- talks - lists.cs.princeton.edu

Yann LeCun's talk at ORFE
by Kai Li 07 Apr '26

07 Apr '26

From: "Ariana M Reyes" <ar8562(a)princeton.edu> To: "Bartolomeo Stellato" <bstellato(a)princeton.edu>, "Kai Li" <li(a)cs.princeton.edu> Cc: "Connie A. Brown" <connieb(a)princeton.edu> Sent: Tuesday, April 7, 2026 12:40:28 PM Subject: RE: Meeting with Yann LeCun Dear Kai, We are glad that you were able to sign up for an individual meeting and to attend dinner. Below is the announcement that we sent out if you can please circulate it to CS students and faculty as suggested. Thank you! ……………………………………………………………………………………………………………………… Today’s ORFE Department Colloquia Date & Time: Tuesday, April 7, 2026 at 4:30 PM Location: 101 - Friend Center Speaker: [ https://orfe.princeton.edu/events/2026/yann-lecun-new-york-university | Yann LeCun, New York University ] Bio: Yann LeCun is the Executive Chairman of AMI Labs and the Jacob T. Schwartz Professor at NYU affiliated with the Courant Institute - School of Mathematics, Computer Science, and Data Science. He was the Chief AI Scientist at Meta (2018-2025), the founding Director of Meta-FAIR (2013-2017), and of the NYU Center for Data Science (2011-2014). He received an Engineering Diploma from ESIEE (Paris) and a PhD from Sorbonne Université (1987). After a postdoc in Toronto he joined AT&T Bell Labs in 1988, and AT&T Labs in 1996 as Head of Image Processing Research. He joined NYU as a professor in 2003 and Meta/Facebook in 2013. His interests include AI, machine learning, computer perception, robotics, and computational neuroscience. He is the recipient of the 2018 ACM Turing Award (with Geoffrey Hinton and Yoshua Bengio) for "conceptual and engineering breakthroughs that have made deep neural networks a critical component of computing", a member of the National Academy of Sciences, the National Academy of Engineering, the French Académie des Sciences. ………………………………………………………………………………………………………………………….. Ariana Reyes Administrative Assistant & Events Operations Research and Financial Engineering (ORFE) Princeton University Sherrerd Hall, Room 119, Princeton NJ 08544 609-258-9076 ar8562(a)princeton.edu From: Bartolomeo Stellato <bstellato(a)princeton.edu> Sent: Saturday, April 4, 2026 10:36 AM To: Kai Li <li(a)cs.princeton.edu> Cc: Ariana M Reyes <ar8562(a)princeton.edu>; Connie A. Brown <connieb(a)princeton.edu> Subject: Re: Meeting with Yann LeCun Dear Kai, Glad you were able to sign up for a meeting and dinner! Sherrerd 117 is the visitor room assigned to Yann, so it's the most convenient one for meetings. That should work well for your 1:30pm slot. Let us know if you have any questions. Best, Bartolomeo On Sat, Apr 4, 2026 at 12:05 AM Kai Li < [ mailto:li@cs.princeton.edu | li(a)cs.princeton.edu ] > wrote: Dear Bartolomeo, I signed up for 1:30pm slot in Sherred 117. I saw Elad meets Yann there. Is it the room for me to meet him or should I meet him in my office in CS building? I also signed up for dinner. Thank you! Kai ----- On Apr 3, 2026, at 7:55 PM, Bartolomeo Stellato < [ mailto:bstellato@princeton.edu | bstellato(a)princeton.edu ] > wrote: BQ_BEGIN Dear Kai, Thanks for reaching out. We still have a couple of slots available on the schedule, and you are welcome to sign up for lunch or dinner as well. The sign-up sheet is here: [ https://princetonu-my.sharepoint.com/:x:/g/personal/orfe-onedrive_princeton… | https://princetonu-my.sharepoint.com/:x:/g/personal/orfe-onedrive_princeton… ] The talk page is here: [ https://orfe.princeton.edu/events/2026/yann-lecun-new-york-university | https://orfe.princeton.edu/events/2026/yann-lecun-new-york-university ] . We are still waiting for the final title and abstract, but we can let you know as soon as they are available. Happy to have you circulate it to CS, of course. Looking forward to meeting you on Tuesday. Best, Bartolomeo On Fri, Apr 3, 2026, 8:15 PM Kai Li < [ mailto:li@cs.princeton.edu | li(a)cs.princeton.edu ] > wrote: BQ_BEGIN Dear Ariana, Connie and Bartolomeo, This is Kai Li, a faculty member at Computer Science. Sean at NYU told me that Yann LeCun is giving a talk at ORFE and you mentioned to Sean that it would be nice for me to meet Yann. For some reason, I couldn't find any message about this in my email. I have known Yann for over 10 years and would be great to see him again. I am flexible to meet him on April 7th. Please let me know what time slots are available. If you would like me to circulate his talk to CS faculty and students, please send me info about his talk so that I can forward. Best wishes, Kai Kai Li Paul & Marcia Wythes Professor Department of Computer Science Princeton University [ https://www.google.com/maps/search/35+Olden+street+Princeton,+NJ+08544?entr… | 35 Olden street ] [ https://www.google.com/maps/search/35+Olden+street+Princeton,+NJ+08544?entr… | Princeton, NJ 08544 ] (609) 258-4637 BQ_END BQ_END

