News

Our paper Geodesic Difference-in-Differences will appear in Biometrika. It extends difference-in-differences methods to structured outcomes, including distributions, networks, and manifold-valued data, with applications to age-at-death distributions and electricity-generation compositions.

The Journal of Machine Learning Research has accepted our paper Sliced Wasserstein Regression. We develop global and local regression methods for multivariate distributional responses using the sliced Wasserstein distance, with theoretical guarantees and applications to excess winter mortality and financial data.

Deep Single-Index Fréchet Regression has been accepted to ICML 2026.

End-to-End Deep Learning for Predicting Metric Space-Valued Outputs will appear in the Journal of Machine Learning Research. The paper introduces E2M, an end-to-end deep learning framework that predicts metric space-valued outputs using geometry-aware weighted Fréchet means.

Deep Fréchet Regression has been selected as the discussion paper for the Journal of the American Statistical Association Theory & Methods Invited Session at the 2026 Joint Statistical Meetings. The session will take place on Monday, August 3rd (10:30am–12:20pm) and will include invited discussions, contributed comments, and a rejoinder.

Two of our papers will be presented at NeurIPS 2025:

• Fréchet Geodesic Boosting: We introduce FGBoost, a gradient boosting framework designed to intrinsically model complex regression relationships with non-Euclidean outputs in geodesic metric spaces.
• Wasserstein Transfer Learning: We propose a novel transfer learning framework for regression where outputs are probability distributions residing in the Wasserstein space.

Our paper Deep Fréchet Regression has been accepted for publication in the Journal of the American Statistical Association. It develops a deep learning framework for modeling complex regression relationships when responses are random objects in general metric spaces.