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

Reinforcement learning algorithms rely on exploration to discover newbehaviors, which is typically achieved by following a stochastic policy. Incontinuous control tasks, policies with a Gaussian distribution have beenwidely adopted. Gaussian exploration however does not result in smoothtrajectories that generally correspond to safe and rewarding behaviors inpractical tasks. In addition, Gaussian policies do not result in an effectiveexploration of an environment and become increasingly inefficient as the actionrate increases. This contributes to a low sample efficiency often observed inlearning continuous control tasks. We introduce a family of stationaryautoregressive (AR) stochastic processes to facilitate exploration incontinuous control domains. We show that proposed processes possess twodesirable features: subsequent process observations are temporally coherentwith continuously adjustable degree of coherence, and the process stationarydistribution is standard normal. We derive an autoregressive policy (ARP) thatimplements such processes maintaining the standard agent-environment interface.We show how ARPs can be easily used with the existing off-the-shelf learningalgorithms. Empirically we demonstrate that using ARPs results in improvedexploration and sample efficiency in both simulated and real world domains,and, furthermore, provides smooth exploration trajectories that enable safeoperation of robotic hardware.

Autoregressive Policies for Continuous Control Deep Reinforcement Learning