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

Non-volatile memory (NVM) crossbars have been identified as a promisingtechnology, for accelerating important machine learning operations, withmatrix-vector multiplication being a key example. Binary neural networks (BNNs)are especially well-suited for use with NVM crossbars due to their use of alow-bitwidth representation for both activations and weights. However, theaggressive quantization of BNNs can result in suboptimal accuracy, and theanalog effects of NVM crossbars can further degrade the accuracy duringinference.  This paper presents a comprehensive study that benchmarks BNNs trained andvalidated on ImageNet and deployed on NeuroSim, a simulator forNVM-crossbar-based PIM architecture. Our study analyzes the impact of variousparameters, such as input precision and ADC resolution, on both the accuracy ofthe inference and the hardware performance metrics. We have found that an ADCresolution of 8-bit with an input precision of 4-bit achieves near-optimalaccuracy compared to the original BNNs. In addition, we have identifiedbottleneck components in the PIM architecture that affect area, latency, andenergy consumption, and we demonstrate the impact that different BNN layershave on hardware performance.

Comprehensive Benchmarking of Binary Neural Networks on NVM Crossbar Architectures