Multi-Trek Separation in Linear Structural Equation Models 

Wednesday, February 26, 2020 - 4:00pm to 4:30pm

Event Calendar Category

LIDS & Stats Tea

Speaker Name

Jean Baptiste Seby



Building and Room Number

LIDS Lounge


Building on the theory of causal discovery from observational data, we study interactions between multiple (sets of) random variables in a linear structural equation model with non-Gaussian error terms. We give a correspondence between the structure in the higher-order cumulants and combinatorial structure in the causal graph. It has previously been shown that low rank of the covariance matrix corresponds to trek separation in the graph. Generalizing this criterion to multiple sets of vertices, we characterize when determinants of subtensors of the higher-order cumulant tensors vanish. This criterion applies when hidden variables are present as well. For instance, it allows us to identify the presence of a hidden common cause of k of the observed variables.

Joint work with Elina Robeva.


Jean-Baptiste graduated from the Ecole Polytechnique (Paris, France) and joined MIT in the Technology & Policy Program. He is currently a Master's student in Computer Science advised by Prof. Ali Jadbabaie, within IDSS (Institute for Data, Systems, and Society). His research focuses broadly on probability theory, statistics and graphical models.