STARTER-AI Workshop 2024, supported by NSF, NIH, and NVIDIA

May 29-31, 2024

This was our inaugural workshop in a series planned over the coming two years for expanding AI capacity (research, training, and computing) at UTRGV. We believe this workshop helped nucleate a thriving community of AI practitioners comprised of faculty and students.

The workshop included:

• Invited Speakers and Panel Discussion
• Hands-on Introductory AI Programming with course material developed both in-house and by industry leader NVIDIA
• Discussion of AI Courses and Resources available at UTRGV
• Career paths and Opportunities in AI
• Topic workshops to explore AI in specific fields
• "Bring Your Own Problems" Panel session to discuss introduction of AI into your work
• AI@UTRGV showcase of ongoing AI-powered research and development

Locations

• Edinburg - EIEAB 1.204
• Brownsville - BINAB 1.208
• Online

Schedule

Day	Time	Event
May 29	9:00 - 9:30	Check-in / Registration / Coffee
	9:30 - 10:00	Welcome, Chair: Dr. Mohanty Introducing Speakers, Chair: Dr. Kumar
	10:00 - 10:50	UT-Dallas Speaker Dr. Tamil
	10:50 - 11:00	Break
	11:00 - 12:00	UT-Dallas Speaker Dr. Gupta
	12:00 - 1:30	Lunch
	1:30 - 2:30	UTRGV Speaker Dr. Kim
	2:30 - 3:50	AI@UTRGV Showcase Oral Presentations, Chair: Dr. Mukherjee Presentations: 1, 2, 3, 4, 5, 6, 7, 8
	3:50 - 4:00	Break
	4:00 - 5:00	AI@UTRGV Research Poster Showcase, Chair: Dr. Mukherjee Posters: Edinburg, Brownsville
May 30	9:00 - 9:30	Check-in / Coffee
	9:30 - 10:30	TAMU Speaker Dr. Chakravorty
	10:30 - 10:50	NVIDIA DLI: Fundamentals of Deep Learning, Chair: Dr. Enriquez
	10:50 - 11:00	Break
	11:00 - 12:00	NVIDIA DLI: Fundamentals of Deep Learning, Chair: Dr. Enriquez
	12:00 - 1:30	Lunch
	1:30 - 3:50	NVIDIA DLI: Fundamentals of Deep Learning, Chair: Dr. Enriquez
	3:50 - 4:00	Break
	4:00 - 5:00	NVIDIA DLI: Fundamentals of Deep Learning, Chair: Dr. Enriquez
May 31	9:00 - 9:30	Check-in / Coffee
	9:30 - 9:50	UTRGV Speaker Dr. Kumar
	9:50 - 10:50	Topic - Getting Started on Cradle HPC, Chair: Dr. Enriquez, Dr. Kim, Dr. Mohanty
	10:50 - 11:00	Break
	11:00 - 12:00	Closing Panel Discussion with Potential Collaborators, Chair: Dr. Martirosyan

 

Section Chairs and Invited Speakers

 

Invited Speakers

Abstract:

As artificial intelligence (AI) continues to revolutionize the healthcare landscape, it is imperative to ensure that these advancements contribute to the goal of achieving inclusive and equitable healthcare for all. This talk will explore the critical intersection of AI and healthcare, Read more. focusing on the imperative need to address disparities and promote fairness in the development and deployment of healthcare AI systems.

---

Abstract:

Computational logic can play a central role in the quest for building interpretable, explainable, and trustworthy AI systems. We discuss how default theories expressed as logic programs represent inductive generalizations that, in turn, can represent interpretable and explainable Read more. machine learning models. Rule-based machine learning algorithms can be subsequently designed that are competitive with mainstream state-of-the-art machine learning systems. We discuss the application of these algorithms to making convolutional neural networks---used for image recognition---explainable. An overview will be given of the s(CASP) goal-directed predicate answer set programming system and show how it can be used in flexible ways to generate explanations for predictions made by machine learning models as well as perform counterfactual reasoning. We will also discuss how s(CASP) and large language models together can be used to develop trustworthy (domain-specific) natural language understanding systems.

