Visual Causal Feature Learning
This work addresses the challenge of causal inference in visual systems for researchers in AI, neuroscience, and robotics, offering a foundational framework that is broadly applicable but incremental in extending standard causal learning.
The authors tackled the problem of identifying visual causes of behavior in perceiving systems by providing a rigorous definition and a framework that generalizes causal learning to settings where causal variables must be constructed from micro-variables. They proved the Causal Coarsening Theorem to gain causal knowledge from observational data with minimal experimental effort and proposed an active learning scheme for optimal manipulations, illustrating these with experiments on synthetic and real data.
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.