1 0

Ilya Maier will present his MSE thesis “Induced Cycles of Many Lengths" on Tuesday, April 21, 2026 at 2:30pm in Fine Hall 224.
by Gradinfo 07 Apr '26

07 Apr '26

Ilya Maier will present his MSE thesis “Induced Cycles of Many Lengths" on Tuesday, April 21, 2026 at 2:30pm in Fine Hall 224. Adviser: Maria Chudnovsky, Reader: Pravesh Kothari. All are welcome to attend. Abstract: Let $G$ be a graph and let $\mathrm{cl}(G)$ be the number of distinct induced cycle lengths in $G$. We show that for $c,t\in \mathbb N$, every graph $G$ that does not contain an induced subgraph isomorphic to $K_{t+1}$ or $K_{t,t}$ and satisfies $\mathrm{cl}(G) \le c$ has bounded treewidth. As a consequence, we obtain a polynomial-time algorithm for deciding whether a graph $G$ contains induced cycles of at least three distinct lengths.

1 0

Theresa Lim will present her MSE Thesis: "A Functional Introduction to Computer Science" on April 20, 2026 at 11am in CS 302.
by Gradinfo 07 Apr '26

07 Apr '26

Theresa Lim will present her MSE Thesis "A Functional Introduction to Computer Science" on April 20, 2026 at 11am in CS 302. Adviser: David Walker, Reader: Kevin Wayne All are welcome to attend. Abstract below: In recent years, there has been increased demand for computer science (CS) education at the primary and secondary level. Traditionally, introductory CS courses teach students about an imperative language, such as Java or Python, and students practice writing increasingly complex programs from scratch. However, with the rise of AI-generated code, it is more important than ever that students develop the skill of reasoning about unfamiliar code in a systematic and critical manner. Additionally, many researchers have identified discrete mathematics courses as a common difficulty for college computer science students due to students' unfamiliarity with writing proofs. This work presents a 20-hour introductory computer science curriculum that aims to address both of these problems. The course is tailored to high school students with little to no coding experience, and it teaches students to code in OCaml and reason about program correctness using equational reasoning and proof by induction. We incorporate modern pedagogical techniques, including backward design, differentiated instruction, and culturally-responsive pedagogy. This course not only prepares students for future work in computer science but also strengthens the logical reasoning skills that math and science disciplines require. This abridged course serves as a foundation for future proof-based introductory computer science curricula.

