CAIML Newsletter, March 2023
March 15, 2023
Introduction
The Center for Applied AI and Machine Learning (CAIML) was founded in 2019 by Dr. Doug DeGroot and Dr. Gopal Gupta with the goal of bringing the benefits of advanced artificial intelligence and machine learning technologies to the industry. Artificial intelligence and machine learning technologies are having a transformational impact on our economy and society. CAIML aims to help companies adapt to this transformation. CAIML researchers help companies use state-of-the-art AI and Machine Learning technologies to solve their most vexing problems. CAIML also engages in training industry workforce in these new technologies. CAIML also develops AI and Machine Learning technology in-house that it makes available to others for licensing and for use in joint projects. In 2022, CAIML also became part of the Richardson Innovation Quarters (RIQ). CAIML Is largely supported by industry projects. The RIQ component is funded by the UT Dallas Office of VP Research.
Current and Past Projects
CAIML has already completed a number of projects, and several are ongoing. A brief synopsis of each project is given below.
Vistra Corp
The project focused on electricity price prediction in the presence of battery storage using machine learning and optimization technology. The project was motivated by Vistra’s 400-megawatt energy storage system in Moss Landing, CA, which required its price forecasting algorithms to be reworked. 2020.
InfoVision, Inc
The goal of this project was to use drone technology for warehouse automation, specifically, automated inventory-taking. 2021.
Atos Corp
Explainability is a major challenge for machine learning systems. As part of this project, explainable machine learning algorithm developed by CAIML researchers were refined and incorporated in Atos’ machine learning tool chain that it offers to its customers. 2021.
Rayburn Electric
The project’s goal is to predict load on the Texas grid, forecast electricity prices, and optimize bids for Texas electricity market (ERCOT) using machine learning technologies. 2021-23.
Tessolve, Inc:
The project deals with routing and assignment problem for a PCB-based logic tester system that can auto-assign ATE (Automatic Test Equipment) Tester channel pins to the DUT (Device Under Test) signal pins. 2022-23.
Digit7 Corp
Digit7 Corp is in the business of developing touchless self-checkout systems. CAIML researchers will help Digit7 in making their solution scalable and achieve high accuracy. 2023.
Nexco Corp
Nippon Expressway Co. is a Japanese company responsible for maintaining highways in Japan. Nexco uses a machine learning based tool to estimate the health of a road. CAIML researchers will help Nexco develop a system to automatically prioritize road repairs based on various constraints. 2023.
Vistra Personnel Training in AI Technologies
CAIML has taught a series of beginners, intermediate, and advanced courses to Vistra personnel to bring them up to speed on state-of-the-art AI and machine learning technologies. These courses are taught by CAIML researchers, who are CS faculty members.
In-house Technology
Several any in-house CAIML tools are available for licensing and for use in industry collaborative projects.
- Explainable Machine Learning Systems: CAIML researchers have developed a family of explainable machine learning systems that can help companies understand their machine learning models. Machine learning models are black-boxes that make predictions without explaining how they were made. Thus, if the model is biased or unfair is hard to know. Thus, ability to explain a prediction is an important characteristic of a model. We have developed the FOLD family of explainable machine learning tools that are available for licensing. Deprecated versions of these tools are available for free.
- Interactive Chatbots that Understand Humans: CAIML researchers have leveraged ChatGPT (large language models) and its own automated commonsense reasoning tool to develop technology that allows building chatbots that can interact with a human user and “understand” what he/she is saying. The chatbots are domain-specific but are guaranteed to be safe and provide correct response. CAIML can develop these interactive chatbots for its industry customers.
- Automated Reasoning Systems: CAIML researchers have built technology to model the human thought process. This system, called s(CASP), is being used by many groups around the world to develop applications that rely on automating commonsense reasoning. CAIML researchers can help you build these intelligent applications (see Nexco project above, for example).
- What’s Happening?: This system analyzes videos and detects activities of interest through analysis of time series data using statistical relational learning (SRL) models.
- Statistical Relational Learning (SRL): Several tools are available for combining learning and reasoning through the use of SRL technology. These include reasoning in hybrid probabilistic domain, cost sensitive learning, etc.
CAIML Researchers
CAIML has more than a dozen researchers with an array of skills in various areas of AI. Most of these researchers come from the Department of Computer Science and the Jonsson School of Engineering and Computer Science. Once a project comes to CAIML, a team of researchers is assembled that includes faculty members and graduate students to work on the project. The industry projects described above have different CAIML researchers who lead them.
More Information
For more information or to learn more about industrial collaboration opportunities, please contact, Prof. Doug DeGroot (Doug.Degroot@utdallas.edu) or Prof. Gopal Gupta (Gopal.Gupta@utdallas.edu).