English

A Compositional Object-Based Approach to Learning Physical Dynamics

Artificial Intelligence 2017-03-07 v2 Machine Learning

Abstract

We present the Neural Physics Engine (NPE), a framework for learning simulators of intuitive physics that naturally generalize across variable object count and different scene configurations. We propose a factorization of a physical scene into composable object-based representations and a neural network architecture whose compositional structure factorizes object dynamics into pairwise interactions. Like a symbolic physics engine, the NPE is endowed with generic notions of objects and their interactions; realized as a neural network, it can be trained via stochastic gradient descent to adapt to specific object properties and dynamics of different worlds. We evaluate the efficacy of our approach on simple rigid body dynamics in two-dimensional worlds. By comparing to less structured architectures, we show that the NPE's compositional representation of the structure in physical interactions improves its ability to predict movement, generalize across variable object count and different scene configurations, and infer latent properties of objects such as mass.

Keywords

Cite

@article{arxiv.1612.00341,
  title  = {A Compositional Object-Based Approach to Learning Physical Dynamics},
  author = {Michael B. Chang and Tomer Ullman and Antonio Torralba and Joshua B. Tenenbaum},
  journal= {arXiv preprint arXiv:1612.00341},
  year   = {2017}
}

Comments

Published as a conference paper for ICLR 2017. 15 pages, 6 figures

R2 v1 2026-06-22T17:10:51.716Z