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 Plus monthly

Global: Zendy Tools annual

Global: Zendy Tools monthly

Global: Zendy Free Plan

Learning to control an agent from data collected offline in a richpixel-based visual observation space is vital for real-world applications ofreinforcement learning (RL). A major challenge in this setting is the presenceof input information that is hard to model and irrelevant to controlling theagent. This problem has been approached by the theoretical RL community throughthe lens of exogenous information, i.e, any control-irrelevant informationcontained in observations. For example, a robot navigating in busy streetsneeds to ignore irrelevant information, such as other people walking in thebackground, textures of objects, or birds in the sky. In this paper, we focuson the setting with visually detailed exogenous information, and introduce newoffline RL benchmarks offering the ability to study this problem. We find thatcontemporary representation learning techniques can fail on datasets where thenoise is a complex and time dependent process, which is prevalent in practicalapplications. To address these, we propose to use multi-step inverse models,which have seen a great deal of interest in the RL theory community, to learnAgent-Controller Representations for Offline-RL (ACRO). Despite being simpleand requiring no reward, we show theoretically and empirically that therepresentation created by this objective greatly outperforms baselines.

Agent-Controller Representations: Principled Offline RL with Rich Exogenous Information