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/

Zendy Plus

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

Zendy Tools

Zendy Open

Non-stationary environments are challenging for reinforcement learningalgorithms. If the state transition and/or reward functions change based onlatent factors, the agent is effectively tasked with optimizing a behavior thatmaximizes performance over a possibly infinite random sequence of MarkovDecision Processes (MDPs), each of which drawn from some unknown distribution.We call each such MDP a context. Most related works make strong assumptionssuch as knowledge about the distribution over contexts, the existence ofpre-training phases, or a priori knowledge about the number, sequence, orboundaries between contexts. We introduce an algorithm that efficiently learnspolicies in non-stationary environments. It analyzes a possibly infinite streamof data and computes, in real-time, high-confidence change-point detectionstatistics that reflect whether novel, specialized policies need to be createdand deployed to tackle novel contexts, or whether previously-optimized onesmight be reused. We show that (i) this algorithm minimizes the delay untilunforeseen changes to a context are detected, thereby allowing for rapidresponses; and (ii) it bounds the rate of false alarm, which is important inorder to minimize regret. Our method constructs a mixture model composed of a(possibly infinite) ensemble of probabilistic dynamics predictors that modelthe different modes of the distribution over underlying latent MDPs. Weevaluate our algorithm on high-dimensional continuous reinforcement learningproblems and show that it outperforms state-of-the-art (model-free andmodel-based) RL algorithms, as well as state-of-the-art meta-learning methodsspecially designed to deal with non-stationarity.

Minimum-Delay Adaptation in Non-Stationary Reinforcement Learning via Online High-Confidence Change-Point Detection