Join us at our next event.

Save the Date for a MATRIX event hosted by our Augmenting Human Capabilities Thrust

Featuring a keynote from Microsoft’s Chief Scientific Officer Eric Horvitz, MD, PhD

This event is dedicated to exploring the latest innovations and applications of Artificial Intelligence in the field of human health.

This event brings together AI researchers, clinicians, industry leaders in high tech, biotech and drug discovery, science foundations, and policymakers to discuss how AI is transforming diagnostics, treatment, personalized medicine, and public health.

Join the 2nd AI Spring School

A three-day event, hosted by UTSA MATRIX, NSF NAIAD, ATHENA, and the UTSA Office of Research!

Dive into the latest advancements in AI and machine learning (ML) for edge systems. Discover how cutting-edge neuro-inspired and distributed (federated) algorithms are enhancing energy efficiency and robustness in AI.

Learn from top experts from Google, NSF, IBM, CMU, Duke, and more. Engage in hands-on activities and connect with the vibrant AI/ML community. Open to all students and researchers, this event is your opportunity to deepen your knowledge and shape the future of AI at the edge. Don’t miss out on this exciting journey!

Topic: Resource Allocation and Learning for Wireless Networks: Underwater and Radio Frequency Use Cases

Next in our spring seminar series, we welcome Dr. Urbashi Mitra, IEEE Signal Processing Society Distinguished Lecturer Professor, and IEEE Fellow at the University of Southern California.

You can see the presentation live at the UTSA Main Campus in the Student Union Mesquite room (2.01.24).

Those who cannot attend in-person are welcome to tune in virtually through Zoom: https://utsa.zoom.us/j/94807623288.

Topic: Learning Dynamics and Implicit Bias of Gradient Flow in Overparameterized Linear Models

Next in our spring seminar series, we welcome Dr. René Vidal, a Professor from the University of Pennsylvania.

Abstract:
Contrary to the common belief that overparameterization may hurt generalization and optimization, recent work suggests that overparameterization may bias the optimization algorithm towards solutions that generalize well — a phenomenon known as implicit regularization or implicit bias — and may also accelerate convergence — a phenomenon known as implicit acceleration. This talk will provide a detailed analysis of the dynamics of gradient flow in overparameterized two-layer linear models showing that convergence to equilibrium depends on the imbalance between input and output weights (which is fixed at initialization) and the margin of the initial solution. The talk will also provide an analysis of the implicit bias, showing that large hidden layer width, together with (properly scaled) random initialization, constrains the network parameters to converge to a solution which is close to the min-norm solution.

You can see the presentation live at the UTSA Main Campus in the Student Union Mesquite room (2.01.24).

Those who cannot attend in-person are welcome to tune in virtually through Zoom: https://utsa.zoom.us/j/94807623288.

Topic: Revolutionizing Public Safety:
6G Wireless Unleashes Precision Positioning and Navigation

Next in our spring seminar series, we welcome Dr. Eirini Tsiropoulou, an Associate Professor from the University of New Mexico.

Abstract:
The imperative need for accurate Position, Navigation, and Timing (PNT) systems, vital to critical infrastructure and advanced location-based services, has spurred the development of alternative technologies. This talk addresses the potential degradation of Global Navigation Satellite Systems (GNSS) under varying conditions by proposing an innovative PNT solution. Leveraging Reconfigurable Intelligent Surfaces (RISs), we present a symbiotic game-theoretic framework that enhances PNT accuracy and efficiency. The heart of our approach lies in the orchestration and configuration of RISs. Game Theory and Reinforcement Learning (RL) are harnessed to address this challenge. Through a novel Satisfaction Game, targets autonomously determine the optimal count of contributing RISs for PNT services. The specific RISs ensemble is selected via an advanced RL algorithm. Further optimization involves phase shifts of the RISs, maximizing signal strength at targets through reflected signals. The proposed approach is grounded in a mutualistic symbiotic concept drawn from biological systems. Collaborator nodes, supported by RISs and anchor nodes with known coordinates, forge a symbiotic relationship with targets of unknown positions. Minimizing positioning and timing errors becomes a non-cooperative game among these entities, showcasing the existence of a Nash Equilibrium through the potential game principles. Performance evaluation is conducted through modeling and simulation, unveiling trade-offs between game-theoretic and RL approaches. The proposed methodology presents a promising avenue for robust PNT solutions, bridging accuracy and reliability gaps in GNSS and catering to critical sectors dependent on precise positioning, navigation, and timing.

You can see the presentation live at the UTSA Main Campus in the Student Union Mesquite room (2.01.24).

Those who cannot attend in-person are welcome to tune in virtually through Zoom: https://utsa.zoom.us/j/94807623288.

Topic: Anomaly Detection from Hyperspectral Remote Sensing Imagery

Next in our spring seminar series, we welcome Dr. Jenny Du, a professor from Mississippi State University.

Abstract:
Anomaly detection from remote sensing images is to detect pixels whose spectral signatures are different from their background. Anomalies are often man-made targets. With such target signatures being unknown, anomaly detection has many important applications, such as crop stress surveying, water quality monitoring, and law enforcement related uses, where prior information of targets is often unavailable. The key to success is accurate background modeling. Anomaly detection from remote sensing images is challenging, because spatial coverage is very large and background is highly heterogeneous. Hyperspectral imaging with very high spectral resolution offers the advantage in background-anomaly separation. In this talk, the major machine learning techniques in anomaly detection from hyperspectral remote sensing images will be introduced, and the recent advances using transfer learning and deep learning for anomaly detection will be presented. In particular, autoencoder, generative adversarial network, and adversarial autoencoder are deployed and compared with traditional approaches. Practical issues and future development trends will also be discussed.

You can see the presentation live at the UTSA Main Campus in the Student Union Mesquite room (2.01.24).

Those who cannot attend in-person are welcome to tune in virtually through Zoom: https://utsa.zoom.us/j/94807623288.