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

We show that, during inference with Convolutional Neural Networks (CNNs),more than 2x to $8x ineffectual work can be exposed if instead of targetingthose weights and activations that are zero, we target different combinationsof value stream properties. We demonstrate a practical application withBit-Tactical (TCL), a hardware accelerator which exploits weight sparsity, perlayer precision variability and dynamic fine-grain precision reduction foractivations, and optionally the naturally occurring sparse effectual bitcontent of activations to improve performance and energy efficiency. TCLbenefits both sparse and dense CNNs, natively supports both convolutional andfully-connected layers, and exploits properties of all activations to reducestorage, communication, and computation demands. While TCL does not requirechanges to the CNN to deliver benefits, it does reward any technique that wouldamplify any of the aforementioned weight and activation value properties.Compared to an equivalent data-parallel accelerator for dense CNNs, TCLp, avariant of TCL improves performance by 5.05x and is 2.98x more energy efficientwhile requiring 22% more area.

Bit-Tactical: Exploiting Ineffectual Computations in Convolutional Neural Networks: Which, Why, and How