arXiv 2307.08982

Neural Network Pruning as Spectrum Preserving Process

By Shibo Yao, Dantong Yu, et al.

Published 2023-07-18

Citation lineage

Review the prior work and downstream research connected to this paper.

Neural networks have achieved remarkable performance in various application domains. Nevertheless, a large number of weights in pre-trained deep neural networks prohibit them from being deployed on smartphones and embedded systems. It is highly desirable to obtain lightweight versions of neural networks for inference in edge devices. Many cost-effective approaches were proposed to prune dense and convolutional layer…

View the original paper on arXiv