We are using ideas inspired by causal inference to address a difficult problem in machine learning: unsupervised domain adaptation. For example, we wish to train on data from one hospital and succeed on other, unseen hospitals; or train on images from one setting and test on images from many different settings.
We are building theoretical and practical models that take as input both a mechanistic world model (for example and ordinary differential equation describing the cardio-vascular system) and data (for example ICU patient vital signs). The goal is to get the best of both worlds: the robustness, interpretability, and causal grounding of mechanistic models, together with the flexibility of black-box deep learning models.
In collaboration with health providers such as Clalit Health Services and Rambam Health Campus we are developing individual-level causal inference tools that will give accurate and safe treatment recommendations to patients.