---

Abstract:

This presentation provides an insightful overview of Artificial Intelligence (AI), elucidating its fundamental concepts and detailing the historical evolution from early innovations like the perceptron to advanced applications in deep learning technologies. It also showcases AI's Read more. expansive role in fostering interdisciplinary research, demonstrating how AI tools and methodologies can be harnessed to bridge various academic and industrial fields, thus enhancing innovation and solving multifaceted problems. The session highlights the transformative potential of AI in sectors such as healthcare, science, engineering, business, and education. Additionally, it introduces significant past and ongoing projects within Dr. Kim’s research lab that exemplifies the practical application of AI in interdisciplinary contexts. Attendees will be introduced to the lab’s resources and the collaborative environment that propels the frontiers of AI research in UTRGV.

---

Abstract:

The national computing landscape offers a rich set of opportunities for researchers to engage different computing modalities in their research. In addition to computing resources, we today consider opportunities to develop the human component of this Dr. Chakravorty will talk about Read more. some of the recent developments that offer researchers access to computing and training resources. be used for research, training, and educational activities, often at no cost to researchers. Campus, local, regional, and national aspects need to be considered as one considers using these resources. Dr. Chakravorty will discuss relevant programs and the mechanisms for researchers to apply to them. He will also highlight the training and computing resources offered by the High Performance Research Computing center at Texas A&M University. In particular, he will talk about the recently deployed Launch cluster, a regional resource for computing that was funded by the National Science Foundation.

Bio:

Dhruva Chakravorty is the Director of User Services and Research at Texas A&M High Performance Research Computing and actively supports the software and research needs of researchers. His research interests lie in using large scale computational methods and Read more. cyberinfrastructure technologies. He has developed application performance sets for composable cyberinfrastructure as part of NSF FASTER and NSF ACES. He leads the NSF SWEETER (Role PI, OAC-1925764) and BRICCs programs (Role PI, OAC-2019136) that connect researchers at universities and community colleges in Texas, New Mexico, and Arizona to regional and national CI resources. Dhruva coordinates the extensive HPRC training program. As part of this he has developed education and training materials for formal and informal instruction in CI. He leads the efforts to offer micro-credential in research-CI skills and leads the “Summer Computing Academy” program for K-12 students that has been supported by the National Security Agency GenCyber, and the Texas Workforce Commission. He serves as the Project Manager and co-PI for NSF ACES testbed (OAC-2112356) and is the PI on the forthcoming NSF Launch regional computing resource (OAC-222895) that connects 11 smaller institutions across Texas.

---

Abstract:

Cybersecurity attacks continue to hamper today's Internet-connected society and are making headline news frequently. This presentation will provide some of significant threats posed by these cybersecurity vulnerabilities and attacks. Efficient detection and mitigation of Read more. some of these security attacks have been challenging for traditional network systems. To help in this effort, Artificial Intelligence and Machine learning schemes are being developed for more efficient protection against such attacks. In this presentation, we will discuss these challenges and the mitigation efforts using Machine learning schemes being developed in our research group at UTRGV.

 

AI@UTRGV Research Showcase Presentations

Oral Presentations

 

Presenter 1:

AI-Driven Design and Synthesis of Nanomaterials for the detection of fentanyl and derivatives

Author(s): Amber Garcia and Dr. Julie P. Vanegas

Abstract: The opioid crisis, significantly intensified by the widespread abuse of synthetic opioids like fentanyl, represents a formidable public health challenge. Conventional opioid detection methodologies lack the sensitivity and specificity required to effectively counter Read more. the rapid emergence and proliferation of these potent synthetic substances. This project proposes the novel development of aptamer-based nanosensors, employing cutting-edge advancements in nanotechnology, biochemistry, and artificial intelligence (AI) to overcome this inadequacy.

