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

The high computation and memory storage of large deep neural networks (DNNs)models pose intensive challenges to the conventional Von-Neumann architecture,incurring substantial data movements in the memory hierarchy. The memristorcrossbar array has emerged as a promising solution to mitigate the challengesand enable low-power acceleration of DNNs. Memristor-based weight pruning andweight quantization have been seperately investigated and proven effectivenessin reducing area and power consumption compared to the original DNN model.However, there has been no systematic investigation of memristor-basedneuromorphic computing (NC) systems considering both weight pruning and weightquantization. In this paper, we propose an unified and systematicmemristor-based framework considering both structured weight pruning and weightquantization by incorporating alternating direction method of multipliers(ADMM) into DNNs training. We consider hardware constraints such as crossbarblocks pruning, conductance range, and mismatch between weight value and realdevices, to achieve high accuracy and low power and small area footprint. Ourframework is mainly integrated by three steps, i.e., memristor-based ADMMregularized optimization, masked mapping and retraining. Experimental resultsshow that our proposed framework achieves 29.81X (20.88X) weight compressionratio, with 98.38% (96.96%) and 98.29% (97.47%) power and area reduction onVGG-16 (ResNet-18) network where only have 0.5% (0.76%) accuracy loss, comparedto the original DNN models. We share our models at link http://bit.ly/2Jp5LHJ.

An Ultra-Efficient Memristor-Based DNN Framework with Structured Weight Pruning and Quantization Using ADMM