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

Filter pruning has been widely used in the compression and acceleration ofconvolutional neural networks (CNNs). However, most existing methods are stillchallenged by heavy compute cost and biased filter selection. Moreover, mostdesigns for filter evaluation miss interpretability due to the lack ofappropriate theoretical guidance. In this paper, we propose a novel filterpruning method which evaluates filters in a interpretable, multi-persepectiveand data-free manner. We introduce information capacity, a metric thatrepresents the amount of information contained in a filter. Based on theinterpretability and validity of information entropy, we propose to use that asa quantitative index of information quantity. Besides, we experimently showthat the obvious correlation between the entropy of the feature map and thecorresponding filter, so as to propose an interpretable, data-driven scheme tomeasure the information capacity of the filter. Further, we introduceinformation independence, another metric that represents the correlation amongdifferrent filters. Consequently, the least impotant filters, which have lessinformation capacity and less information independence, will be pruned. Weevaluate our method on two benchmarks using multiple representative CNNarchitectures, including VGG-16 and ResNet. On CIFAR-10, we reduce 71.9% offloating-point operations (FLOPs) and 69.4% of parameters for ResNet-110 with0.28% accuracy increase. On ILSVRC-2012, we reduce 76.6% of floating-pointoperations (FLOPs) and 68.6% of parameters for ResNet-50 with only 2.80%accuracy decrease, which outperforms the state-of-the-arts.

Filter Pruning based on Information Capacity and Independence