Seminars:

Monday 20th Feb
"Statistically Efficient Offline Reinforcement Learning" (Masatoshi Uehara) | 10AM

"Towards the Statistically Principled Design of ML Algorithms" (Frederic Koehler) | 12PM

Wednesday 22nd Feb

"Scalable Natural Language Processing for Transforming Medicine" (Monica Agrawal) | 10AM

"On the Complexity of Quantum Many Body Systems" (Chinmay Nirkhe)  |12PM

Events:

Call to Join the MS Planning Committee

Tech Talk with Noah Levine | Feb 20th 4PM

Kiki & Conversation with Marquis Bey | Feb 22nd 3PM
VentureCat Info Session | Feb 22nd 6:15PM

RAISO Discussion | Feb 22nd 7:00PM

Evening with Sarah Lim | Feb 23rd 5PM

Bagel Friday | Feb 24th 9:30AM
Wild Hacks Registration | Open Till Feb 28th 11:59

PhD Visit Day | March 10th

News

CS Seminars

Monday / CS Seminar

February 20th / 10:00 AM

Mudd 3514

Title: Statistically Efficient Offline Reinforcement Learning

Speaker: Masatoshi Uehara
 

Abstract:

Despite the remarkable achievements of reinforcement learning (RL) in the realm of gaming, such as AlphaGo and OpenAI Five, its implementation in scientific domains, such as economics and medicine, remains limited. This limitation is partly due to the costly and potentially hazardous nature of conducting experiments that involve human interaction. To address this challenge, statistically efficient offline RL, which enables sequential decision-making in a sample-efficient manner using offline data, is crucial. During this presentation, I will elaborate on my research in this area, with a primary focus on our "double minimax RL framework." This framework satisfies several desiderata such as (1) it can integrate any rich function approximation such as deep neural networks, (2) it is statistically efficient, and (3) it enables us to carry out statistical inference. For the remainder of the time, I will delve into the realm of model-based offline RL with general function approximation, introducing a novel algorithm referred to as constrained pessimistic policy optimization (CPPO). This algorithm has been developed to tackle the most challenging issue in offline RL, known as "distributional shift," which occurs in scenarios where the coverage of offline data is insufficient. I will demonstrate our CPPO algorithm is able to learn high-quality policies even if the coverage of offline data is not sufficient.
 

Biography:
Masatoshi Uehara is a third-year Ph.D. student at Cornell CS, advised by Nathan Kallus. He previously received a bachelor’s degree in applied mathematics and computer science from the University of Tokyo and a master’s of science in statistics at Harvard University. His research is at the intersection of reinforcement learning, causal ML, and Econ ML. He has won two scholarships, awarded to the most outstanding students from Japan, during his Ph.D. program. His works have been selected as spotlight/oral papers (2–5%) in top machine learning conferences such as ICML, NeurIPS, and ICLR.

//

Monday / CS Seminar

February 20th / 12:00 PM

Mudd 3514

Title: Towards the Statistically Principled Design of ML Algorithms

Speaker: Frederic Koehler

Abstract:

What are the optimal algorithms for learning from data? Have we found them already, or are better ones out there to be discovered? Making these questions precise, and answering them, requires taking on the mathematically deep interplay between statistical and computational constraints. It also requires reconciling our theoretical toolbox with surprising new phenomena arising from practice, which seem to violate conventional rules of thumb regarding algorithm and model design. I will discuss progress along these lines: in terms of designing new algorithms for basic learning problems, controlling generalization in large statistical models, and understanding key statistical questions for generative modeling.

Biography:

Frederic is currently a Rajeev Motwani Postdoctoral Fellow in the Department of Computer Science at Stanford University. He was previously a research fellow at the Simons Institute, and before that received his PHD in "Mathematics and Statistics" from the Massachusetts Institute of Technology, where he was coadvised by Ankur Moitra and Elchanan Mossel.

//

Wednesday / CS Seminar
February 22nd / 10:00 AM
Mudd 3514

Title: Scalable Natural Language Processing for Transforming Medicine
Speaker: Monica Agrawal

Abstract:
The data in electronic health records (EHRs) have immense potential to transform medicine both at the point-of-care and through retrospective research. However, structured data alone can only tell a fraction of patients' clinical narratives, as many clinically important variables are trapped within clinical notes. Automated extraction is difficult since clinical notes are written in their own jargon-heavy dialect, patient histories can contain hundreds of notes, and there is often minimal labeled data available. In this talk, I will discuss scalable natural language processing (NLP) solutions to overcome these technical challenges in clinical information extraction. These include the development of label-efficient modeling methodology and novel techniques for leveraging large language models. I will also describe a new paradigm for EHR documentation that incentivizes the creation of high-quality data at the point-of-care. I will end by discussing opportunities for NLP to impact a variety of healthcare workflows.

Biography:
Monica Agrawal recently completed her PhD in Computer Science at MIT CSAIL, advised by Professor David Sontag in the Clinical Machine Learning Group. Her research has been published at venues in machine learning, natural language processing, computational health, and human-computer interaction. She has been the recipient of a Takeda Fellowship, a Tau Beta Pi Fellowship, and an MIT EECS Edgerton Fellowship. Previously, she graduated from Stanford with a BS and MS in Computer Science.

//

Wednesday / CS Seminar

February 22th / 12:00 PM
Mudd 3514

Title: On the Complexity of Quantum Many Body Systems

Speaker: Chinmay Nirkhe

Abstract:
The ground-state of a quantum system of n particles is the eigenvector of minimal eigenvalue of a 2^n sized matrix called the Hamiltonian. The ground-state is the quantum generalization of the solution to a classical Constraint Satisfaction Problem. In this talk, I will describe a set of works in the intersection of computation complexity theory and quantum computation that show that ground-states of quantum systems likely lack succinct classical classical representation. I will discuss the implications for robust entanglement at room temperature and the quantum PCP conjecture. En route, I will discuss the resolution of the No Low-energy Trivial States (NLTS) Problem and oracle separations between QMA and QCMA.

Biography:
Chinmay Nirkhe is a research staff scientist with IBM Quantum at the MIT-IBM Watson AI Lab, primarily focusing on the intersection of complexity theory and quantum computing. Some of his research interests include error correction, hardness of approximation, and demonstrations of quantum/classical separations. His favorite open questions are the quantum PCP conjecture and whether QCMA equals QMA. He received his Ph.D. in Computer Science from UC Berkeley and his B.S. in Mathematics and Computer Science from Caltech.

CS Department Events

Other Events

News

Northwestern Department of Computer Science

Mudd Hall, 2233 Tech Drive, Third Floor, Evanston, Illinois, 60208

Unsubscribe