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

Slimmable networks are a family of neural networks that can instantly adjustthe runtime width. The width can be chosen from a predefined widths set toadaptively optimize accuracy-efficiency trade-offs at runtime. In this work, wepropose a systematic approach to train universally slimmable networks(US-Nets), extending slimmable networks to execute at arbitrary width, andgeneralizing to networks both with and without batch normalization layers. Wefurther propose two improved training techniques for US-Nets, named thesandwich rule and inplace distillation, to enhance training process and boosttesting accuracy. We show improved performance of universally slimmableMobileNet v1 and MobileNet v2 on ImageNet classification task, compared withindividually trained ones and 4-switch slimmable network baselines. We alsoevaluate the proposed US-Nets and improved training techniques on tasks ofimage super-resolution and deep reinforcement learning. Extensive ablationexperiments on these representative tasks demonstrate the effectiveness of ourproposed methods. Our discovery opens up the possibility to directly evaluateFLOPs-Accuracy spectrum of network architectures. Code and models are availableat: https://github.com/JiahuiYu/slimmable_networks

Universally Slimmable Networks and Improved Training Techniques