2025
Student Seminar: Joonhyung Shin
2:00–3:30 pm DSI Building, Room 301
Monday, December 1, 2025, at 2:00 PM, in Room 301, 5460 S. University Ave
Dissertation Proposal Presentation
Joonhyung Shin, Department of Statistics, The University of Chicago
“Statistical problems with computational hardness: a statistical physics approach”
Statistics Colloquium: Matteo Sesia
11:30 am–12:30 pm Jones 303
Matteo Sesia, Department of Data Sciences and Operations, University of Southern California
Title: “Conformal Inference for Open-Set and Imbalanced Classification”
Abstract: This talk presents a conformal prediction method for classification in highly imbalanced and open-set settings, where there are many possible classes and not all may be represented in the data. Existing approaches require a finite, known label space and typically involve random sample splitting, which works well when there is a sufficient number of observations from each class. Consequently, they have two limitations: (i) they fail to provide adequate coverage when encountering new labels at test time, and (ii) they may become overly conservative when predicting previously seen labels. To obtain valid prediction sets in the presence of unseen labels, we compute and integrate into our predictions a new family of conformal p-values that can test whether a new data point belongs to a previously unseen class. We study these p-values theoretically, establishing their optimality, and uncover an intriguing connection with the classical Good–Turing estimator for the probability of observing a new species. To make more efficient use of imbalanced data, we also develop a selective sample splitting algorithm that partitions training and calibration data based on label frequency, leading to more informative predictions. Despite breaking exchangeability, this allows maintaining finite-sample guarantees through suitable re-weighting. With both simulated and real data, we demonstrate our method leads to prediction sets with valid coverage even in challenging open-set scenarios with infinite numbers of possible labels, and produces more informative predictions under extreme class imbalance.
Student Seminars: Ziyang Wei
11:00 am–12:30 pm Jones 304
Tuesday, November 18, 2025, at 11:00 AM, in Jones 304, 5747 S. Ellis Avenue
Dissertation Proposal Presentation
Ziyang Wei, Department of Statistics, The University of Chicago
“Asymptotic and finite-sample theory for stochastic gradient descent”
Statistics Colloquium: Aaditya Ramdas
11:30 am–12:30 pm Jones 303
Aaditya Ramdas, Department of Statistics and Data Science and Machine Learning, Carnegie Mellon University
Title: “Bringing closure to FDR control: a general principle for multiple testing”
Abstract: The closure principle (Biometrika’76) states that every procedure for controlling the familywise error rate (FWER) can be recovered or improved via “closed testing”. Since the publication of the seminal Benjamini-Hochberg paper (the most cited paper in statistics), it has been an open problem how the “closure principle” applies to controlling the false discovery rate (FDR). We fully settle this open problem by developing a closure principle not only for FDR, but every error metric that is an expectation (including the classical one for FWER as a special case). Our developments crucially hinge on the modern concept of e-values, which perhaps explains the delay in its discovery. This theoretical advance has immediate implications for practice: it leads to surprising improvements to both modern and classical FDR methods (eg: Benjamini-Yekutieli’s famous 2001 procedure is strictly improved, as is the e-Benjamini-Hochberg procedure), and it also allows for practitioners to choose the error metric and error level post-hoc.
https://arxiv.org/abs/2509.02517 is the preprint, joint work with Ziyu Xu, Aldo Solari, Lasse Fischer, Rianne de Heide, Jelle Goeman (it is actually a merge of two simultaneous papers).
Student Seminar: John Hood
9:00 am–11:00 pm Jones 111
Monday, November 17, 2025, at 9:00 AM, in Jones 111, 5747 S. Ellis Avenue
Dissertation Proposal Presentation
John Hood, Department of Statistics, The University of Chicago
“Scalable Nonnegative Tensor Decomposition Models for Latent Structure Discovery in Large Complex Networks”
Student Seminar: Ian Joffe
8:00–8:30 am Jones 111
Monday, November 17, 2025, at 8:00 AM, in Jones 111, 5747 S. Ellis Avenue
Master’s Thesis l Presentation
Ian Joffe, Department of Statistics, The University of Chicago
“Searching for Linear Representation of Political Sentiment in Large Language Models”
Student Seminar: Nimish Janey
1:00–1:30 pm Jones 111
Thursday, November 13, 2025, at 1:00 PM, in Jones 111, 5747 S. Ellis Avenue
Master’s Thesis l Presentation
Nimish Janey, Department of Statistics, The University of Chicago
“Expected Shortfall Under Historical Simulation: A Monte Carlo Study ofConditional and Unconditional Backtests”
Student Seminar: Dillon Jones
12:30–1:00 pm Jones 111
Thursday, November 13, 2025, at 12:30 PM, in Jones 111, 5747 S. Ellis Avenue
Master’s Thesis l Presentation
Dillan Jones, Department of Statistics, The University of Chicago
“Robust and Range-Based GARCH Models with Leverage Effects: A Study of VaR–ES Forecasting”
Student Seminar: Nick Bourdeau
12:00–12:30 pm Jones 111
Thursday, November 13, 2025, at 12:00 PM, in Jones 111, 5747 S. Ellis Avenue
Master’s Thesis l Presentation
Nick Bourdeau, Department of Statistics, The University of Chicago
“Information-Theoretic Feature Discovery and Forecasting of US Yield Curve Movements”
Student Seminar: Ping Chi Shih
10:00 am–10:30 pm Jones 111
Thursday, November 13, 2025, at 10:00 AM, in Jones 111, 5747 S. Ellis Avenue
Master’s Thesis l Presentation
Ping Chi Shih, Department of Statistics, The University of Chicago
“Task-Aligned Factors: Supervised Autoencoders for CPI Forecasting and Futures Cross-Sections”