---

Presenter 2:

Autoencoders in Reinforcement Learning

Author(s): Hector Lugo, Arturo Meza-Canales, Osvaldo Garza, Dr. Erik Enriquez, and Dr. Dong-Chul Kim

Abstract: This project aims to enhance the efficiency of machine learning algorithms, with a particular focus on reinforcement learning (RL) models. Our objective is to reduce the complexity of environmental observations in order to increase the overall efficiency and Read more. effectiveness of RL agents. To achieve this, we will integrate an autoencoder, a neural network designed to compress data while preserving its key features. Our research will employ several benchmark models from MuJoCo, allowing us to compare the model’s performance when trained with encoded versus raw observational data. Through this investigation, we anticipate uncovering insights that could lead to more streamlined training, and create avenues to train RL agents on more complex tasks.

---

Presenter 3:

Control Strategies for Cooperative Multi-Robot Tasks Using Multi-Agent Reinforcement Learning

Author(s): Daniel Masamba, Tyler Morgan, Dr. Erik Enriquez, and Dr. Dong-Chul Kim

Abstract: This research project used a curriculum learning approach to train simulated agents with Reinforcement Learning (RL) techniques. By training agents on tasks simpler than the one meant for them to achieve, we saw more efficient and effective training for desired tasks than Read more. agents with no previous training. This could also lead to multiple, previously trained agents being able to collaborate on tasks more effectively. The research project’s priorities shifted from exploring the applications of Multi-Agent Reinforcement Learning (MARL) as we explored in more depth the variations of most effectively training a single agent as a precursor. However, our methods could lead to a more meaningful exploration of MARL strategies in the future.

---

Presenter 4:

The Development of a Disability Assessment Tool using Biosensors and Machine Learning for Persons with Multiple Sclerosis

Author(s): Jofred Gonzalez, Dr. Damian Valles, and Dr. John W. Farrell III CSCS, CPSS

Abstract: The development of a tool to assess Multiple Sclerosis severity using Inertial Mass Units (Motion Sensors) and Machine Learning aims to simplify the assessment of MS severity for the lower extremities of the human body. More specifically, this presentation Read more. discusses how data was collected and prepared to be processed by a non-binary classification machine learning algorithm to create a model that will output a number from 0 to 10 emulating the Expanded Disability Status Scale assessment.

---

Presenter 5:

AI based prediction of the charge storage capability of Lithium-ion battery materials

Author(s): Manoj Chhetri and Dr. Karen Martirosyan

Abstract: In this report, a machine learning approach was implemented to a dataset of 2345 entries of rechargable Li-ion battery materials, obtained from MaterialsProject online portal, to predict the gravimetric charge storing capacity (mAh/g) which is a crucial parameter in Read more. a battery for its energy storage capabilities. This study highlighted the possibility of using machine learning models in materials science by integrating domain-specific knowledge with advance statistical techniques. It provides insight for informed decision-making and tuning the important parameters to achieve the highest possible charge storage capacity in battery materials design and optimization process. By accurately predicting the gravimetric capacity of battery materials taking into account of intrinsic battery properties such as voltage, energy density and stability, our model contributes to discover and deploy more efficient and sustainable battery solutions.

---

Presenter 6:

IntelliBeeHive: An Automated Honey Bee, Pollen, and Varroa Destructor Monitoring System

Author(s): Christian I. Narcia-Macias, Joselito Guardado, Jocell Rodriguez, Dr. Joanne Rampersad-Ammons, Dr. Erik Enriquez, and Dr. Dong-Chul Kim

