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

Most existing deep neural networks are static, which means they can only doinference at a fixed complexity. But the resource budget can vary substantiallyacross different devices. Even on a single device, the affordable budget canchange with different scenarios, and repeatedly training networks for eachrequired budget would be incredibly expensive. Therefore, in this work, wepropose a general method called MutualNet to train a single network that canrun at a diverse set of resource constraints. Our method trains a cohort ofmodel configurations with various network widths and input resolutions. Thismutual learning scheme not only allows the model to run at differentwidth-resolution configurations but also transfers the unique knowledge amongthese configurations, helping the model to learn stronger representationsoverall. MutualNet is a general training methodology that can be applied tovarious network structures (e.g., 2D networks: MobileNets, ResNet, 3D networks:SlowFast, X3D) and various tasks (e.g., image classification, object detection,segmentation, and action recognition), and is demonstrated to achieveconsistent improvements on a variety of datasets. Since we only train the modelonce, it also greatly reduces the training cost compared to independentlytraining several models. Surprisingly, MutualNet can also be used tosignificantly boost the performance of a single network, if dynamic resourceconstraint is not a concern. In summary, MutualNet is a unified method for bothstatic and adaptive, 2D and 3D networks. Codes and pre-trained models areavailable at \url{https://github.com/taoyang1122/MutualNet}.

MutualNet: Adaptive ConvNet via Mutual Learning from Different Model Configurations