English (United Kingdom)

https://curated-unify.zendy.io/wp-json/zendy-region/v1/featured_content/oa?rat=en

https://curated-unify.zendy.io/wp-json/zendy-region/v1/highlighted_journal/

Global: Zendy Plus annual

Presents the access of premium content as premium feature

Premium Content

Presents the keyphrase highlighting as premium feature

Keyphrase Highlighting

Presents the summarisation as premium feature

Summarisation

Insights

Presents the pdf analysis as premium feature

PDF Analysis

Presents the zaia usage as premium feature

ZAIA

Global: Zendy Plus monthly

Global: Zendy Tools annual

Global: Zendy Tools monthly

Global: Zendy Free Plan

Deep convolutional neural networks (CNNs) with a large number of parametersrequire intensive computational resources, and thus are hard to be deployed inresource-constrained platforms. Decomposition-based methods, therefore, havebeen utilized to compress CNNs in recent years. However, since the compressionfactor and performance are negatively correlated, the state-of-the-art workseither suffer from severe performance degradation or have relatively lowcompression factors. To overcome this problem, we propose to compress CNNs andalleviate performance degradation via joint matrix decomposition, which isdifferent from existing works that compressed layers separately. The idea isinspired by the fact that there are lots of repeated modules in CNNs. Byprojecting weights with the same structures into the same subspace, networkscan be jointly compressed with larger ranks. In particular, three joint matrixdecomposition schemes are developed, and the corresponding optimizationapproaches based on Singular Value Decomposition are proposed. Extensiveexperiments are conducted across three challenging compact CNNs for differentbenchmark data sets to demonstrate the superior performance of our proposedalgorithms. As a result, our methods can compress the size of ResNet-34 by 22Xwith slighter accuracy degradation compared with several state-of-the-artmethods.

Joint Matrix Decomposition for Deep Convolutional Neural Networks Compression