Fall Seminar Series 2023 – Dr. Julia Lindberg

Estimating Gaussian mixtures using sparse polynomial moment systems

Dr. Julia Lindberg

The University of Texas Austin

9/29/2023

11AM – 12PM CST

Location: SAN PEDRO 1 YOTTA ROOM 430

506 Dolorosa St, San Antonio, TX 78204

Zoom: https://utsa.zoom.us/j/91049981814

Abstract: The method of moments is a statistical technique for density estimation that solves a system of moment equations to estimate the parameters of an unknown distribution. A fundamental question critical to understanding identifiability asks how many moment equations are needed to get finitely many solutions and how many solutions there are. We answer this question for classes of Gaussian mixture models using the tools of polyhedral geometry. Using these results, we present a homotopy method to perform parameter recovery, and therefore density estimation, for high dimensional Gaussian mixture models. The number of paths tracked in our method scales linearly in the dimension.

Fall Seminar Series 2023 *Distinguished Speaker* Dr. Yiran Chen of Duke University

Unleashing the Potential of Edge Computing with AI Hardware and Software Innovations

Dr. Yiran Chen

Professor of Electrical and Computer Engineering

Director of the NSF AI Institute for Edge Computing Leveraging the Next-generation Networks (Athena),

Director of NSF Industry-University Cooperative Research Center (IUCRC) for Alternative Sustainable and Intelligent Computing (ASIC)

Co-director of Duke Center for Computational Evolutionary Intelligence (DCEI)

Duke University

10/13/2023

11AM – 12PM CST

Location: UTSA Main Campus, Student Union Denman Room 2.01.28 (Second Floor)

Zoom: https://utsa.zoom.us/j/91049981814

Abstract: Edge computing has emerged as a powerful paradigm, shifting data processing and analysis closer to its origin points, resulting in real-time insights, reduced latency, and improved privacy. AI hardware and software innovations further unleash the potential of edge computing by enhancing the intelligence, efficiency, and elasticity of the systems. In this talk, we will discuss the customization of AI hardware and the optimization of AI models, considering the computing resource constraints and communication bandwidth limitations of edge devices, the security and privacy requirements of the systems, and the heterogeneity of data and devices.

Fall Seminar Series 2023 *In-Person Event* – Dr. Christopher Kanan

Efficient Continual Deep Learning

Dr. Christopher Kanan

University of Rochester

10/27/2023

11AM – 12PM CST

Location: SAN PEDRO 1 YOTTA ROOM 430

506 Dolorosa St, San Antonio, TX 78204

Zoom: https://utsa.zoom.us/j/91049981814

Continual learning methods enable progressive knowledge accumulation in deep neural networks, and it has the potential to make training much more efficient and has many exciting applications such as on-device learning. Whereas conventional deep learning can suffer from catastrophic forgetting of previously acquired abilities when updating a network, over the past five years, continual learning has largely overcome this limitation. Despite this achievement, I argue the vast majority of continual learning algorithms have little practical value. In this talk, I show that methods are highly inefficient in terms of compute, memory, and storage. Some methods even require more computation than re-training from scratch as a dataset grows! For continual learning to have real-world applicability, the research community cannot ignore the resources used by these algorithms. Subsequently, I demonstrate that three new neuroscience-inspired algorithms created by my lab can vastly improve computational efficiency on large-scale vision and text understanding datasets. Our work has the potential to make continual learning more practical and widely adopted as well as advancing the goals of GreenAI.

Fall Seminar Series 2023 *Distinguished Speaker* Dr. David Pan of UT Austin

Closing the Virtuous Cycle of Photonics for AI and AI for Photonics

Dr. David Z. Pan

Department of Electrical and Computer Engineering

The University of Texas at Austin

11/3/2023

11AM – 12PM CST

Location: UTSA Main Campus, Student Union Hidalgo Room (2.214)

Zoom: https://utsa.zoom.us/j/91049981814

Photonics is a promising emerging technology that offers huge performance and efficiency benefits for AI/ML computing. There has been increasing interest in developing photonic AI accelerators for the next generation datacenters, smart sensing, intelligent edge, and automotive computing. However, realizing its full potential requires overcoming significant design challenges through a cross-layer co-design approach, where devices, circuits, architectures, and algorithms are optimized together. In this talk, I will show how photonics and AI interplay with each other. On the front of Photonics for AI, I will show some case studies of customized photonic device, circuit, architecture, and software co-design for efficient machine learning. On the front of AI for Photonics, I will show a novel machine learning framework for real-time photonic device simulations, which has the potential to speed up the customized photonic structure design through AI-enhanced optimizations. Closing the virtuous cycle of photonics for AI and AI for photonics shall hold great potential to significantly advance the state-of-the-art of each other.