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

Searching for network width is an effective way to slim deep neural networkswith hardware budgets. With this aim, a one-shot supernet is usually leveragedas a performance evaluator to rank the performance \wrt~different width.Nevertheless, current methods mainly follow a manually fixed weight sharingpattern, which is limited to distinguish the performance gap of differentwidth. In this paper, to better evaluate each width, we propose a locally freeweight sharing strategy (CafeNet) accordingly. In CafeNet, weights are morefreely shared, and each width is jointly indicated by its base channels andfree channels, where free channels are supposed to loCAte FrEely in a localzone to better represent each width. Besides, we propose to further reduce thesearch space by leveraging our introduced FLOPs-sensitive bins. As a result,our CafeNet can be trained stochastically and get optimized within a min-minstrategy. Extensive experiments on ImageNet, CIFAR-10, CelebA and MS COCOdataset have verified our superiority comparing to other state-of-the-artbaselines. For example, our method can further boost the benchmark NAS networkEfficientNet-B0 by 0.41\% via searching its width more delicately.

Locally Free Weight Sharing for Network Width Search