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

In offline imitation learning (IL), we generally assume only a handful ofexpert trajectories and a supplementary offline dataset from suboptimalbehaviors to learn the expert policy. While it is now common to minimize thedivergence between state-action visitation distributions so that the agent alsoconsiders the future consequences of an action, a sampling error in an offlinedataset may lead to erroneous estimates of state-action visitations in theoffline case. In this paper, we investigate the effect of controlling theeffective planning horizon (i.e., reducing the discount factor) as opposed toimposing an explicit regularizer, as previously studied. Unfortunately, itturns out that the existing algorithms suffer from magnified approximationerrors when the effective planning horizon is shortened, which results in asignificant degradation in performance. We analyze the main cause of theproblem and provide the right remedies to correct the algorithm. We show thatthe corrected algorithm improves on popular imitation learning benchmarks bycontrolling the effective planning horizon rather than an explicitregularization.

Offline Imitation Learning by Controlling the Effective Planning Horizon