1 0

Computational Math IDeAS Seminar
by Emily C. Lawrence 07 Apr '26

07 Apr '26

COMPUTATIONAL MATH – IDeAS SEMINAR Recurring weekly series · Thursdays, 2:00 – 3:00 PM This week's talk: Speaker: Daniel B. Szyld (Temple) Title: Convergence of Randomized and Greedy Block Gauss-Seidel methods, as well as Asynchronous Iterations Date: Thursday, April 9, 2026 Time: 2:00 PM – 3:00 PM Room: 224 Fine Hall Abstract: We extend results known for the randomized (point and block) Gauss-Seidel and the Gauss-Southwell methods for the case of a Hermitian and positive definite matrix to certain classes of non-Hermitian matrices. We consider cases with overlapping variables (as in Domain Decomposition). We obtain convergence results for a whole range of parameters describing the probabilities in the randomized method or the greedy choice strategy in the Gauss-Southwell-type methods. We identify those choices which make our convergence bounds best possible. One result is that the best convergence bounds that we obtain for the expected values in the randomized algorithm are as good as the best for the deterministic, but more costly algorithms of Gauss-Southwell type. We use these new results to show a provable convergence rate for asynchronous iterations. (Joint work with Andreas Frommer) About the speaker: Daniel B. Szyld is Professor of Mathematics at Temple University. Until recently he was the President of the International Linear Algebra Society (2020-2026). He was Vice-President (at large) of SIAM, the Society for Industrial and Applied Mathematics in 2014-2015. He was the editor-in-chief of the SIAM Journal on Matrix Analysis and Applications (2015-2020) and of the Electronic Transactions on Numerical Analysis (2005-2013). He is a SIAM Fellow, and an AMS Fellow. He has worked on many aspects of numerical linear algebra and matrix computations, including eigenvalue problems, sparse matrix techniques, Schwarz preconditioning and domain decomposition, Krylov subspace methods, and asynchronous iterative methods. His PhD is from the Courant Institute, New York University (1983), which he attended after undergraduate studies in his native Buenos Aires. Up next: Thursday, April 16 – Elad Romanov (Yale University). - Liza Rebrova, Marc Aurèle Gilles, and Jorge Garza Vargas.

1 0

Emerging Frontiers in Computer Architecture and Systems Seminar: Wed. April 8, 4:30 PM, B205 EQuad
by Emily C. Lawrence 07 Apr '26

07 Apr '26

Design in Motion: Specialization as an Ongoing Architectural Journey Emerging Frontiers in Computer Architecture and Systems Seminar Date: Apr 8, 2026, 4:30 pm – 5:30 pm Location: B205 Engineering Quadrangle Speaker: Martha Kim, Columbia University Abstract For decades, computing advanced through Moore’s Law and Larus’ virtuous cycle, where hardware improvements automatically accelerated existing software, enabling higher abstraction and continuously fueling demand for faster chips. However, as emerging workloads like deep learning outstrip general-purpose CPU capabilities, the industry increasingly turns to specialized accelerators that sacrifice generality for performance, introducing fragmented programming models, scarce expertise requirements, and complex system integration challenges. As a result, while accelerators deliver significant speedups, they also create a less-virtuous cycle in which software brittleness, high development costs, and rapid workload evolution risk rendering even highly optimized hardware obsolete. There is a strong and understandable predisposition to design accelerators in an ad hoc, reactive manner, but this limits the architecture community’s ability to systematically shape long-term computing trends. Without a structured understanding of acceleration dynamics, it is difficult to anticipate end states, evaluate tradeoffs, or avoid wasted design effort. This talk will present a simple model of the acceleration process, grounded in practical experience, to enable more proactive and principled decisions that steer innovation toward sustainable and virtuous outcomes. Bio Martha Kim is an Associate Professor of Computer Science at Columbia University. Kim leads the ARCADE Lab where she conducts research in computer architecture, parallel programming, compilers, and low-power computing. Her work has explored low-cost chip manufacturing systems, reconfigurable communication networks, and fine-grained parallel application profiling techniques. Kim holds a PhD in Computer Science and Engineering from the University of Washington and a bachelors in Computer Science from Harvard University.

