Specify Robust Causal Representation from Mixed Observations
This work addresses the need for more robust machine learning models in applications prone to adversarial attacks and distribution shifts, though it appears incremental as it builds on existing causal representation learning methods.
The paper tackles the problem of learning robust and generalizable representations from observational data by proposing a method to learn causal representations based on a hypothetical causal graph, showing that models using these representations are more robust to adversarial attacks and distribution shifts compared to baselines.
Learning representations purely from observations concerns the problem of learning a low-dimensional, compact representation which is beneficial to prediction models. Under the hypothesis that the intrinsic latent factors follow some casual generative models, we argue that by learning a causal representation, which is the minimal sufficient causes of the whole system, we can improve the robustness and generalization performance of machine learning models. In this paper, we develop a learning method to learn such representation from observational data by regularizing the learning procedure with mutual information measures, according to the hypothetical factored causal graph. We theoretically and empirically show that the models trained with the learned causal representations are more robust under adversarial attacks and distribution shifts compared with baselines. The supplementary materials are available at https://github.com/ymy $4323460 / \mathrm{CaRI} /$.