Abstract: Utilizing computer vision and the latest technological advancements, in this study, we developed a honey bee monitoring system that aims to enhance our understanding of Colony Collapse Disorder, honey bee behavior, population decline, and overall hive health. The system is Read more. positioned at the hive entrance providing real-time data, enabling beekeepers to closely monitor the hive's activity and health through an account-based website. Using machine learning, our monitoring system can accurately track honey bees, monitor pollen-gathering activity, and detect Varroa mites, all without causing any disruption to the honey bees. Moreover, we have ensured that the development of this monitoring system utilizes cost-effective technology, making it accessible to apiaries of various scales, including hobbyists, commercial beekeeping businesses, and researchers. The inference models used to detect honey bees, pollen, and mites are based on the YOLOv7-tiny architecture trained with our own data. The F1-score for honey bee model recognition is 0.95 and the precision and recall value is 0.981. For our pollen and mite object detection model F1-score is 0.95 and the precision and recall value is 0.821 for pollen and 0.996 for "mite". The overall performance of our IntelliBeeHive system demonstrates its effectiveness in monitoring the honey bee's activity, achieving an accuracy of 96.28 % in tracking and our pollen model achieved a F1-score of 0.831.

---

Presenter 7:

Artificial Intelligence in Health Equity

Author(s): Jose Arizpe and Dr. Sanjeev Kumar

Abstract: Artificial Intelligence has been integrated in various aspects of modern life with deep learning to create automated self driving cars and machine learning used to create chatbots among other things. The reason AI has grown in such popularity involves the amount of Read more. data that is produced allowing for AI models to be more sophisticated than previously possible. Thus, the introduction of AI in healthcare is an area of focus that has continuously grown as researchers attempt to decipher the best course of action when utilizing AI in health. There is undoubtably many benefits that could be gained from utilizing AI in health such as for disease prediction, bridging gaps between language barriers, and personalized treatment. However, there are many barriers that must be overcome to reap the benefits as equity in health is an important factor that must be considered to ensure that everyone can benefit equally.

---

Presenter 8:

Application of Machine Learning for Security of Medical Devices

Author(s): Oscar Chaidez Amaya and Dr. Sanjeev Kumar

Abstract: Over recent years the number of medical devices has increased and more people have become dependent on them, with these many security threats have risen. Since these medical devices exchange information about the medical aspect of the patient, Read more. personal data is exchanged and stored regularly, which can be stolen or even interfered with to diminish the functionality of this device. By implementing machine learning into the authentication process between the sensors collecting data on the device and the servers storing information, unauthorized access can be minimized, making these devices more secure from attacks. Through this presentation, a more secure authentication process between the server and sensors is presented with the use of machine learning classifiers.

 

Poster Showcase

 

Edinburg:

Ad Hoc Community Building through Chatbot during Power Outage

Author(s): Dr. Jun Sun

Abstract: This study delves into the design and impact of a novel chatbot system aimed at fostering community resilience during power outages by facilitating ad hoc community building and resource sharing. Using SMS to bypass internet connectivity issues, the chatbot Read more. facilitates communication, updates, and peer assistance among those affected by outages. This research highlights digital tools' role in boosting community resilience and emergency strategies.

---

Particle Swarm Optimization for Training Quadrotor PID Controller

Author(s): Eric X. Rodriguez and Dr. Qi Lu

Abstract: The PID (Proportional-Integral-Derivative) controller is a simple and effective method employed for system monitoring and control. A notable drawback of the PID control scheme is utilizing disproportional gains that can lead to system failure. A common approach to Rread more. addressing this is manually tuning each PID parameter for every assigned response system, but it is time-consuming and may still leave the system vulnerable to inadequate flight responses. PSO (Particle Swarm Optimization) demonstrates rapid convergence speed in obtaining optimal parameters, requiring adjustment of only a few setting parameters for convergence and exhibits high calculation efficiency. Moreover, it often achieves high-quality solutions in a shorter time compared to other stochastic methods.

