April 2026 - talks - lists.cs.princeton.edu

Henri Schmidt will present his FPO "Inferring cellular phylogenies using optimization” on Thursday, May 7, 2026 at 10:00 AM in CS 402.
by CS Grad Department 05 May '26

05 May '26

Henri Schmidt will present his FPO "Inferring cellular phylogenies using optimization” on Thursday, May 7, 2026 at 10:00 AM in CS 402. The members of Henri’s committee are as follows: Examiners: Ben Raphael (Adviser), Yuri Pritykin, Pravesh Kothari Readers: Olga Troyanskaya, Mohammed El-Kebir (UIUC) A copy of his thesis is available upon request. Please email gradinfo(a)cs.princeton.edu if you would like a copy of the thesis. Everyone is invited to attend the talk. Abstract follows below: Reconstructing the phylogenetic tree of cells is a central challenge in developmental and cancer biology. Advances in sequencing and experimental design have brought this goal within reach, producing high-resolution single-cell measurements that enable phylogenetic reconstruction. However, phylogenetic reconstruction from such measurements is difficult and requires phylogenetic models that balance biological plausibility and computational tractability. We present a collection of research whose goal is, broadly, to design tractable phylogenetic models along with efficient algorithms to reconstruct phylogenetic trees. In particular, we adapt modern tools and techniques from combinatorial and convex optimization to the phylogenetic setting, leading to highly efficient and accurate algorithms for large-scale phylogenetic inference.

2 1

CS Colloquium Speaker: Hamsa Bastani, Tues May 5 at 12:10pm
by Emily C. Lawrence 05 May '26

05 May '26

CS Colloquium Speaker: Hamsa Bastani, University of Pennsylvania Date & Time: Tuesday, May 5 - 12:10pm Location: CS 105 Event page: https://www.cs.princeton.edu/events/unpacking-unintended-consequences-ai-ed… Webinar registration: https://princeton.zoom.us/webinar/register/WN_lKB_0wIFQnWF3YfnIgL7tw Title: Unpacking the Unintended Consequences of AI in Education Abstract: The rapid integration of AI into educational settings presents challenges and opportunities—this talk will discuss findings from three large-scale field studies investigating the impact of AI on student learning. First, we found that unfettered access to ChatGPT negatively impacted short-term student learning outcomes in high school mathematics. Second, to understand longer-term effects, we examined learning in chess academies. Contrary to the popular strategy of promoting student agency, our findings show that self-regulated learning (where students decide when to request AI help) can substantially harm learning by undermining productive struggle and motivation. Third, we found that carefully designed AI can be a potent way to boost productive struggle—we leverage student-chatbot interactions (which encode rich information about the student’s knowledge state) to personalize practice problems, significantly improving learning over a 5-month Python certification course. Taken together, these studies suggest that while providing students with unguided AI tools can be detrimental, targeted AI tools that are designed to critically engage students can be beneficial. Bio: Hamsa Bastani is an Associate Professor of Operations, Information, and Decisions at the Wharton School, University of Pennsylvania. Her research focuses on developing novel machine learning algorithms for social impact, including in healthcare and education. She has worked with national governments to deploy algorithms at the country scale for targeted border COVID-19 screening and essential medicine access, and co-led one of the first large field studies of generative AI tutors in high schools. Her research has been published in leading outlets including Nature, Management Science, Operations Research, and PNAS, and has garnered numerous recognitions, including the Wagner Prize for Excellence in Operations Research, the INFORMS Pierskalla Award for best healthcare paper, and the George Nicholson Prize.

1 2

Nicholas Richardson will present his FPO "Gradient-based Optimization of Geometry with Application to Optical Computing" on Monday, May 4, 2026 at 11:00 AM in CS 401.
by CS Grad Department 04 May '26

