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 achieve very efficient deep learning model deployment that designs neuralnetwork architectures to fit different hardware constraints. Given aconstraint, most neural architecture search (NAS) methods either sample a setof sub-networks according to a pre-trained accuracy predictor, or adopt theevolutionary algorithm to evolve specialized networks from the supernet. Bothapproaches are time consuming. Here our key idea for very efficient deploymentis, when searching the architecture space, constructing a table that stores thevalidation accuracy of all candidate blocks at all layers. For a stricterhardware constraint, the architecture of a specialized network can be veryefficiently determined based on this table by picking the best candidate blocksthat yield the least accuracy loss. To accomplish this idea, we proposeProgressive One-shot Neural Architecture Search (PONAS) that combinesadvantages of progressive NAS and one-shot methods. In PONAS, we propose atwo-stage training scheme, including the meta training stage and thefine-tuning stage, to make the search process efficient and stable. Duringsearch, we evaluate candidate blocks in different layers and construct theaccuracy table that is to be used in deployment. Comprehensive experimentsverify that PONAS is extremely flexible, and is able to find architecture of aspecialized network in around 10 seconds. In ImageNet classification, 75.2%top-1 accuracy can be obtained, which is comparable with the state of the arts.

PONAS: Progressive One-shot Neural Architecture Search for Very Efficient Deployment