Dear all,
Tomorrow's AI seminar will be by Yoon Kim, who will be talking about large language models. This seminar will be remote, but please attend in person if possible.
Prasad
Location: KEC 1001
Zoom link: https://oregonstate.zoom.us/j/96491555190?pwd=azJHSXZ0TFQwTFFJdkZCWFhnTW04U…<https://nam04.safelinks.protection.outlook.com/?url=https%3A%2F%2Foregonsta…>
Large Language Models and Symbolic Structures
Yoon Kim
EECS, Massachusetts Institute of Technology
Abstract:
Over the past decade the field of NLP has shifted from a pipelined approach (wherein intermediate symbolic structures such as parse trees are explicitly predicted and utilized for downstream tasks) to an end-to-end approach wherein pretrained large language models (LLMs) are adapted to various downstream tasks via finetuning or prompting. What role (if any) can symbolic structures play in the era of LLMs? In the first part of the talk, we will see how latent symbolic structures in the form of hierarchical alignments can be used to guide LM-based neural machine translation systems to improve translation of low resource languages and even enable the use of new translation rules during inference. In the second part, we will see how expert-derived grammars can be used to control LLMs via prompting for tasks such as semantic parsing where the output structure must obey strict domain-specific constraints.
Speaker Biography:
Yoon Kim is an assistant professor at MIT EECS. He received his PhD from Harvard University, advised by Alexander Rush.
Dear all,
Our next AI seminar is scheduled to be on February 23rd , 2-3 PM. It will be followed by a 30-minute Q&A session with the graduate students.
Location: KEC 1001
Zoom link: https://oregonstate.zoom.us/j/96491555190?pwd=azJHSXZ0TFQwTFFJdkZCWFhnTW04U…<https://nam04.safelinks.protection.outlook.com/?url=https%3A%2F%2Foregonsta…>
Gradient-free bilevel programming with applications in machine learning
Alireza Aghasi
Assistant Professor
Electrical and Computer Engineering
Oregon State University
Abstract:
Our talk is mainly about a scalable and easy way of addressing bilevel optimization problems when we only have access to noisy evaluations of the objective functions, and absolutely no access to their gradient or Hessian. The talk starts with some motivating applications of bilevel programming in machine learning. We will then briefly overview zeroth-order methods for general optimization problems, and specifically focus on a more recent technique called Gaussian smoothing, which allows estimating the first and second order derivatives of an objective function solely through random queries to the function. Finally, we propose a computational algorithm to address general bilevel programs, without the need to access the Jacobian or Hessian of the underlying objectives. We present theoretical guarantees on the performance of the algorithm, and ultimately present iteration complexity (sample complexity) results. To the best of our knowledge, the proposed algorithm is the first result in the literature about a fully zeroth-order scheme to address general bilevel programs.
Speaker Bio:
Alireza Aghasi joined the School of Electrical Engineering and Computer Science at Oregon State University in Fall 2022. Between 2017 and 2022, he was an assistant professor in the Department of Data Science and Analytics at the school of business, Georgia State University. Prior to this position he was a research scientist with the Department of Mathematical Sciences, IBM T.J. Watson research center, Yorktown Heights. From 2015 to 2016 he was a postdoctoral associate with the computational imaging group at the Massachusetts Institute of Technology, and between 2012 and 2015 he served as a postdoctoral research scientist with the compressed sensing group at Georgia Tech. His research fundamentally focuses on optimization theory, probability theory and statistical analysis, with applications to various areas of data science, artificial intelligence, and modern signal processing.
Please watch this space for future AI Seminars :
https://engineering.oregonstate.edu/EECS/research/AI<https://nam04.safelinks.protection.outlook.com/?url=https%3A%2F%2Fengineeri…>
Rajesh Mangannavar,
Graduate Student
Oregon State University
----
AI Seminar Important Reminders:
-> For graduate students in the AI program, attendance is strongly encouraged
Dear all,
Our next AI seminar is scheduled to be on February 16th , 2-3 PM. It will be followed by a 30-minute Q&A session with the graduate students.
Location: KEC 1001
Zoom link: https://oregonstate.zoom.us/j/96491555190?pwd=azJHSXZ0TFQwTFFJdkZCWFhnTW04U…<https://nam04.safelinks.protection.outlook.com/?url=https%3A%2F%2Foregonsta…>
Towards AI-Powered 3D Reconstruction and Generation
Zexiang Xu
Research Scientist
Adobe Research
Abstract:
3D reconstruction and generation are essential for a variety of vision and graphics applications, including 3D content creation, augmented reality, virtual reality, gaming, and robotics. Our research aims to address the core challenges in these areas by re-inventing neural 3D representations and architectural designs, leveraging advanced AI technologies. This talk will cover the recent progress in the fields and showcase our latest works. We will begin by discussing the recent advancements in neural field representations, which have enabled compact and realistic 3D modeling, thereby facilitating both reconstruction and generation tasks. We will then investigate the efficient estimation of neural fields using deep networks in a feed-forward manner, enabling fast and generalizable 3D reconstruction. The talk will also demonstrate the strong inherent connection between 3D reconstruction and generation, exploring how reconstruction techniques can serve as a foundation for powerful generation models. Our recent research introduces transformer-based 3D large reconstruction models, achieving fast and realistic 3D reconstruction and generation across various tasks, ranging from multi-view reconstruction to single-view reconstruction and text-to-3D generation.
