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

The pioneer deep neural networks (DNNs) have emerged to be deeper or widerfor improving their accuracy in various applications of artificialintelligence. However, DNNs are often too heavy to deploy in practice, and itis often required to control their architectures dynamically given computingresource budget, i.e., anytime prediction. While most existing approaches havefocused on training multiple shallow sub-networks jointly, we study trainingthin sub-networks instead. To this end, we first build many inclusive thinsub-networks (of the same depth) under a minor modification of existingmulti-branch DNNs, and found that they can significantly outperform thestate-of-art dense architecture for anytime prediction. This is remarkable dueto their simplicity and effectiveness, but training many thin sub-networksjointly faces a new challenge on training complexity. To address the issue, wealso propose a novel DNN architecture by forcing a certain sparsity pattern onmulti-branch network parameters, making them train efficiently for the purposeof anytime prediction. In our experiments on the ImageNet dataset, itssub-networks have up to $43.3\%$ smaller sizes (FLOPs) compared to those of thestate-of-art anytime model with respect to the same accuracy. Finally, we alsopropose an alternative task under the proposed architecture using ahierarchical taxonomy, which brings a new angle for anytime prediction.

Anytime Neural Prediction via Slicing Networks Vertically