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 Tools annual

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 Tools monthly

Global: Zendy Plus monthly

Presents the access of premium content as premium feature

Premium Content

Global: Zendy Plus annual

Global: Zendy Free Plan

Realizing today's cloud-level artificial intelligence functionalitiesdirectly on devices distributed at the edge of the internet calls for edgehardware capable of processing multiple modalities of sensory data (e.g. video,audio) at unprecedented energy-efficiency. AI hardware architectures todaycannot meet the demand due to a fundamental "memory wall": data movementbetween separate compute and memory units consumes large energy and incurs longlatency. Resistive random-access memory (RRAM) based compute-in-memory (CIM)architectures promise to bring orders of magnitude energy-efficiencyimprovement by performing computation directly within memory. However,conventional approaches to CIM hardware design limit its functional flexibilitynecessary for processing diverse AI workloads, and must overcome hardwareimperfections that degrade inference accuracy. Such trade-offs betweenefficiency, versatility and accuracy cannot be addressed by isolatedimprovements on any single level of the design. By co-optimizing across allhierarchies of the design from algorithms and architecture to circuits anddevices, we present NeuRRAM - the first multimodal edge AI chip using RRAM CIMto simultaneously deliver a high degree of versatility for diverse modelarchitectures, record energy-efficiency $5\times$ - $8\times$ better than priorart across various computational bit-precisions, and inference accuracycomparable to software models with 4-bit weights on all measured standard AIbenchmarks including accuracy of 99.0% on MNIST and 85.7% on CIFAR-10 imageclassification, 84.7% accuracy on Google speech command recognition, and a 70%reduction in image reconstruction error on a Bayesian image recovery task. Thiswork paves a way towards building highly efficient and reconfigurable edge AIhardware platforms for the more demanding and heterogeneous AI applications ofthe future.

Edge AI without Compromise: Efficient, Versatile and Accurate Neurocomputing in Resistive Random-Access Memory