Symmetry-Based Disentangled Representation Learning requires Interaction with Environments

Hugo Caselles-Dupré (1, 2, 3) , Michael Garcia-Ortiz (1) , David Filliat (2, 3)

1 SoftBank Robotics Europe
2 U2IS - Unité d'Informatique et d'Ingénierie des Systèmes
3 Flowers - Flowing Epigenetic Robots and Systems

Finding a generally accepted formal definition of a disentangled representation in the context of an agent behaving in an environment is an important challenge towards the construction of data-efficient autonomous agents. Higgins et al. (2018) recently proposed Symmetry-Based Disentangled Representation Learning, a definition based on a characterization of symmetries in the environment using group theory. We build on their work and make observations, theoretical and empirical, that lead us to argue that Symmetry-Based Disentangled Representation Learning cannot only be based on static observations: agents should interact with the environment to discover its symmetries. Our experiments can be reproduced in Colab and the code is available on GitHub.

Domaines: Machine Learning [stat.ML]

Hugo Caselles-Dupré, Michael Garcia-Ortiz, David Filliat. Symmetry-Based Disentangled Representation Learning requires Interaction with Environments. NeurIPS 2019 6 Neural Information Processing Conference, Dec 2019, Vancouver, Canada. ⟨hal-02379399⟩