Speaker Bio:
Zexiang Xu is a research scientist at Adobe Research. His research interests lie at the intersection of computer vision, computer graphics, and machine learning, with the primary focus on enabling efficient and realistic 3D reconstruction and rendering. His work covers key areas in 3D vision and graphics, including 3D reconstruction, 3D generation, neural representations, view synthesis, relighting, rendering, and appearance acquisition. Previously, he earned his Ph.D. in 2020 from UC San Diego working with Prof. Ravi Ramamoorthi.
Please watch this space for future AI Seminars :
https://engineering.oregonstate.edu/EECS/research/AI<https://nam04.safelinks.protection.outlook.com/?url=https%3A%2F%2Fengineeri…>
Rajesh Mangannavar,
Graduate Student
Oregon State University
----
AI Seminar Important Reminders:
-> For graduate students in the AI program, attendance is strongly encouraged
Dear all,
Our next AI seminar is scheduled to be on February 9th , 2-3 PM. It will be followed by a 30-minute Q&A session with the graduate students.
Location: KEC 1001
Zoom link: https://oregonstate.zoom.us/j/96491555190?pwd=azJHSXZ0TFQwTFFJdkZCWFhnTW04U…<https://nam04.safelinks.protection.outlook.com/?url=https%3A%2F%2Foregonsta…>
Learning Universal Sequence Representations and Interpretable ML Models for Microbial Communities
Gail Rosen
Professor
Electrical and Computer Engineering
Drexel University
Abstract:
Microbes are found in every nook and cranny of the world and regulate the planetary carbon and oxygen cycles that are vital to life. Through high-throughput DNA/RNA sequencing, we can collect a vast amount of data about communities of microbes in diverse environments. In this talk, I will discuss development of large language models to tackle the first step in understanding this vast amount of data for taxonomy/gene identification, including antimicrobial resistance. Then, I will demonstrate our efforts to predict overall microbiome phenotype from metagenomic data and to explain important features, such as species composition, community hierarchy, functions, and mutations. Finally, I will address scalability of methods for sustainability.
Speaker Bio:
Gail Rosen received a B.S., M.S., and Ph.D. from the Georgia Institute of Technology. She is a recipient of an NSF CAREER award, a Drexel Faculty Career Development award, and Drexel Provost’s Fellowship. She serves on the editorial board of the Association for Microbiology’s mSystems and BMC Microbiome journals. She heads the Ecological and Evolutionary Signal-processing and Informatics (EESI) lab, organizes the Center for Biological Discovery from Big Data, and serves on the board and is a founding member of the University Research Computing Facility at Drexel. For the past 3 summers, she organized 2-week Drexel-Rowan-UChicago Biological Data Science summer workshops, which reached 8000+ participants around the world. In 2022-2023, she spent a year as a visiting scholar at the National Institute of Standards and Technology, benchmarking cancer-somatic variant calling methods. Her interests are in machine learning and evolution.
Please watch this space for future AI Seminars :
https://engineering.oregonstate.edu/EECS/research/AI<https://nam04.safelinks.protection.outlook.com/?url=https%3A%2F%2Fengineeri…>
Rajesh Mangannavar,
Graduate Student
Oregon State University
----
AI Seminar Important Reminders:
-> For graduate students in the AI program, attendance is strongly encouraged
Dear all,
Our next AI seminar is scheduled to be on February 2nd (Today), 2-3 PM. It will be followed by a 30-minute Q&A session with the graduate students.
Location: KEC 1001
Zoom link: https://oregonstate.zoom.us/j/96491555190?pwd=azJHSXZ0TFQwTFFJdkZCWFhnTW04U…<https://nam04.safelinks.protection.outlook.com/?url=https%3A%2F%2Foregonsta…>
Cross-validation for Geospatial Problems
Rebecca Hutchinson
Associate Professor
School of Electrical Engineering and Computer Science
Department of Fisheries, Wildlife, and Conservation Sciences
Oregon State University
Abstract:
Most machine learning practitioners have used cross-validation to assess the generalization performance of a model. However, typical cross-validation methods assume that the data are independent and identically distributed (iid). In geospatial problems, data points have geolocated coordinates (like latitude and longitude) as well as associated features, which are often spatially autocorrelated. Furthermore, geospatial problems may entail prediction to new regions, where the features are distributed differently. These properties violate the iid assumption. With a motivating ecological example of evaluating species distribution models, this talk will address theoretical and practical challenges that arise when conducting cross-validation for geospatial applications.
Speaker Bio:
Rebecca Hutchinson is an associate professor at Oregon State University, with a joint appointment across the School of Electrical Engineering and Computer Science and the Department of Fisheries, Wildlife, and Conservation Sciences. She is also affiliated with the Center for Quantitative Life Sciences and the Collaborative Robotics and Intelligent Systems (CoRIS) Institute. She became interested in interdisciplinary work in machine learning and quantitative ecology during her postdoctoral studies with the Institute for Computational Sustainability and as an NSF SEES Fellow, advised by Tom Dietterich and Matt Betts. Prior to that, she completed her Ph.D. at Carnegie Mellon University with Tom Mitchell. She is the recipient of an NSF CAREER Award.
Please watch this space for future AI Seminars :
https://engineering.oregonstate.edu/EECS/research/AI<https://nam04.safelinks.protection.outlook.com/?url=https%3A%2F%2Fengineeri…>
Rajesh Mangannavar,
Graduate Student
Oregon State University
----
AI Seminar Important Reminders:
-> For graduate students in the AI program, attendance is strongly encouraged
