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

Recently several structured pruning techniques have been introduced forenergy-efficient implementation of Deep Neural Networks (DNNs) with lessernumber of crossbars. Although, these techniques have claimed to preserve theaccuracy of the sparse DNNs on crossbars, none have studied the impact of theinexorable crossbar non-idealities on the actual performance of the prunednetworks. To this end, we perform a comprehensive study to show how highlysparse DNNs, that result in significant crossbar-compression-rate, can lead tosevere accuracy losses compared to unpruned DNNs mapped onto non-idealcrossbars. We perform experiments with multiple structured-pruning approaches(such as, C/F pruning, XCS and XRS) on VGG11 and VGG16 DNNs with benchmarkdatasets (CIFAR10 and CIFAR100). We propose two mitigation approaches -Crossbar column rearrangement and Weight-Constrained-Training (WCT) - that canbe integrated with the crossbar-mapping of the sparse DNNs to minimize accuracylosses incurred by the pruned models. These help in mitigating non-idealitiesby increasing the proportion of low conductance synapses on crossbars, therebyimproving their computational accuracies.

Examining and Mitigating the Impact of Crossbar Non-idealities for Accurate Implementation of Sparse Deep Neural Networks