Policy manifold search

Abstract

Neuroevolution is an alternative to gradient-based optimisation that has the potential to avoid local minima and allows parallelisation. The main limiting factor is that usually it does not scale well with parameter space dimensionality. Inspired by recent work examining neural network intrinsic dimension and loss landscapes, we hypothesise that there exists a low-dimensional manifold, embedded in the policy network parameter space, around which a high-density of diverse and useful policies are located. This paper proposes a novel method for diversity-based policy search via Neuroevolution, that leverages learned representations of the policy network parameters, by performing policy search in this learned representation space. Our method relies on the Quality-Diversity (QD) framework which provides a principled approach to policy search, and maintains a collection of diverse policies, used as a dataset for learning policy representations. Further, we use the Jacobian of the inversemapping function to guide the search in the representation space. This ensures that the generated samples remain in the high-density regions, after mapping back to the original space. Finally, we evaluate our contributions on four continuous-control tasks in simulated environments, and compare to diversity-based baselines. CCS CONCEPTS• Computing methodologies → Dimensionality reduction and manifold learning; Evolutionary robotics; Neural networks.

Evolving Populations of Diverse RL Agents with MAP-Elites

“…In this work, we use the isoline variation operator (Vassiliades & Mouret, 2018) that, given two parent policies, say policies θ 1 and θ 2 , adds Gaussian noise N (0, σ 1 ) to θ 1 and offsets the results along the line θ 2 −θ 1 by a magnitude randomly sampled from a zero-mean Gaussian distribution with variance N (0, σ 2 ). This strategy has proved to be particularly effective to evolve neural networks (Rakicevic et al, 2021). Pseudocode for MAP-ELITES is provided in the Appendix.…”