AI & Quantum Computing Symposium

Artificial Intelligence (AI) and Quantum Computing are transformational technologies that will fundamentally change how we observe, interpret, and interact with the world of data that surrounds us. This symposium will focus on the state of the art in AI and Quantum Computing, touch upon the convergence of these two transformational technologies, and provide insights on future R&D opportunities from thought leaders in the field.

There will be a mix of individual speakers and panels. Due to ongoing health concerns, in-person is limited and will be by invitation only.

Register in advance for this event!

MATRIX Fall Seminar Series – Dr. Farinaz Koushanfar

Safe and Robust Machine Learning for Real Systems

Dr. Farinaz Koushanfar

University of California San Diego

Friday, October 15, 2021

11 AM – 12 PM CST

https://utsa.zoom.us/j/98953233499

We are at the CUSP of the fourth industrial revolution empowered by machine learning and application automation: seamlessly connecting people, data, and computing machines. Such intelligent technologies however bring numerous potential security vulnerabilities and threats that might severely compromise their safety. In this talk, I present our work in providing end-to-end solutions for practicable robust machine learning based upon co-design and optimization of ML, data, hardware and software. The goal is to characterize the potential ML attack surface, identify and address the nefarious threats in real-time, and devise novel solutions to tackle these problems. In particular, I will show how our work marks the first set of metrics, as well as real-time resource-efficient solutions for machine learning vulnerability characterization, adversarial attacks, and data poisoning.  I conclude by briefly discussing the challenges and opportunities moving forward.

MATRIX Fall Seminar Series – Dr. Jeffrey L. Krichmar

Flexible Path Planning and Navigation in Humans, Rats, and Robots

Dr. Jeffrey L. Krichmar

University of California, Irvine

Friday, October 8, 2021

11 AM – 12 PM CST

https://utsa.zoom.us/j/98953233499

Flexible path planning is necessary for reaching goals and adapting when conditions change. We introduce a biologically plausible path planning model that learns its environment, rapidly adapts to change, and plans efficient routes to goals. Because the algorithm is suitable for spike-based neuromorphic hardware, it has the potential of realizing orders of magnitude gains in power efficiency and computational gains through parallelization. In the first part of the talk, I will introduce the algorithm and show its effectiveness in autonomous robot navigation. In the second part of the talk, I will explore how this biologically inspired path planning algorithm can explain navigation strategies in humans and in rodents. I suggest that mental effort may be the reason we observe variability in human navigation. I also show that the model activity resembles neural correlates of hippocampal preplay and vicarious trial and error (VTE) observed when rodents are navigating mazes. This variation and flexibility may be beneficial for autonomous agents during information gathering and for robotic swarms.

MATRIX Fall Seminar Series – Dr. Gabriela Ciocarlie

Everything is Connected: Security and Reliability for Critical Infrastructure

Dr. Gabriela Ciocarlie

University of Texas San Antonio

Friday, October 1, 2021

11 AM – 12 PM CST

https://utsa.zoom.us/j/98953233499

Abstract

This talk will present different approaches for security and reliability for critical infrastructure. Critical infrastructure often connects cyber and physical layers; in my work, I considered approaches that operate across both dimensions. The talk will focus on IoT, with automated methods to identify and monitor IoT devices, and cellular networks, with methods to detect anomalous cells and diagnose them.

MATRIX Fall Seminar Series – Dr. Gabriel de Erausquin

Use of deep learning strategies to account for biological complexity and the genetics of neuropsychiatric disease

Dr. Gabriel de Erausquin

University of Texas Health San Antonio

Friday, September 24, 2021

11 AM – 12 PM CST

https://utsa.zoom.us/j/98953233499

Abstract:

Complex phenotypes present several challenges for genetic studies including the presence of epistasis, pleiotropy, and heterogeneity. We approached these problems in a data-driven fashion to test the hypothesis that the heritability expected from twin studies but unexplained by genetic studies is distributed in heterogeneous partitions of a complex trait, each with distinct genotypic-phenotypic associations. We designed a machine learning algorithm termed PGMRA to identify naturally occurring partitions in the data in an unsupervised fashion. PGMRA first dissects genome-wide data and uncovers a genotypic architecture composed of sets of variations shared by subsets of individuals. Next, phenotypic data are independently organized into natural sets of features such as clinical manifestations, voxels of neuroimages, or personality traits in a phenomic-like approach. Cross-matching of the two types of sets reveals multiple associations restricted to subgroups of individuals, thereby uncovering the genotypic-phenotypic architecture of a trait and accounting for its distributed genetic risk or propensity.

MATRIX Fall Seminar Series – Dr. Ioannis Agadakos

Chameleons’ Oblivion: Complex-Valued Deep Neural Networks for Protocol-Agnostic RF Device Fingerprinting

Dr. Ioannis Agadakos

Northeastern University

Friday, September 17, 2021

11 AM – 12 PM CST

https://utsa.zoom.us/j/98953233499

Abstract: Prior work has demonstrated techniques for fingerprinting devices based on their network traffic or transmitted signals, which use software artifacts or characteristics of the underlying protocol. However, these approaches are not robust or applicable in many real-world scenarios. In this paper we explore the feasibility of device fingerprinting under challenging realistic settings, by identifying artifacts in the transmitted signals caused by devices’ unique hardware “imperfections”. We develop RF-DCN, a novel Deep Complex-valued Neural Network (DCN) that operates on raw RF signals and is completely agnostic of the underlying applications and protocols. We introduce two DCN variations: a retrofitted Convolutional DCN (CDCN) originally created for acoustic signals, and a novel Recurrent DCN (RDCN) for modeling time series. Our work demonstrates the feasibility of operating on raw I/Q data collected within a narrowband spectrum from open air captures across vastly different modulation schemes. In contrast to prior work, we do not utilize knowledge of the modulation scheme or protocol intricacies such as carrier frequencies. We conduct an extensive experimental evaluation on large and diverse datasets as part of a DARPA red team evaluation, and investigate the effects of different environmental factors as well as neural network architectures and hyperparameters on our system’s performance. Our novel RDCN consistently outperforms all baseline neural network architectures, is robust to noise, and can identify a target device even when numerous devices are concurrently transmitting within the band of interest under the same or different protocols. While our experiments demonstrate the applicability of our techniques under challenging conditions where other neural network architectures break down, we identify additional challenges in signal-based fingerprinting and provide guidelines for future explorations.