Regularizing towards Causal Invariance: Linear Models with Proxies
Michael Oberst 1 Nikolaj Thams 2 Jonas Peters 2 David Sontag 1
Abstract A
βx βy
We propose a method for learning linear mod-
α
els whose predictive performance is robust to X Y
causal interventions on unobserved variables,
when noisy ...


雷达卡




京公网安备 11010802022788号







