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

Neural approximate computing gains enormous energy-efficiency at the cost oftolerable quality-loss. A neural approximator can map the input data to outputwhile a classifier determines whether the input data are safe to approximatewith quality guarantee. However, existing works cannot maximize the invocationof the approximator, resulting in limited speedup and energy saving. Byexploring the mapping space of those target functions, in this paper, weobserve a nonuniform distribution of the approximation error incurred by thesame approximator. We thus propose a novel approximate computing architecturewith a Multiclass-Classifier and Multiple Approximators (MCMA). Theseapproximators have identical network topologies and thus can share the samehardware resource in a neural processing unit(NPU) clip. In the runtime, MCMAcan swap in the invoked approximator by merely shipping the synapse weightsfrom the on-chip memory to the buffers near MAC within a cycle. We also proposeefficient co-training methods for such MCMA architecture. Experimental resultsshow a more substantial invocation of MCMA as well as the gain ofenergy-efficiency.

Invocation-driven Neural Approximate Computing with a Multiclass-Classifier and Multiple Approximators