Intrinsic Physical Concepts Discovery with Object-Centric Predictive Models

TL;DR

This paper introduces PHYCINE, a system that unsupervisedly discovers and represents intrinsic physical concepts like mass and charge, advancing object-centric models for better physical reasoning.

Contribution

The paper presents a novel unsupervised approach to infer and reason about intrinsic physical concepts at multiple abstraction levels using object-centric predictive models.

Findings

Variables inferred align with physical properties.

Discovered concepts improve causal reasoning performance.

System captures concepts like mass and charge without supervision.

Abstract

The ability to discover abstract physical concepts and understand how they work in the world through observing lies at the core of human intelligence. The acquisition of this ability is based on compositionally perceiving the environment in terms of objects and relations in an unsupervised manner. Recent approaches learn object-centric representations and capture visually observable concepts of objects, e.g., shape, size, and location. In this paper, we take a step forward and try to discover and represent intrinsic physical concepts such as mass and charge. We introduce the PHYsical Concepts Inference NEtwork (PHYCINE), a system that infers physical concepts in different abstract levels without supervision. The key insights underlining PHYCINE are two-fold, commonsense knowledge emerges with prediction, and physical concepts of different abstract levels should be reasoned in a bottom-up fashion. Empirical evaluation demonstrates that variables inferred by our system work in accordance with the properties of the corresponding physical concepts. We also show that object representations containing the discovered physical concepts variables could help achieve better performance in causal reasoning tasks, i.e., ComPhy.