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

Many recent works have designed accelerators for Convolutional NeuralNetworks (CNNs). While digital accelerators have relied on near dataprocessing, analog accelerators have further reduced data movement byperforming in-situ computation. Recent works take advantage of highly parallelanalog in-situ computation in memristor crossbars to accelerate the manyvector-matrix multiplication operations in CNNs. However, these in-situaccelerators have two significant short-comings that we address in this work.First, the ADCs account for a large fraction of chip power and area. Second,these accelerators adopt a homogeneous design where every resource isprovisioned for the worst case. By addressing both problems, the newarchitecture, Newton, moves closer to achieving optimal energy-per-neuron forcrossbar accelerators.  We introduce multiple new techniques that apply at different levels of thetile hierarchy. Two of the techniques leverage heterogeneity: one adapts ADCprecision based on the requirements of every sub-computation (with zero impacton accuracy), and the other designs tiles customized for convolutions orclassifiers. Two other techniques rely on divide-and-conquer numeric algorithmsto reduce computations and ADC pressure. Finally, we place constraints on how aworkload is mapped to tiles, thus helping reduce resource provisioning intiles. For a wide range of CNN dataflows and structures, Newton achieves a 77%decrease in power, 51% improvement in energy efficiency, and 2.2x higherthroughput/area, relative to the state-of-the-art ISAAC accelerator.

Newton: Gravitating Towards the Physical Limits of Crossbar Acceleration