Entropic Desired Dynamics for Intrinsic ControlDownload PDF

21 May 2021, 20:51 (edited 25 Jan 2022)NeurIPS 2021 PosterReaders: Everyone
  • Keywords: intrinsic control, skill discovery, exploration, unsupervised reinforcement learning
  • TL;DR: Discriminating between globally consistent codes yields exploratory behavior and a map of the controllable environment without the requirement of supervisory signals or rewards.
  • Abstract: An agent might be said, informally, to have mastery of its environment when it has maximised the effective number of states it can reliably reach. In practice, this often means maximizing the number of latent codes that can be discriminated from future states under some short time horizon (e.g. \cite{eysenbach2018diversity}). By situating these latent codes in a globally consistent coordinate system, we show that agents can reliably reach more states in the long term while still optimizing a local objective. A simple instantiation of this idea, \textbf{E}ntropic \textbf{D}esired \textbf{D}ynamics for \textbf{I}ntrinsic \textbf{C}on\textbf{T}rol (EDDICT), assumes fixed additive latent dynamics, which results in tractable learning and an interpretable latent space. Compared to prior methods, EDDICT's globally consistent codes allow it to be far more exploratory, as demonstrated by improved state coverage and increased unsupervised performance on hard exploration games such as Montezuma's Revenge.
  • Supplementary Material: pdf
  • Code Of Conduct: I certify that all co-authors of this work have read and commit to adhering to the NeurIPS Statement on Ethics, Fairness, Inclusivity, and Code of Conduct.
  • Code: https://github.com/deepmind/EDDICT
20 Replies