A novel tensor-based model compression method via tucker and tensor train decompositions

2017 IEEE 26th Conference on Electrical Performance of Electronic Packaging and Systems (EPEPS) Pub Date : 2017-10-01 DOI:10.1109/EPEPS.2017.8329724

Cong Chen, Kim Batselier, N. Wong

引用次数: 0

Abstract

We develop a novel tensor-based Tucker-Tensor-Train-Model-Compression (T3MC) scheme for speeding up nonlinear circuit simulation. Experiment shows that T3MC achieves high efficiency with significantly higher accuracy than state-of-the-art nonlinear model order reduction (MOR) methods.

查看原文本刊更多论文

基于tucker和张量列分解的新型张量模型压缩方法

提出了一种新的基于张量的塔克-张量-训练-模型-压缩(T3MC)方案，以加速非线性电路的仿真。实验表明，T3MC算法效率高，精度显著高于当前的非线性模型降阶(MOR)算法。

本文章由计算机程序翻译，如有差异，请以英文原文为准。

求助全文

约1分钟内获得全文求助全文

来源期刊

2017 IEEE 26th Conference on Electrical Performance of Electronic Packaging and Systems (EPEPS)

自引率

0.00%

发文量