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

State of the art deep learning models have made steady progress in the fieldsof computer vision and natural language processing, at the expense of growingmodel sizes and computational complexity. Deploying these models on low powerand mobile devices poses a challenge due to their limited compute capabilitiesand strict energy budgets. One solution that has generated significant researchinterest is deploying highly quantized models that operate on low precisioninputs and weights less than eight bits, trading off accuracy for performance.These models have a significantly reduced memory footprint (up to 32xreduction) and can replace multiply-accumulates with bitwise operations duringcompute intensive convolution and fully connected layers.  Most deep learning frameworks rely on highly engineered linear algebralibraries such as ATLAS or Intel's MKL to implement efficient deep learningoperators. To date, none of the popular deep learning directly support lowprecision operators, partly due to a lack of optimized low precision libraries.In this paper we introduce a work flow to quickly generate high performance lowprecision deep learning operators for arbitrary precision that target multipleCPU architectures and include optimizations such as memory tiling andvectorization. We present an extensive case study on low power ARM Cortex-A53CPU, and show how we can generate 1-bit, 2-bit convolutions with speedups up to16x over an optimized 16-bit integer baseline and 2.3x better than handwrittenimplementations.

Automating Generation of Low Precision Deep Learning Operators