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

Convolutional Neural Networks (CNN) are becoming a common presence in manyapplications and services, due to their superior recognition accuracy. They areincreasingly being used on mobile devices, many times just by porting largemodels designed for server space, although several model compression techniqueshave been considered. One model compression technique intended to reducecomputations is channel pruning. Mobile and embedded systems now have GPUswhich are ideal for the parallel computations of neural networks and for theirlower energy cost per operation. Specialized libraries perform these neuralnetwork computations through highly optimized routines. As we find in ourexperiments, these libraries are optimized for the most common network shapes,making uninstructed channel pruning inefficient. We evaluate higher levellibraries, which analyze the input characteristics of a convolutional layer,based on which they produce optimized OpenCL (Arm Compute Library and TVM) andCUDA (cuDNN) code. However, in reality, these characteristics and subsequentchoices intended for optimization can have the opposite effect. We show that areduction in the number of convolutional channels, pruning 12% of the initialsize, is in some cases detrimental to performance, leading to 2x slowdown. Onthe other hand, we also find examples where performance-aware pruning achievesthe intended results, with performance speedups of 3x with cuDNN and above 10xwith Arm Compute Library and TVM. Our findings expose the need forhardware-instructed neural network pruning.

Performance Aware Convolutional Neural Network Channel Pruning for Embedded GPUs