Neural Engineering Journal Club (Thursday Oct. 12)

Event Time and Location: Thursday 10/12 at noon in CDI 3.352

Lukas Hirsch will lead a discussion of causal inference, including the attached paper:
“Nonlinear causal discovery with additive noise models ”

Abstract
The discovery of causal relationships between a set of observed variables is a fun damental problem in science. For continuous-valued data linear acyclic causal models with additive noise are often used because these models are well under-stood and there are well-known methods to fit them to data. In reality, of course, many causal relationships are more or less nonlinear, raising some doubts as to the applicability and usefulness of purely linear methods. In this contribution we show that the basic linear framework can be generalized to nonlinear models. In this extended framework, nonlinearities in the data-generating process are in fact a blessing rather than a curse, as they typically provide information on the underlying causal system and allow more aspects of the true data-generating mechanisms to be identified. In addition to theoretical results we show simulations and some simple real data experiments illustrating the identification power provided by non-linearities.