Video Content by Date
Speaker: Paweł Morzywolek
We provide an inferential framework to assess variable importance for heterogeneous treatment effects. This assessment is especially useful in high-risk domains such as medicine, where decision makers hesitate to rely on black-box treatment recommendation algorithms. The variable importance measures we consider are local in that they may differ across individuals, while the inference is global in that it tests whether a given variable is important for any individual. Our approach builds on recent developments in semiparametric theory for function-valued parameters, and is valid even when statistical machine learning algorithms are employed to quantify treatment effect heterogeneity. We demonstrate the applicability of our method to infectious disease prevention strategies.
Speaker: Vinod Vaikuntanathan
Integer lattices play a central role in mathematics and computer science, with applications ranging from number theory and coding theory to combinatorial optimization. Over the past three decades, they have also become a cornerstone of modern cryptography.
In this talk, I will describe the evolution of lattices in cryptography: from the early use of lattices to break classical cryptosystems; to their application in designing new encryption and digital signature schemes with (conjectured) post-quantum security; and to their role in achieving long-standing cryptographic goals such as fully homomorphic encryption that allow us to compute directly on encrypted data.
The talk will not assume any prior background in cryptography.
Speaker: Trisha Lawrence
The global population with access to electricity is constantly increasing from 84 to 92 percent. However, as the world continues to advance towards sustainable energy targets, there still exist 900 million people living without access to electricity.
In this talk, we provide a modeling framework for analyzing mini‑grid project performance and evaluating the economic impact of battery energy storage in competitive electricity markets. Using a dataset of 104 rural mini‑grid installations, we estimate the probability of project success through both a Probit regression model and a Bayesian hierarchical model. Community ownership and the presence of storage systems emerge as statistically significant predictors, with Bayesian posterior estimates closely aligning with frequentist results while providing improved predictive stability.
Furthermore, we analyze the bidding behavior of the Alberta electricity market and construct a mixed‑integer self‑scheduling model that determines optimal charging and discharging strategies. Through numerical experiments we demonstrate how storage can enhance arbitrage profitability, influence market clearing prices, and support system reliability.
Our results highlight the value of combining statistical inference with optimization‑based modeling to guide investment decisions.
Speaker: Puneet Velidi
Advances in protein folding and structure prediction models have enabled new computational approaches to immunotherapeutic research by providing access to high-quality structural information at scale. In this talk, we present three core application areas. (1) Antigen structure prediction, where folding models are used to characterize the three-dimensional structure of viral, tumor-associated, and neoantigen targets in the absence of experimental data. (2) Antibody–antigen complex prediction, where multimeric and joint modeling approaches are leveraged to infer binding modes, paratope–epitope interactions, and structural determinants of specificity. (3) Immunogenicity prediction, where predicted structures are analyzed to assess surface accessibility, conformational stability, and geometric features that influence immune recognition. Together, these applications illustrate how protein folding models function not only as structure predictors, but as foundational components in quantitative pipelines for immunotherapeutic discovery and design.
Speaker: Kostya Druzhkov
Differential equations can be studied from a purely geometric point of view, translating many constructions from finite-dimensional differential geometry into their language. This approach helps to clarify such notions as symmetries, conservation laws, presymplectic structures, and others. However, a number of questions arise in this framework whose answers are either incomplete or currently unknown. In particular, the problem of defining the cotangent equation in terms of the intrinsic geometry of PDEs remains open. This problem is directly related to the Hamiltonian formalism for differential equations.
From an applied perspective, methods for constructing exact solutions of differential equations are of particular interest. One of the most powerful approaches is based on the study of solutions invariant under certain symmetries of the given equation. A question of practical importance in this context is how the systems describing such invariant solutions inherit geometric structures from the original system.
In this talk, I will explain how these two topics are brought together within a reduction mechanism, which in particular clarifies how Hamiltonian operators are inherited by systems describing solutions of a given equation that are invariant under some of its symmetries. To fully implement this mechanism, an interpretation of cotangent equations in intrinsic geometric terms is also required. This can be achieved in the case where the reduced system turns out to be finite-dimensional.
Speaker: Valentina Zapata Castro
Model categories provide a powerful framework for abstract homotopy theory, but their complexity often makes them difficult to classify. By focusing on finite categories, especially grids, we gain a combinatorial setting where the problem becomes explicit. In this talk, we explore model structures through weak factorization systems (WFS) on posets, which are in one-to-one correspondence with transfer systems and their duals, both introduced here. This perspective leads to a method for constructing model structures and a characterization theorem for finding weak equivalence sets in posets. Our approach offers a pathway towards classifying model structures in a controlled setting.
This is joint work with Kristen Mazur, Angélica Osorno, Constanze Roitzheim, Rekha Santhanam and Danika Van Niel.
