Each year, the ÌÇÐÄlogo Information Theory Society organizes the North American School of Information Theory (NASIT), a gathering designed for graduate students and postdoctoral researchers from across North America and around the world. The school provides a unique opportunity for participants to interact with leading experts, exchange ideas, and gain first-hand insights into the latest advances in information theory and its related disciplines.
NASIT 2025 started on Monday, June 16th featuring three tutorials. The first one, presented by Prof. Ilan Shomorony (University of Illinois at Urbana Champaign), explored how information theory can be used to address some fundamental challenges in genomic data processing and related problems. Prof. Nihar Shah (Carnegie Mellon University) presented the next tutorial on the theory of peer review and related experiments, highlighting theoretically grounded solutions to address challenges. The last tutorial of the first day was given by the 2025 Goldsmith Lecturer, Prof. Gauri Joshi (Carnegie Mellon University), who provided an insightful overview on optimization algorithms for federated learning in heterogeneous environments and multi-tasking learning. The day concluded with a welcome reception featuring Mediterranean cuisine.
Tuesday’s program opened with a tutorial on private distributed computing offered by Prof. Viveck Cadambe (Georgia Institute of Technology), who discussed fundamental trade-offs between accuracy and privacy when executing computational tasks across unreliable distributed servers. The next tutorial was an elegant chalkboard lecture by the 2025 Padovani Lecturer, Prof. Emre Telatar (École Polytechnique Fédérale de Lausanne), who provided a deep dive into the theory of universal compression, highlighting thought-provoking connections to learning theory. The day concluded with a tutorial by Prof. Sanghamitra Dutta (University of Maryland) who gave a lecture on explainable machine learning and its interplay with information theory.
Ìý
Ìý
Wednesday began with a tutorial by Prof. Ramya Vinayak (University of Wisconsin), who presented the foundations of comparison-based preference and metric learning for pluralistic alignment in heterogeneous environments. The second tutorial, delivered by Prof. Ali Tajer (Rensselaer Polytechnic Institute), focused on methods for estimating both the latent ground-truth representations of data and the structures governing their interactions. The final tutorial of the day was given by prof. Qian Yu (University of California, Santa Barbara), who demonstrated how information-theoretic tools can be used to design sample-optimal algorithms for stochastic optimization problems. The day concluded with a delightful banquet, providing a memorable social highlight and offering a scenic view of Minnesota’s natural beauty at the height of summer.
Thursday featured a thought-provoking tutorial by Prof. Elza Erkip (New York University) on learning techniques for compression and communication across distributed terminals. In the afternoon, participants enjoyed a boat cruise on the mighty Mississippi River, complete with food, drinks, and a firsthand look at the operation of the St. Anthony lock and dam system.
Ìý
Ìý
Friday, the final day of NASIT, was deliberately kept short to allow participants greater flexibility in their departure plans. The program featured a tutorial by Prof. Mahdi Soltanolkotabi (University of Southern California), who presented an in-depth mathematical foundation of generative AI and demonstrated how it can address challenging problems such as feature learning, prompting, and enhancing inference-time reasoning.
Each day, NASIT provided breakfast and lunch, which participants enjoyed together. On Monday and Tuesday, attendees also presented their ongoing work in poster sessions held during the lunch break, creating valuable opportunities to engage with fellow participants and speakers.Ìý
Abstracts, videos, and slides of the tutorials, as well as the posters, are available at no cost at
The success of NASIT 2025 was made possible through the generous support of the ÌÇÐÄlogo Information Theory Society, the National Science Foundation, Seagate Technology, UMN’s Data Science Initiative, and the ECE Department at UMN. The Organizing Committee was chaired by Prof. Martina Cardone, Prof. Mohammad Ali Maddah-Ali, and Prof. Soheil Mohajer, with Prof. Shirin Jalali (Rutgers University) serving as Financial Chair. Special thanks also to Prof. Shirin Saeedi Bidokhti and Prof. Hamed Hassani (University of Pennsylvania) for their feedback and suggestions based on their experience organizing NASIT 2023. The success of this event would not have been possible without the invaluable help and tireless efforts of Jeanine Maiden (UMN ECE Department).