04 May '26

Nicholas Richardson will present his FPO "Gradient-based Optimization of Geometry with Application to Optical Computing" on Monday, May 4, 2026 at 11:00 AM in CS 401. The members of Nicholas’s committee are as follows: Examiners: Ryan Adams (Adviser) Margaret Martonosi, Benjamin Eysenbach, Readers: Szymon Rusinkiewicz, Elif Ertekin (UIUC) A copy of his thesis is available upon request. Please email gradinfo(a)cs.princeton.edu if you would like a copy of the thesis. Everyone is invited to attend the talk. Abstract follows below: The automated design of engineered systems — reducing human expertise to a specification and delegating search to computation — is one of the central ambitions of computational science and optimization. In this thesis, I present work at the intersection of optimization techniques and their application to inverse design of computational hardware. First, I present a novel method, Fiber Monte Carlo, which enables gradient-based optimization in optimization problems containing parametric discontinuities; this is a class of problems that arise in a wide variety of applications, including graphics, topology optimization, computational geometry, and physical simulation. Next, I turn to the computational design of electromagnetic systems. I first detail the development of a custom differentiable electromagnetic field solver, JaxEM, with native support for inverse design applications. The construction of this simulator enables straightforward gradient-based design of structured hardware. Finally, I apply Fiber Monte Carlo and JaxEM to the inverse design of specialized hardware for an optical classifier. This chapter contains a standalone contribution by way of a novel method to achieve nonlinear computation using only linear optics. The hardware is designed in silico using Fiber Monte Carlo based topology optimization, and we use gradient methods by differentiating through simulations performed with JaxEM. The resulting design was fabricated as a silicon photonic chip, providing an end-to-end demonstration of gradient-based automated hardware design in physical silicon.

2 1

Runpeng Luo will present his General Exam "Inferring Allele-Specific Copy Number Aberrations across Different Sequencing Technologies" on Tuesday, May 5, 2026 at 3:15 PM in CS 302.
by CS Grad Department 30 Apr '26

30 Apr '26

Runpeng Luo will present his General Exam "Inferring Allele-Specific Copy Number Aberrations across Different Sequencing Technologies" on Tuesday, May 5, 2026 at 3:15 PM in CS 302. Committee Members: Ben Raphael (advisor), Yuri Pritykin, Sneha Goenka Abstract: Copy-number aberrations (CNAs) are somatic mutations that alter the number of copies of a genomic segment on one or both parental chromosomes. Inferring CNAs from DNA or RNA sequencing of individual cells is important for quantifying tumor heterogeneity, reconstructing tumor evolution, and identifying the spatial organization of clones, or groups of cells with the same set of CNAs. Traditionally, CNAs were inferred from DNA sequencing of bulk tumor samples containing DNA from thousands to millions of cells, which limits the ability to identify CNAs that are present in only a subpopulation of cells in a tumor. More recently, researchers have inferred CNAs from single-cell RNA sequencing and single-cell ATAC sequencing data from tumor samples. However, neither of these technologies directly measure DNA copy number: single-cell RNA sequencing measures gene expression, and single-cell ATAC sequencing profiles open-chromatin accessibility. Thus, the inference of CNAs from these technologies is complicated by transcriptional or epigenetic variation as well as the sparse and noisy measurement in individual cells. To address these challenges, we studied the setting in which both single-cell sequencing data and bulk DNA sequencing data are available from the same patient. We formulated the single-cell clone inference problem as follows: Given the allele-specific CNAs (copy numbers specific to each parental chromosome) of multiple clones and haplotype phasing (which alleles belong to each parental chromosome) both inferred from bulk DNA sequencing data, infer the clone label of each cell profiled by single-cell sequencing data. Here I describe two advances. First, we developed a probabilistic inference method Copy-typing to solve the single-cell clone inference problem. Copy-typing utilizes allele-specific CNAs and haplotype phasing to aggregate single-cell sequencing data across multiple chromosomes, and infer the clone labels via maximum a posteriori inference. Second, we made various improvements on existing allele-specific CNA inference methods using bulk DNA sequencing data to infer more accurate haplotype phasing and allele-specific CNAs. We validated Copy-typing on multiple real datasets. Our result shows strong agreement with orthogonal validation data. Using spatial transcriptomics data, Copy-typing identified spatially contiguous tumor clones which were consistent with pathology images and published analyses. Reading List: https://docs.google.com/document/d/1Okg8ke948Wj1oqMnOufyObxVqZTcSboYYzRVZ_t… Everyone is invited to attend the talk, and those faculty wishing to remain for the oral exam following are welcome to do so.

