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

In the early stages of human life, babies develop their skills by exploringdifferent scenarios motivated by their inherent satisfaction rather than byextrinsic rewards from the environment. This behavior, referred to as intrinsicmotivation, has emerged as one solution to address the exploration challengederived from reinforcement learning environments with sparse rewards. Diverseexploration approaches have been proposed to accelerate the learning processover single- and multi-agent problems with homogeneous agents. However, scarcestudies have elaborated on collaborative learning frameworks betweenheterogeneous agents deployed into the same environment, but interacting withdifferent instances of the latter without any prior knowledge. Beyond theheterogeneity, each agent's characteristics grant access only to a subset ofthe full state space, which may hide different exploration strategies andoptimal solutions. In this work we combine ideas from intrinsic motivation andtransfer learning. Specifically, we focus on sharing parameters in actor-criticmodel architectures and on combining information obtained through intrinsicmotivation with the aim of having a more efficient exploration and fasterlearning. We test our strategies through experiments performed over a modifiedViZDooM's My Way Home scenario, which is more challenging than its originalversion and allows evaluating the heterogeneity between agents. Our resultsreveal different ways in which a collaborative framework with little additionalcomputational cost can outperform an independent learning process withoutknowledge sharing. Additionally, we depict the need for modulating correctlythe importance between the extrinsic and intrinsic rewards to avoid undesiredagent behaviors.

Collaborative Training of Heterogeneous Reinforcement Learning Agents in Environments with Sparse Rewards: What and When to Share?