Invariant Risk Minimization

Slides: https://github.com/bayesgroup/bayesgroup.github.io/blob/master/bmml_sem/2019/Kodryan_Invariant%20Risk%20Minimization.pdf Maxim Kodryan, Research Intern at Samsung-HSE lab This talk is dedicated to the correlation-versus-causation dilemma. Minimizing training error leads machines into recklessly absorbing all the correlations found in training data. Understanding which patterns are actually useful (causal) is important if we want our models to generalize to new test distributions. It seems that there exists an intimate link between invariance and causation useful for generalization. We will consider the concept of Invariant Risk Minimization (as opposed to Empirical Risk Minimization) — a novel learning paradigm that estimates nonlinear, invariant, causal predictors from multiple training environments, to enable out-of-distribution generalization. We will also provide an information-theoretic view on the topic.