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/

Zendy Plus

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

Zendy Tools

Zendy Open

With the flourish of differentiable neural architecture search (NAS),automatically searching latency-constrained architectures gives a newperspective to reduce human labor and expertise. However, the searchedarchitectures are usually suboptimal in accuracy and may have large jittersaround the target latency. In this paper, we rethink three freedoms ofdifferentiable NAS, i.e. operation-level, depth-level and width-level, andpropose a novel method, named Three-Freedom NAS (TF-NAS), to achieve both goodclassification accuracy and precise latency constraint. For theoperation-level, we present a bi-sampling search algorithm to moderate theoperation collapse. For the depth-level, we introduce a sink-connecting searchspace to ensure the mutual exclusion between skip and other candidateoperations, as well as eliminate the architecture redundancy. For thewidth-level, we propose an elasticity-scaling strategy that achieves preciselatency constraint in a progressively fine-grained manner. Experiments onImageNet demonstrate the effectiveness of TF-NAS. Particularly, our searchedTF-NAS-A obtains 76.9% top-1 accuracy, achieving state-of-the-art results withless latency. The total search time is only 1.8 days on 1 Titan RTX GPU. Codeis available at https://github.com/AberHu/TF-NAS.

TF-NAS: Rethinking Three Search Freedoms of Latency-Constrained Differentiable Neural Architecture Search