IFML's signature Seminar Series: Spring 2026 Recap

This spring 2026, IFML hosted its Semester Seminar Series given by speakers from four academic institutions, including two from Yale, and one research laboratory. Designated by the National Science Foundation (NSF) in 2020, the Institute for Foundations of Machine Learning (IFML) develops the key foundational tools for the next decade of Al innovation, and guests of IFML faculty give talks on research that push AI research forward. Well attended in person and by Zoom by professors and students, the seminars can also be accessed post-event through recorded talks on IFML’s YouTube channel. 

Jiaxin Shi, former research scientist at Google DeepMind, kicked off the spring semester series with his talk on “Are Diffusion and Autoregression Truly Different? Insights from Masked Diffusion Models." Watch talk here. 

Larry Wasserman from Carnegie Mellon University spoke about particle physicists developing an algorithm called COWs (Customized Orthogonal Weights) for separating signals from backgrounds in certain experiments. They look at COWs from a statistical perspective and consider several extensions of the method. In particular, a modified version of the method leads to a robust method for estimating arbitrary mixtures of conditionally independent distributions. Check out his talk here.

Andre Wibisono of Yale University gave a talk on “A survey of the mixing times of the Proximal Sampler algorithm,” (check out video here), while IFML’s postdoctoral fellow, Arsen Vasilyan, spoke on “Foundations of Reliable Learning with Imperfect Data" (watch video here). 

Manolis Zampetakis from Yale University gave a talk on “Learning Mixture Models via Efficient High-dimensional Sparse Fourier Transforms" (watch video here) and Kevin Tian, assistant professor of Computer Science at UT Austin, shared on his joint work with Syamantak Kumar, Purnamrita Sarkar, and Yusong Zhu: “High-Magnetization Sampling at Low Temperatures: Ising Models and Bayesian Sparse Linear Regression” (check out talk here).

Visiting professor from the University of Minnesota Mehmet Akçakaya wrapped up the semester series nicely with his talk on “Learning and guidance approaches for generative and physics-driven models in computational MRI” (watch video here).

Looking ahead, IFML will continue to host and convene researchers at the cutting edge of AI to the UT community and looks forward to dynamic presentations in the upcoming academic year.