MATRIX Spring Seminar Series – Dr. Furong Huang

Trustworthy Machine Learning in Complex Environments

Dr. Furong Huang

Assistant Professor of the Department of Computer Science

University of Maryland

1/27/2023

11AM – 12PM CST

https://utsa.webex.com/utsa/j.php?MTID=m9ab3786b04dc901b8cd82f477b43ddc3

Abstract: With the burgeoning use of machine learning models in an assortment of applications, there is a need to rapidly and reliably deploy models in a variety of environments. These trustworthy machine learning models must satisfy certain criteria, namely the ability to: (i) adapt and generalize to previously unseen worlds although trained on data that only represent a subset of the world, (ii) allow for non-iid data, (iii) be resilient to (adversarial) perturbations, and (iv) conform to social norms and make ethical decisions. In this talk, towards trustworthy and generally applicable intelligent systems, I will cover some reinforcement learning algorithms that achieve fast adaptation by guaranteed knowledge transfer, principled methods that measure the vulnerability and improve the robustness of reinforcement learning agents, and ethical models that make fair decisions under distribution shifts.

MATRIX AI Fall Seminar Series – Dr. Mohamed Amer

Responsible AI: How do we get there?

Dr. Mohamed Amer

Robust AI

11/11/2022

11AM CST – 12PM CST

https://utsa.webex.com/utsa/j.php?MTID=m66e32a039daa62769074b57712455b0c

Abstract: Machine learning models are getting larger and more complex. While machine learning provides a powerful set of tools, the models fail unexpectedly, harbor biases, and are poorly understood. Nonetheless, they are deployed at scale. A machine learning-driven system requires extensive engineering & computational resources, and massive amounts of data. Currently, these resources are only available to big players, governments & corporations, and there is a lack of tools or processes to enable auditing, compliance, and evaluation of such complex systems. The models are sometimes governed by a set of rules defined by the developers, which are often subjective and not well thought out. This talk will cover a wide variety of topics around responsible AI highlighting deep learning current challenges, vulnerabilities, and possible ways to mitigate them.

MATRIX Fall Seminar Series – Dr. Ananth Grama

Computational Methods for Analysis of Functional Brain Images

Dr. Ananth Grama

Purdue University

11/4/2022

11AM – 12PM CST

WebEx: https://utsa.webex.com/utsa/j.php?MTID=m0dd130eaf6ddc33055fb62e2e1c06923

Abstract:  Rapid advances in neuroimaging have resulted in large repositories of images with high temporal and spatial resolutions. This has motivated complex connectomic analyses aimed at understanding representation and processing of stimuli . Such analyses require novel computational tools that significantly extend the state-of-the-art in machine learning and data science. These studies have far-reaching implications for the fields of precision psychiatry, behavioural analysis, and neurodegeneration, in addition to applications in AR/VR and advanced human interfaces.

In this talk, I will discuss recent work in our group on understanding neuronal response to naturalistic visual stimulus — I present two methods based on archetypal analysis and deep learning to: (a) find interpretable representations of fMRI response, (b) predict objects in visual frames, and (c) reconstruct visual inputs. I will also briefly describe computational methods to characterize individual-level uniqueness of functional connectomes. I will present three methods based on matrix sampling, graph alignment, and linear algebraic techniques to accurately identify the identity of individuals, and the nature of the cognitive task being performed.  I will conclude my talk by highlighting the immense significance and challenges posed by problems in the field of neuroimage analyses.

Distinguished Speaker Series – Dr. John Gustafson

MATRIX Fall Seminar Series – Dr. Soheil Kolouri

Lifelong Learning Machines – Approaches and Challenges

Soheil Kolouri

Vanderbilt University

Friday, October 28, 2022

11 AM – 12 PM CST

WebEx: https://utsa.webex.com/utsa/j.php?MTID=m0d75dea1a1670d0f72425cd9d59073ac

Abstract: The next wave of AI demands new machine learning frameworks that can continually learn and adapt to the stream of nonstationary multimodal information. This challenge is called continual, lifelong, or incremental learning in the ML community. Despite their phenomenal performance in various applications, deep neural networks are prone to forgetting their previously learned information upon learning new ones. This phenomenon is called “catastrophic forgetting” and is deeply rooted in the stability-plasticity dilemma. Overcoming catastrophic forgetting in deep neural networks has become an active field of research. In this talk, I will review some of our recent works on overcoming catastrophic forgetting in continual learning, from novel regularization-based approaches to sparsity and heterogeneous dropout. I conclude my talk by discussing our ongoing research on continual learning and some of the open problems and challenges.

MATRIX Fall Seminar Series – Dr. Kal Clark

Eye-AI: A New Approach to Quality, Labeling, and Vetting in Radiology

Dr. Kal Clark

University of Texas Health San Antonio

10/21/2022

11AM – 12PM CST

Abstract: The predominant source of medical errors in radiology is surprisingly not deficient medical knowledge. Rather, errors primarily stem from the methods radiologists use to visually inspect the image, referred to as “perceptual errors.” Perception-related errors are estimated to reach 33% on abnormal chest radiology exams. To mitigate this problem, radiologists employ personalized and high-dimensional visual search strategies, otherwise known as search patterns. Quantitative descriptions of these search patterns via direct observation or self-report are not only unreliable, but also interfere with quality improvement interventions, which therefore negatively impact patient care.  Through the use of eye-tracking technology and feature extraction we can employ novel techniques to improve health care quality, passively label datasets, and vet the next cohort of medical practitioners.