Virtual presentation / poster accept
Composing Task Knowledge With Modular Successor Feature Approximators
Wilka Carvalho · Angelos Filos · Richard Lewis · Honglak Lee · Satinder Singh
Keywords: [ Reinforcement Learning ] [ generalization ] [ deep reinforcement learning ] [ compositional generalization ] [ successor features ]
Recently, the Successor Features and Generalized Policy Improvement (SF&GPI) framework has been proposed as a method for learning, composing and transferring predictive knowledge and behavior. SF&GPI works by having an agent learn predictive representations (SFs) that can be combined for transfer to new tasks with GPI. However, to be effective this approach requires state features that are useful to predict, and these state-features are typically hand-designed. In this work, we present a novel neural network architecture, “Modular Successor Feature Approximators” (MSFA), where modules both discover what is useful to predict, and learn their own predictive representations. We show that MSFA is able to better generalize compared to baseline architectures for learning SFs and a modular network that discovers factored state representations.