Visual Causal Feature Learning

TL;DR

This paper introduces a rigorous framework for identifying visual causes of behavior across various systems, leveraging causal coarsening and active learning to efficiently discover causal features from visual data.

Contribution

It generalizes causal learning to construct causal variables from micro-variables, proves the Causal Coarsening Theorem, and develops active learning methods for visual cause identification.

Findings

The Causal Coarsening Theorem enables causal inference from observational data.

Active learning scheme efficiently identifies visual causes through optimal image manipulations.

Algorithms demonstrated effective on both synthetic and real datasets.

Abstract

We provide a rigorous definition of the visual cause of a behavior that is broadly applicable to the visually driven behavior in humans, animals, neurons, robots and other perceiving systems. Our framework generalizes standard accounts of causal learning to settings in which the causal variables need to be constructed from micro-variables. We prove the Causal Coarsening Theorem, which allows us to gain causal knowledge from observational data with minimal experimental effort. The theorem provides a connection to standard inference techniques in machine learning that identify features of an image that correlate with, but may not cause, the target behavior. Finally, we propose an active learning scheme to learn a manipulator function that performs optimal manipulations on the image to automatically identify the visual cause of a target behavior. We illustrate our inference and learning algorithms in experiments based on both synthetic and real data.