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 (RL) algorithms are often known for sampleinefficiency and difficult generalization. Recently, Unsupervised EnvironmentDesign (UED) emerged as a new paradigm for zero-shot generalization bysimultaneously learning a task distribution and agent policies on the generatedtasks. This is a non-stationary process where the task distribution evolvesalong with agent policies; creating an instability over time. While past worksdemonstrated the potential of such approaches, sampling effectively from thetask space remains an open challenge, bottlenecking these approaches. To thisend, we introduce CLUTR: a novel unsupervised curriculum learning algorithmthat decouples task representation and curriculum learning into a two-stageoptimization. It first trains a recurrent variational autoencoder on randomlygenerated tasks to learn a latent task manifold. Next, a teacher agent createsa curriculum by maximizing a minimax REGRET-based objective on a set of latenttasks sampled from this manifold. Using the fixed-pretrained task manifold, weshow that CLUTR successfully overcomes the non-stationarity problem andimproves stability. Our experimental results show CLUTR outperforms PAIRED, aprincipled and popular UED method, in the challenging CarRacing and navigationenvironments: achieving 10.6X and 45\% improvement in zero-shot generalization,respectively. CLUTR also performs comparably to the non-UED state-of-the-artfor CarRacing, while requiring 500X fewer environment interactions.

CLUTR: Curriculum Learning via Unsupervised Task Representation Learning