Variational Temporal Abstraction

NeurIPS Poster (2019)

초록

We introduce a variational approach to learning and inference of temporally hierarchical structure and representation for sequential data. We propose the Variational Temporal Abstraction, a hierarchical recurrent state space model that can infer the latent temporal structure and thus perform the stochastic state transition hierarchically. We also propose to apply this model to implement the jumpy-imagination ability in imagination-augmented agent-learning in order to improve the efficiency of the imagination. In experiments, we demonstrate that our proposed method can model 2D and 3D visual sequence datasets with interpretable temporal structure discovery and that its application to jumpy imagination enables more efficient agent-learning in a 3D navigation task.

저자

김태섭(MILA/카카오브레인), 안성진 (럿거스 대학교), Yoshua Bengio (MILA)

키워드

Generative Models AGI

발행 날짜

2019.06.14