1 0

Cyrus Vachha will present his General Exam "Enabling Transitions Across Physical Reality and Virtual Representations of Spaces" on Friday, May 1, 2026 at 10:00 AM in CS 301 and via zoom.
by CS Grad Department 30 Apr '26

30 Apr '26

Cyrus Vachha will present his General Exam "Enabling Transitions Across Physical Reality and Virtual Representations of Spaces" on Friday, May 1, 2026 at 10:00 AM in CS 301 and via zoom. Zoom link: https://princeton.zoom.us/j/98265539965?pwd=SNOZdUkcuap7MnyDv6zS2ahotL8KFc.1 Committee Members: Parastoo Abtahi (advisor), Felix Heide, Szymon Rusinkiewicz Abstract: Recent advances in Gaussian splatting, 3D foundation models, and video-passthrough AR/VR offer unprecedented fidelity for capturing and blending physical and virtual worlds. Yet, a fundamental limitation remains: the inability to act upon these layered representations of reality with physical reality itself. While we can visualize environments with photorealistic detail, current systems lack the architectural support to make these captures truly interoperable with agents or physical actions. To bridge this disconnect, I reconceptualize XR as a multi-layered reality, where the challenge lies in facilitating bidirectional transitions: allowing interactions to propagate from physical reality to virtual representations and back again. I first present SplatXR, a Unity toolkit bridging the gap from physical reality to visual virtuality. By providing building-block abstractions, SplatXR enables developers to build end-user applications leveraging splats and facilitate transitions across layers, such as reconstructing real objects into splats or converting static captures into interactive mesh-enhanced splats. This effectively transforms passive captures into a programmable reality. I discuss future research directions extending this framework toward controllable, situated world models for composing generative interactable elements implicitly in XR. Building on this layer traversal framework, I present preliminary concepts for a robotic tele-operation system demonstrating the inverse transition: where virtual interactions propagate back into physical reality. By selectively rendering 3D portals of remote scenes and increasing visual and instruction fidelity where robotic uncertainty is highest, the system would allow users to resolve linguistic ambiguities through spatial interaction. This bidirectional interoperability creates a natural interface for communicating with reality, illustrating a future where the digital-physical boundary becomes a traversable interface. Reading List: https://docs.google.com/document/d/1XrDTOyRBRv-oP8zvbjmT-efdVy6iBvW2xWRoAnK… Everyone is invited to attend the talk, and those faculty wishing to remain for the oral exam following are welcome to do so.

1 0

Rishwanth Raghu will present his General Exam "Guiding protein diffusion models with cryo-EM imaging data" on Tuesday, May 5, 2026 at 10:00 AM in CS 302 and via zoom.
by CS Grad Department 30 Apr '26

30 Apr '26

Rishwanth Raghu will present his General Exam "Guiding protein diffusion models with cryo-EM imaging data" on Tuesday, May 5, 2026 at 10:00 AM in CS 302 and via zoom. Zoom link: https://princeton.zoom.us/j/97312957225?pwd=3jz5x5tUukbWiHpmlC346C8GZHdYPb.1 Committee Members: Ellen Zhong (advisor), Ben Raphael, Mona Singh Abstract: Protein structure prediction models can accurately infer 3D molecular coordinates from amino acid sequence alone, but often struggle to capture the multiple structural states adopted by flexible biomolecules. In parallel, cryo-electron microscopy (cryo-EM) has emerged as a powerful tool for imaging near-native structural variability, but is challenged by arduous pipelines to go from raw experimental data to atomic structure. We present our work bridging the gap between these modalities, combining heterogeneous cryo-EM data with the rich biophysical priors learned by protein structure prediction models. Our method guides the diffusion sampling trajectory of a pretrained prediction model using constraints derived from cryo-EM density maps, driving samples towards conformational states consistent with experimental data. We demonstrate that this flexible inference-time approach allows us to build atomic structures into diverse cryo-EM maps across dynamic biomolecular systems including transporters and antibodies. Finally, we share ongoing work in scaling these approaches to model full ensembles of atomic structures from continuous distributions of density maps. Reading List: https://docs.google.com/document/d/1V1GfmvXZJwdSlDTkTSouDERQh7gRSW9yE5vL6gJ… Everyone is invited to attend the talk, and those faculty wishing to remain for the oral exam following are welcome to do so.