1 0

CS Colloquium Speaker: Luay Nakhleh, Mon, April 6 at 12:10pm
by Emily C. Lawrence 06 Apr '26

06 Apr '26

CS Colloquium Speaker: Luay Nakhleh , Rice University Date & Time: Monday, April 6 - 12:10pm Location: CS 105 Host: Ben Raphael Event page: https://www.cs.princeton.edu/events/towards-building-phylogenetic-network-l… Webinar registration: https://princeton.zoom.us/webinar/register/WN_37fnuL32R1OvNSUqK2wsOw Title: Towards Building the Phylogenetic Network of Life: Models, Methods, and Challenges Abstract: In Charles Darwin’s Origin of Species, the evolutionary history of species takes the shape of a tree. Ever since, trees have been a mainstay of evolutionary biology and, indeed, almost all of biology. However, genomic data is revealing a picture of evolution that is too complex to be captured by a single, simple tree. Horizontal gene transfer (HGT) is rampant in the bacterial world. Hybridization is estimated to be involved in the evolutionary history of over 25% of all plants. Gene flow has occurred between humans and Neanderthals and its extent in many groups of eukaryotic species, including our own order—the primates. In the presence of horizontal gene transfer, hybridization, or gene flow, the evolutionary history takes the shape of a rooted, directed, acyclic graph (rDAG), commonly referred to as ‘phylogenetic network.’ The main computational task in this field is inferring these phylogenetic networks (the rDAG structure itself as well as parameters associated with its nodes and edges) from genomic sequences that are available only for the leaves of the rDAG. Accurate reconstruction of these networks facilitates multiple biological tasks including, but not limited to, understanding the mode of speciation, adaptation to environmental changes (e.g., antibiotic resistance in bacteria), gene functions, and how traits evolve. The talk is divided into three parts. I will first describe models and optimization criteria of increasing complexity for phylogenetic networks. I will then describe methods for inferring phylogenetic networks under these models and criteria. Finally, I will discuss computational challenges that currently limit the scalability of phylogenetic network inference to large data sets. Bio: Luay Nakhleh is the William and Stephanie Sick Dean of Engineering and Professor of Computer Science at Rice University in Houston, Texas. He received a B.Sc. degree in Computer Science (1996) from the Technion (Israel), a master’s degree in Computer Science (1998) from Texas A&M University, and a Ph.D. degree in Computer Science (2004) from The University of Texas in Austin. He conducts research in the areas of bioinformatics and computational biology, focusing mainly on questions in evolutionary biology and genomics. Luay is a recipient of multiple awards including the Department of Energy CAREER award, the National Science Foundation CAREER award, the Sloan Fellowship, and the Guggenheim Fellowship. He is Fellow of the International Society for Computational Biology, the American Institute for Medical and Biological Engineering, and the American Association for the Advancement of Science.

1 1

Romina Mahinpei will present her MSE talk "AI Mediation for Discretionary Work: Reducing the Activation Energy of Feedback Provision" on Monday, April 6 , 2026 at 11am in CS 302.
by Gradinfo 02 Apr '26

02 Apr '26

Romina Mahinpei will present her MSE talk "AI Mediation for Discretionary Work: Reducing the Activation Energy of Feedback Provision" on Monday, April 6 , 2026 at 11am in CS 302. Thesis Committee: Manoel Horta Ribeiro (adviser), Andrés Monroy-Hernández (reader) Thesis Title: AI Mediation for Discretionary Work: Reducing the Activation Energy of Feedback Provision Thesis Abstract: AI systems are increasingly integrated into human workflows, often as mediators that shape human behavior and decision-making by generating intermediate artifacts. While prior work shows that AI mediation can improve task efficiency and accuracy, it remains unclear whether it can increase participation in discretionary but beneficial tasks that people often intend to perform but frequently skip. To explore this idea, we study one such task: feedback provision in a university course. We conduct a randomized field experiment in a 300-level machine learning course in which students' question-level submissions are assigned to either (1) a treatment condition where teaching assistants (TAs) receive AI-mediated feedback templates that they can choose to use or (2) a control condition where TAs do not receive any templates. Using behavioral logs, surveys, and interviews, we examine how AI mediation influences participation, interaction patterns, and downstream student experiences. We find that AI-mediated feedback templates significantly increase feedback provision and that they lower the small but persistent barriers required to initiate feedback, which we refer to as the task's activation energy, rather than reducing the total effort required to produce it. Importantly, this increased provision does not degrade downstream experience of students, suggesting that AI mediation can scale access to beneficial practices while maintaining satisfactory quality. Together, our results position AI mediation as a mechanism for increasing participation in discretionary work and surface important tradeoffs for the design of systems that position AI as mediators. All are welcome to attend.