---

Integrating AI for Investigating Pt(0)-Mediated C–CP Bond Activation in Phosphaalkynes

Author(s): Roberto Escobar, Dr. Abdurrahman C. Ateşin, Christian Müller, William D. Jones, and Dr. Tülay Ateşin

Abstract: Carbon–carbon (C–C) bond activation has gained increased attention as a direct method for the synthesis of pharmaceuticals. Due to the thermodynamic stability and kinetic inaccessibility of the C–C bonds, however, activation of C–C bonds by homogeneous Read more. transition-metal catalysts under mild homogeneous conditions is still a challenge. Most of the systems in which the activation occurs either have aromatization or relief of ring strain as the primary driving force. The activation of unstrained C–C bonds of phosphaalkynes does not have this advantage. This study employs Density Functional Theory (DFT) calculations using the Gaussian16 package to elucidate Pt(0)-mediated C–CP bond activation in phosphaalkynes. Leveraging AI-driven methods, the research integrates file preparation, advanced optimization techniques, and comprehensive visualization tools to analyze and optimize potential energy surfaces (PES). Additionally, robust optimization techniques using `scipy.optimize.minimize` function, enable efficient identification of optimal molecular and critical structures such as a saddle point. The utilization of `RegularGridInterpolator` ensures accurate energy estimation, even with sparse data points. This approach not only addresses challenges in molecular configuration identification but also contributes to a deeper understanding of molecular energy landscapes, with implications for molecular design and analysis. The importance in AI in a field where coding literacy is not widespread, chemists may now have a powerful means to unravel complex mechanisms and expedite tasks, potentially revolutionizing new ways chemistry is approached.

---

Artificial Intelligence in Health Equity

Author(s): Jose Arizpe and Dr. Sanjeev Kumar

Abstract: Artificial Intelligence has been integrated in various aspects of modern life with deep learning to create automated self driving cars and machine learning used to create chatbots among other things. The reason AI has grown in such popularity involves the amount of Read more. data that is produced allowing for AI models to be more sophisticated than previously possible. Thus, the introduction of AI in healthcare is an area of focus that has continuously grown as researchers attempt to decipher the best course of action when utilizing AI in health. There is undoubtably many benefits that could be gained from utilizing AI in health such as for disease prediction, bridging gaps between language barriers, and personalized treatment. However, there are many barriers that must be overcome to reap the benefits as equity in health is an important factor that must be considered to ensure that everyone can benefit equally.

---

Application of Machine Learning for Security of Medical Devices

Author(s): Oscar Chaidez Amaya and Dr. Sanjeev Kumar

Abstract: Over recent years the number of medical devices has increased and more people have become dependent on them, with these many security threats have risen. Since these medical devices exchange information about the medical aspect of the patient, Read more. personal data is exchanged and stored regularly, which can be stolen or even interfered with to diminish the functionality of this device. By implementing machine learning into the authentication process between the sensors collecting data on the device and the servers storing information, unauthorized access can be minimized, making these devices more secure from attacks. Through this presentation, a more secure authentication process between the server and sensors is presented with the use of machine learning classifiers.

---

Brownsville:

A machine learning based search pipeline to detect gravitational waves from core collapse supernovae

Author(s): Raul Espinoza Perez and Dr. Soma Mukherjee

Abstract: Numerous core collapse supernovae (CCSN) waveforms have been simulated in high-performance computing facilities worldwide, each showing unique characteristics influenced by factors like progenitor mass, angular momentum, gravitational wave (GW) energy, Read more. peak frequency, duration, and equation of state. These supernovae are key targets in LIGO's fourth observation run (O4), with efforts to detect them boosted by machine learning (ML) techniques, particularly convolutional neural networks (CNN).

---

Smart traffic light with the aid of AIg

Author(s): Jose Garcia

Abstract: Traffic lights offer a system that can facilitate the flow in big cities and small towns, yet the there are some time gaps where the traffic flow slows down, and some drivers have to wait for a long period for the light to turn green. Smart traffic lights with the aid of AI can Read more. detect when the traffic flow increases and when it decreases, thus allowing the waiting time to either be longer or reduce depending on the number of cars passing during a period of time.