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

AlphaGo's astonishing performance has ignited an explosive interest indeveloping deep reinforcement learning (DRL) for numerous real-worldapplications, such as intelligent robotics. However, the often prohibitivecomplexity of DRL stands at the odds with the required real-time control andconstrained resources in many DRL applications, limiting the great potential ofDRL powered intelligent devices. While substantial efforts have been devoted tocompressing other deep learning models, existing works barely touch the surfaceof compressing DRL. In this work, we first identify that there exists anoptimal model size of DRL that can maximize both the test scores andefficiency, motivating the need for task-specific DRL agents. We thereforepropose an Auto-Agent-Distiller (A2D) framework, which to our best knowledge isthe first neural architecture search (NAS) applied to DRL to automaticallysearch for the optimal DRL agents for various tasks that optimize both the testscores and efficiency. Specifically, we demonstrate that vanilla NAS can easilyfail in searching for the optimal agents, due to its resulting high variance inDRL training stability, and then develop a novel distillation mechanism todistill the knowledge from both the teacher agent's actor and critic tostabilize the searching process and improve the searched agents' optimality.Extensive experiments and ablation studies consistently validate our findingsand the advantages and general applicability of our A2D, outperforming manuallydesigned DRL in both the test scores and efficiency. All the codes will bereleased upon acceptance.

Auto-Agent-Distiller: Towards Efficient Deep Reinforcement Learning Agents via Neural Architecture Search