Intervention-based Recurrent causal Model for Nonstationary Video Causal Discovery

Anonymous

Keywords: causal discovery

TL;DR: Intervention-based Recurrent causal Model for Nonstationary Video Causal Discovery

Abstract:

Nonstationary causal structures are prevalent in real-world physical systems. For example, the stacked blocks interact until they fall apart, while the billiard balls move independently until they collide. However, most video causal discovery methods can not discover such nonstationary casual structures due to the lack of modeling for the instantaneous change and the dynamics of the causal structure.
In this work, we propose the Intervention-based Recurrent Casual Model (IRCM) for nonstationary video causal discovery. First, we extend the existing intervention-based casual discovery framework for videos to formulate the instantaneous change of the causal structure in a principled manner. Then, we use a recurrent model to sequentially predict the causal structure model based on previous observations to capture the nonstationary dynamic of the causal structure.
We evaluate our method on two popular physical system simulation datasets with various types of multi-body interactions. Experiments show that the proposed IRCM achieves the state-of-the-art performance on both the counterfactual reasoning and future forecasting tasks.