

Artificial neural networks,Training,Logic gates,Neural networks,Stochastic processes,Adders,Quantization (signal)

AI 理解论文

溯源树

样例

生成溯源树，研究论文发展脉络

Chat Paper

正在生成论文摘要