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

The ability to transfer skills across tasks has the potential to scale upreinforcement learning (RL) agents to environments currently out of reach.Recently, a framework based on two ideas, successor features (SFs) andgeneralised policy improvement (GPI), has been introduced as a principled wayof transferring skills. In this paper we extend the SFs & GPI framework in twoways. One of the basic assumptions underlying the original formulation of SFs &GPI is that rewards for all tasks of interest can be computed as linearcombinations of a fixed set of features. We relax this constraint and show thatthe theoretical guarantees supporting the framework can be extended to any setof tasks that only differ in the reward function. Our second contribution is toshow that one can use the reward functions themselves as features for futuretasks, without any loss of expressiveness, thus removing the need to specify aset of features beforehand. This makes it possible to combine SFs & GPI withdeep learning in a more stable way. We empirically verify this claim on acomplex 3D environment where observations are images from a first-personperspective. We show that the transfer promoted by SFs & GPI leads to very goodpolicies on unseen tasks almost instantaneously. We also describe how to learnpolicies specialised to the new tasks in a way that allows them to be added tothe agent's set of skills, and thus be reused in the future.

Transfer in Deep Reinforcement Learning Using Successor Features and Generalised Policy Improvement