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 Tools annual

Global: Zendy Free Plan

Global: Zendy Tools monthly

Global: Zendy Plus monthly

In many personalized recommendation scenarios, the generalization ability ofa target task can be improved via learning with additional auxiliary tasksalongside this target task on a multi-task network. However, this method oftensuffers from a serious optimization imbalance problem. On the one hand, one ormore auxiliary tasks might have a larger influence than the target task andeven dominate the network weights, resulting in worse recommendation accuracyfor the target task. On the other hand, the influence of one or more auxiliarytasks might be too weak to assist the target task. More challenging is thatthis imbalance dynamically changes throughout the training process and variesacross the parts of the same network. We propose a new method: MetaBalance tobalance auxiliary losses via directly manipulating their gradients w.r.t theshared parameters in the multi-task network. Specifically, in each trainingiteration and adaptively for each part of the network, the gradient of anauxiliary loss is carefully reduced or enlarged to have a closer magnitude tothe gradient of the target loss, preventing auxiliary tasks from being sostrong that dominate the target task or too weak to help the target task.Moreover, the proximity between the gradient magnitudes can be flexiblyadjusted to adapt MetaBalance to different scenarios. The experiments show thatour proposed method achieves a significant improvement of 8.34% in terms ofNDCG@10 upon the strongest baseline on two real-world datasets. The code of ourapproach can be found at here: https://github.com/facebookresearch/MetaBalance

MetaBalance: Improving Multi-Task Recommendations via Adapting Gradient Magnitudes of Auxiliary Tasks