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

Phenomenally successful in practical inference problems, convolutional neuralnetworks (CNN) are widely deployed in mobile devices, data centers, and evensupercomputers. The number of parameters needed in CNNs, however, are oftenlarge and undesirable. Consequently, various methods have been developed toprune a CNN once it is trained. Nevertheless, the resulting CNNs offer limitedbenefits. While pruning the fully connected layers reduces a CNN's sizeconsiderably, it does not improve inference speed noticeably as the computeheavy parts lie in convolutions. Pruning CNNs in a way that increase inferencespeed often imposes specific sparsity structures, thus limiting the achievablesparsity levels.  We present a method to realize simultaneously size economy and speedimprovement while pruning CNNs. Paramount to our success is an efficientgeneral sparse-with-dense matrix multiplication implementation that isapplicable to convolution of feature maps with kernels of arbitrary sparsitypatterns. Complementing this, we developed a performance model that predictssweet spots of sparsity levels for different layers and on different computerarchitectures. Together, these two allow us to demonstrate 3.1--7.3$\times$convolution speedups over dense convolution in AlexNet, on Intel Atom, Xeon,and Xeon Phi processors, spanning the spectrum from mobile devices tosupercomputers. We also open source our project athttps://github.com/IntelLabs/SkimCaffe.

Faster CNNs with Direct Sparse Convolutions and Guided Pruning