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

Structured network pruning is a practical approach to reduce computation costdirectly while retaining the CNNs' generalization performance in realapplications. However, identifying redundant filters is a core problem instructured network pruning, and current redundancy criteria only focus onindividual filters' attributes. When pruning sparsity increases, theseredundancy criteria are not effective or efficient enough. Since thefilter-wise interaction also contributes to the CNN's prediction accuracy, weintegrate the filter-wise interaction into the redundancy criterion. In ourcriterion, we introduce the filter importance and filter utilization strengthto reflect the decision ability of individual and multiple filters. Utilizingthis new redundancy criterion, we propose a structured network pruning approachSNPFI (Structured Network Pruning by measuring Filter-wise Interaction). Duringthe pruning, the SNPFI can automatically assign the proper sparsity based onthe filter utilization strength and eliminate the useless filters by filterimportance. After the pruning, the SNPFI can recover pruned model's performanceeffectively without iterative training by minimizing the interactiondifference. We empirically demonstrate the effectiveness of the SNPFI withseveral commonly used CNN models, including AlexNet, MobileNetv1, andResNet-50, on various image classification datasets, including MNIST, CIFAR-10,and ImageNet. For all experimental CNN models, nearly 60% of computation isreduced in a network compression while the classification accuracy remains.

Structured Network Pruning by Measuring Filter-wise Interactions