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

Deep reinforcement learning (DRL) is increasingly applied in large-scaleproductions like Netflix and Facebook. As with most data-driven systems, DRLsystems can exhibit undesirable behaviors due to environmental drifts, whichoften occur in constantly-changing production settings. Continual Learning (CL)is the inherent self-healing approach for adapting the DRL agent in response tothe environment's conditions shifts. However, successive shifts of considerablemagnitude may cause the production environment to drift from its originalstate. Recent studies have shown that these environmental drifts tend to driveCL into long, or even unsuccessful, healing cycles, which arise frominefficiencies such as catastrophic forgetting, warm-starting failure, and slowconvergence. In this paper, we propose Dr. DRL, an effective self-healingapproach for DRL systems that integrates a novel mechanism of intentionalforgetting into vanilla CL to overcome its main issues. Dr. DRL deliberatelyerases the DRL system's minor behaviors to systematically prioritize theadaptation of the key problem-solving skills. Using well-established DRLalgorithms, Dr. DRL is compared with vanilla CL on various driftedenvironments. Dr. DRL is able to reduce, on average, the healing time andfine-tuning episodes by, respectively, 18.74% and 17.72%. Dr. DRL successfullyhelps agents to adapt to 19.63% of drifted environments left unsolved byvanilla CL while maintaining and even enhancing by up to 45% the obtainedrewards for drifted environments that are resolved by both approaches.

An Intentional Forgetting-Driven Self-Healing Method For Deep Reinforcement Learning Systems