1 0

Computational Math IDeAS seminar: Thursday at 2pm
by Emily C. Lawrence 31 Mar '26

31 Mar '26

COMPUTATIONAL MATH – IDeAS SEMINAR Recurring weekly series · Thursdays, 2:00 – 3:00 PM This week's talk: Speaker: Florian Schäfer (NYU) Title: Toward Information Geometric Mechanics Date: Thursday, April 2, 2026 Time: 2:00 PM – 3:00 PM Room: 224 Fine Hall Abstract: Shock waves in high-speed gas dynamics cause severe numerical difficulties for classical solvers and scientific machine learning. They are fundamentally a multiscale problem: While viscous effects ensure smoothness on microscopic scales, shocks manifest as macroscopic discontinuities. This talk begins with the observation that shock formation arises from the flow map reaching the boundary of the manifold of diffeomorphisms. We modify its geometry such that geodesics approach but never reach the boundary. The resulting information geometric regularization (IGR) has smooth solutions while avoiding the excessive dissipation of viscous regularizations, accelerating and simplifying the simulation of flows with shocks. We prove the existence of global strong IGR solutions in the unidimensional pressureless case and illustrate its practical utility on multidimensional examples with complex shock interactions. With S. Bryngelson and other collaborators, we use IGR to conduct the first compressible flow simulation exceeding a quadrillion degrees of freedom. The modified geometry of the diffeomorphism manifold is the information geometry of the mass density. The last part of the talk explains how this observation motivates information geometric mechanics that views the solutions of continuum mechanical PDEs as parameters of probability distributions originating from statistical physics. Replacing the Euclidean geometry of individual particles with the information geometry of statistical families promises performant numerical methods that preserve the positivity of densities and energies and readily integrate with scientific machine learning. About the speaker: Florian Schäfer is an assistant professor at the Courant Institute of Mathematical Sciences at New York University. Prior to joining Courant, he was an assistant professor at Georgia Tech and received his PhD in applied and computational mathematics at Caltech, working with Houman Owhadi. Before that, he received bachelor's and master's degrees in mathematics at the University of Bonn. His research addresses fundamental problems in computational science and engineering using insights from statistics and geometry. It was awarded a SIGGRAPH best technical paper award, a Gordon Bell Prize finalist designation, and a Sloan Research Fellowship in Mathematics. Up next: Thursday, April 9 – Daniel B. Szyld (Temple). There are a few slots to meet with the speakers, please let us know if you would like one (gilles(a)princeton.edu). - Liza Rebrova, Marc Aurèle Gilles, and Jorge Garza Vargas.

1 0

CS Colloquium Speaker: Stefano Soatto, Tues. Feb 24 at 3:20pm
by Emily C. Lawrence 31 Mar '26