1 0

Robotics Seminar TOMORROW at 11am
by Emily C. Lawrence 30 Apr '26

30 Apr '26

Hi everyone, We're excited to be hosting Professor Tom Dietterich from OSU this Friday, May 1 . The seminar will be held in CS 105 from 11 am - 12 pm . The seminar will be followed by lunch in EQuad F-wing. Title: Can We Make Machine Learning Safe for Safety-Critical Systems? Abstract : The impressive new capabilities of systems created using deep learning are encouraging engineers to apply these techniques in safety-critical applications such as medicine, aeronautics, and self-driving cars. This talk will discuss the ways that machine learning methodologies are changing to operate in safety-critical systems. These changes include (a) building high-fidelity simulators for the domain, (b) adversarial collection of training data to ensure coverage of the so-called Operational Design Domain (ODD) and, specifically, the hazardous regions within the ODD, (c) methods for verifying that the fitted models generalize well, and (d) methods for estimating the probability of harms in normal operation. There are many research challenges to achieving these. But we must do more, because traditional safety engineering only addresses the known hazards. We must design our systems to detect novel hazards as well. We adopt Leveson’s view of safety as an ongoing hierarchical control problem in which controls are put in place to stabilize the system against disturbances. Disturbances include novel hazards but also management changes such as budget cuts, staff turnover, novel regulations, and so on. Traditionally, it has been the human operators and managers who have provided these stabilizing controls. Are there ways that AI methods, such as novelty detection, near-miss detection, diagnosis and repair, can be applied to help the human organization manage these disturbances and maintain system safety? Bio : Dr. Dietterich is University Distinguished Professor Emeritus in the School of Electrical Engineering and Computer Science at Oregon State University. Dietterich is one of the pioneers of the field of Machine Learning and has authored more than 220 refereed publications and two books. His current research topics include robust artificial intelligence, robust human-AI systems, and applications in sustainability. Dietterich is the 2025 recipient of the Feigenbaum prize for applied AI and the 2024 recipient of the IJCAI Award for Research Excellence. Dietterich is also the recipient of the 2022 AAAI Distinguished Service Award and the 2020 ACML Distinguished Contribution Award, both recognizing his many years of service to the research community. He is a former President of the Association for the Advancement of Artificial Intelligence and the founding president of the International Machine Learning Society. Other major roles include Executive Editor of the journal Machine Learning, co-founder of the Journal for Machine Learning Research, and program chair of AAAI 1990 and NIPS 2000. He currently chairs the Computer Science Section of [ http://arxiv.org/ | arXiv.org ] .

1 0

Xinhao Liu will present his FPO "Computational Tools for Alignment and Integration of Spatial Multimodality Data" on Wednesday, May 6, 2026 at 1:30p in COS 402.
by CS Grad Department 29 Apr '26

29 Apr '26

Xinhao Liu will present his FPO "Computational Tools for Alignment and Integration of Spatial Multimodality Data" on Wednesday, May 6, 2026 at 1:30p in COS 402. The members of Xinhao’s committee are as follows: Examiners: Ben Raphael (Adviser), Yuri Pritykin, Ellen Zhong Readers: Olga Troyanskaya, Li Ding (Washington University in St. Louis) A copy of his thesis is available upon request. Please email gradinfo(a)cs.princeton.edu if you would like a copy of the thesis. Everyone is invited to attend the talk. Abstract follows below: Recent advances in spatial sequencing technologies, such as spatial transcriptomics, proteomics, and epigenomics, enable the in situ measurement of molecular features across intact tissue sections, providing new insights into the spatial organization of biological processes. As datasets grow in complexity, spanning multiple tissue slices, developmental timepoints, and molecular modalities, there is an increasing need for computational methods that can align and integrate spatial omics data into coherent, high-resolution representations of tissues and organs. Accurate alignment across slices is essential not only for reconstructing tissue architecture but also for comparing molecular programs across spatial scales, developmental stages, and experimental modalities. This thesis presents four computational methods for aligning and integrating spatial omics data across developmental timepoints and experimental modalities. First, we introduce PASTE2, a method for alignment and three-dimensional (3D) reconstruction of multislice spatial transcriptomics data that allows slices to overlap only partially. Next, we present DeST-OT, which aligns spatial transcriptomics slices collected at different developmental timepoints to capture tissue dynamics over time. We then introduce MOSAICField, a framework for aligning spatial slices across arbitrary combinations of experimental modalities, such as spatial transcriptomics, spatial proteomics, and histology images. Finally, we introduce OTVI, a machine learning framework that integrates multiple spatial modalities from related tissue slices into a common coordinate framework while learning biologically meaningful representations for each spatial location.

