Causality Discovery


Causality is a fundamental research topic studying cause-effect relationships among different components of a system and causality study can help explain why the system has certain behaviors. Data-driven causality learning/discovery has been widely studied and applied in many disciplines including climatology and neuroscience. For instance, our study shows ENSO phenomenon causes surface air temperature in many remote areas using a data-driven algorithm, which is consistent with climate models.

There are many state-of-art approaches and applications of solving the causality discovery problem, yet they still face many challenges when they are used to discover causality from the large-scale and complex climate observation and simulation datasets. The three challenges we are addressing include: 1) computing complexity: how to learn/update causality graph from nonlinear or hybrid climate datasets with increasing data size and dimensionality, 2) uncertainty: how to deal with different causality outputs generated by existing causality discovery methods, 3) reproducibility: how to achieve reproducible causality discovery by leveraging cloud computing services.


Open Source Github Repositories