31 Mar '26

CS Colloquium Speaker: Stefano Soatto from UCLA & Amazon Date & Time: Tuesday, February 24, 2026 - 3:20pm Location: CS 105 Host: Jia Deng Webinar registration: [ https://princeton.zoom.us/webinar/register/WN_geQ0rhfcQR-fwkdQGAo7Ug | https://princeton.zoom.us/webinar/register/WN_geQ0rhfcQR-fwkdQGAo7Ug ] Title: AI Agents as Universal Task Solvers: Time, Information, and Intelligence in LLMs viewed as Maximalistic Models of Computation Abstract: Scaling laws predict that AI agents will steadily improve and eventually exceed human performance across a wide range of tasks. Yet at the limit of these scaling laws lies a form of inference that involves no intelligence at all: with increasing compute and memory, a model can brute-force any verifiable task without learning anything from past experience. Universally optimal inference, pioneered by Solomonoff and Levin, requires no insight — only exhaustive search. This raises a basic question: if scaling alone does not foster intelligence, what does? And if performance on downstream tasks is insufficient to measure intelligence, what is? In this talk, I will point to the critical role of time in both analyzing and fostering the emergent reasoning behavior of AI agents. Building on insights that Solomonoff sketched in 1985 but that remained theoretical curiosities for decades, I will show that the value of learning is measured not by a reduction in uncertainty — the core of inductive learning and generalization — but by a reduction in the time needed to solve new tasks. A key result is that data can make a universal solver exponentially faster, with the speed-up tightly characterized by a single quantity: the algorithmic mutual information between past experience and the solution to unforeseen tasks. Connecting these ideas to modern AI requires rethinking what computation means for systems powered by large language models. Unlike minimalistic models of computation such as Turing Machines, LLMs are stochastic dynamical systems whose computational elements — context, weights, activations, chain-of-thought — do not resemble a "program" in the ordinary sense. I will show that LLMs are instead maximalistic models of computation: universal, like Turing Machines, but operating through entirely different and in many ways antithetical mechanisms. Programming such systems can be achieved through two-level control strategies — open-loop planning and closed-loop feedback — in abstract space, a framework we have recently released Strands Agents open-source library (www.strandsagents.com). Once time is properly accounted for, scaling laws reveal an inversion : beyond a critical point, increasing resources improve benchmark accuracy while diminishing conceptual depth— a savant regime in which models improve while learning less. I will discuss what this means for how we build, evaluate, and scale AI agents. Bio: Stefano Soatto is a Vice President at AWS Agentic AI, and a Professor of Computer Science at UCLA. He received his PhD in Control and Dynamical Systems from the California Institute of Technology, his D.Ing from the University of Padova, Italy and was a postdoctoral scholar at Harvard University. He is a Fellow of the ACM and of the IEEE.

1 6

Computational Math IDeAS seminar: Thursday, March 26 at 2pm
by Emily C. Lawrence 24 Mar '26

24 Mar '26

COMPUTATIONAL MATH – IDeAS SEMINAR Recurring weekly series · Thursdays, 2:00 – 3:00 PM This week's talk: Speaker: Leticia Mattos Da Silva (MIT) Title: Leveraging Latent Mathematical Structure in Geometry Processing Date: Thursday, March 26, 2026 Time: 2:00 PM – 3:00 PM Room: 224 Fine Hall Abstract: Myriad problems in geometry processing are readily formulated through continuous mathematics, including variational energies, stochastic dynamics, geometric constraints, and partial differential equations. Often, the challenge in solving these formulations in practice is not the absence of a principled theoretical formulation, but that standard computational approaches are limited by design constraints or technical barriers, resulting in instability or poor scalability. Common numerical techniques resort to fragile nonlinear solvers, repeated large optimizations, or forgo analytical or physical properties of the continuous problem. In this talk, I will present methods that tackle these limitations by recognizing and exploiting latent problem structure, including convexity, variational reformulations, and symmetry. After a prelude illustrating the value of this approach in two problems of interest in geometry---simulating elastodynamics and solving PDEs on surfaces---I will describe a new approach to resampling that leverages a time-symmetric variant of the Schrödinger bridge problem. In each of these works, we discuss how algorithmic design, when carefully guided by underlying mathematical structure, can yield numerical tools that offer practitioners greater control and flexibility. About the speaker: Leticia is a fifth-year Ph.D. student at MIT, where she works in the Geometric Data Processing Group under the supervision of Justin Solomon. Her research involves developing numerical algorithms to solve nonlinear PDEs and other dynamical problems that appear in geometry processing. She is the recipient of two MathWorks Fellowships, as well as a Google Fellowship. Previously, she received a B.S. in Mathematics from the University of California, Los Angeles, and in summers past, she has conducted research as part of internships at the Flatiron Institute, at Adobe, Inc. and at the Fields Institute. Up next: Thursday, April 2 – Florian Schaefer (NYU).

1 0