1 0

Zijian Qin will present his General Exam "Avicenna: Fail-Slow Fault-Tolerant Replicated State Machines Using Counterfactual Evaluation" on Weds 5/6/2026 at 9:30am at 194 Nassau St, R. 213.
by Gradinfo 29 Apr '26

29 Apr '26

Zijian Qin will present his General Exam "Avicenna: Fail-Slow Fault-Tolerant Replicated State Machines Using Counterfactual Evaluation" on Weds 5/6/2026 at 9:30am at 194 Nassau St, R. 213. Committee: Wyatt Lloyd (adviser), Jialin Ding, Mae Milano Abstract: Geo-distributed replicated state machines (RSMs) are at the heart of many production distributed systems, offering linearizability and fault tolerance via consensus protocols. Most existing protocols target crash fault tolerance, however, and are vulnerable to fail-slow faults, where a single slow replica can significantly degrade system latency. Existing protocols that tolerate fail-slow faults do so with much higher normal-case latency in geo-distributed settings. Avicenna is the first consensus protocol for geo-distributed RSMs that maintains low normal-case latency while tolerating a single fail-slow replica. Avicenna uses a single leader to order commands, naturally tolerating a fail-slow follower. To tolerate a fail-slow leader, Avicenna compares the current latency with the counterfactual latency clients would experience if a different replica, the shadow leader, were the leader. When that comparison indicates the current leader might be slow, Avicenna quickly promotes the shadow leader with a fast leader rotation protocol. Our evaluation shows Avicenna has the same normal-case latency as Multi-Paxos while tolerating fail-slow faults. Reading List: [ https://docs.google.com/document/d/1tSJQW1Z5HbBKJgwAZj09oknyg1-nWeyidS04n4-… | https://docs.google.com/document/d/1tSJQW1Z5HbBKJgwAZj09oknyg1-nWeyidS04n4-… ]

1 0

Raj Ghugare will present his General Exam "Normalizing Flows are Capable Models for RL" on Thursday, April 30, 2026 at 3:00 PM in CS 105.
by CS Grad Department 29 Apr '26

29 Apr '26

Raj Ghugare will present his General Exam "Normalizing Flows are Capable Models for RL" on Thursday, April 30, 2026 at 3:00 PM in CS 105. Committee Members: Benjamin Eysenbach (advisor), Karthik Narasimhan, Brenden Lake Abstract: Modern reinforcement learning (RL) algorithms have found success by using powerful probabilistic models, such as transformers, energy-based models, and diffusion/flow-based models. To this end, RL researchers often choose to pay the price of accommodating these models into their algorithms – diffusion models are expressive, but are computationally intensive due to their reliance on solving differential equations, while autoregressive transformer models are scalable but typically require learning discrete representations. Normalizing flows (NFs), by contrast, seem to provide an appealing alternative, as they enable likelihoods and sampling without solving differential equations or autoregressive architectures. However, their potential in RL has received limited attention, partly due to the prevailing belief that normalizing flows lack sufficient expressivity. We show that this is not the case. Building on recent work in NFs, we propose a single NF architecture which integrates seamlessly into RL algorithms, serving as a policy, Q-function, and occupancy measure. Our approach leads to much simpler algorithms, which often achieve higher performance in imitation learning, offline, goal conditioned RL and unsupervised RL. Reading List: https://docs.google.com/document/d/1cMCRPwSPhpgrE_Qnku0INV3z7dVFFrJeThvF0NZ… Everyone is invited to attend the talk, and those faculty wishing to remain for the oral exam following are welcome to do so.

2 1