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 analog nature of computing in Memristive crossbars poses significantissues due to various non-idealities such as: parasitic resistances, non-linearI-V characteristics of the device etc. The non-idealities can have adetrimental impact on the functionality i.e. computational accuracy ofcrossbars. Past works have explored modeling the non-idealities usinganalytical techniques. However, several non-idealities have data dependentbehavior. This can not be captured using analytical (non data-dependent) modelsthereby, limiting their suitability in predicting application accuracy.  To address this, we propose a Generalized Approach to Emulating Non-Idealityin Memristive Crossbars using Neural Networks (GENIEx), which accuratelycaptures the data-dependent nature of non-idealities. We perform extensiveHSPICE simulations of crossbars with different voltage and conductancecombinations. Following that, we train a neural network to learn the transfercharacteristics of the non-ideal crossbar. Next, we build a functionalsimulator which includes key architectural facets such as \textit{tiling}, and\textit{bit-slicing} to analyze the impact of non-idealities on theclassification accuracy of large-scale neural networks. We show that GENIExachieves \textit{low} root mean square errors (RMSE) of $0.25$ and $0.7$ forlow and high voltages, respectively, compared to HSPICE. Additionally, theGENIEx errors are $7\times$ and $12.8\times$ better than an analytical modelwhich can only capture the linear non-idealities. Further, using the functionalsimulator and GENIEx, we demonstrate that an analytical model can overestimatethe degradation in classification accuracy by $\ge 10\%$ on CIFAR-100 and$3.7\%$ on ImageNet datasets compared to GENIEx.

GENIEx: A Generalized Approach to Emulating Non-Ideality in Memristive Xbars using Neural Networks