Skip to content

Dr. Vibhav Gogate Gets CAREER Award for Artificial Intelligence Work

Recipient of National Science Foundation Grant Wants to Refine, Accelerate Markov Logic Networks

UT Dallas assistant professor of computer science Dr. Vibhav Gogate has earned a National Science Foundation Faculty Early Career Development (CAREER) Award for his work to improve a type of computer algorithm used in artificial intelligence and machine learning.

Gogate’s award, which will run for five years, will support his work to develop new scalable approaches for learning and inference in Markov logic networks (MLNs).

“MLNs are used in many artificial intelligence sub-fields, such as computer vision, robotics, natural language processing and computational biology,” Gogate said. “Algorithms developed in this proposal can be immediately leveraged in these domains. We intend to use MLNs to solve much larger and harder reasoning problems than is possible today.”

In computer science, a “logic” is a formal language used for representing knowledge about the world. Such knowledge is used to answer queries using a logical inference system.

“For example, take the assertion, ‘Everything is big in Texas,’” Gogate said. “If you accept that as fact, and then you learn that John owns a house in Texas, you would infer that John owns a big house.”

However, such logics have no representations for uncertainty — there is no “probably,” only “yes” or “no” — and contradictory knowledge is not allowed: No house in Texas could be small in the above logic scenario.

Vibhav Gogate QuoteMarkov logic addresses these limitations by unifying binary logic with probability,” Gogate said. “It attaches a probability, or weight, to the statement, which becomes ‘It is highly likely that most things you see in Texas are big, but not all.’ Then, if you have a stronger-weighted statement that John loves small houses, Markov logic will help us infer that John has a small house despite being in Texas.”

Although representing real-world knowledge in Markov logic networks is straightforward, answering questions in them via inference can be quite complicated, so much so that it’s often computationally infeasible in practice.

Gogate’s proposal is to enhance the usefulness of MLN-derived probabilistic models. If successful, this new set of inference algorithms could have a broad range of uses, he said.

Dr. Gopal Gupta, Erik Jonsson Chair and head of the computer science department in the Erik Jonsson School of Engineering and Computer Science, described Gogate’s efforts as quite revolutionary.

“Dr. Gogate’s work attempts to reconcile two competing approaches to machine learning and AI — one based on logic and the other based on statistics,” Gupta said. “The research he is proposing to do as part of his NSF CAREER award will not only advance research in machine learning significantly, but also have an impact on its practical applications.”

As with all CAREER awards, Gogate’s plan incorporates an educational and outreach component.

“We’ll use high school and undergraduate students in development and model-building exercises, giving them a look at a potential research career,” Gogate said. “We’ll also produce open-source software that will broaden the adoption of MLN technology, and we’ll promote standardization of datasets and evaluation methodologies.”

Source | UT Dallas News Center 


The UT Dallas Computer Science program is one of the largest Computer Science departments in the United States with over 2,100 bachelor’s-degree students, more than 1,000 MS master’s students, 150 PhD students, and 86 faculty members, as of Fall 2016. With The University of Texas at Dallas’ unique history of starting as a graduate institution first, the CS Department is built on a legacy of valuing innovative research and providing advanced training for software engineers and computer scientists.

UT Dallas and NCWIT Honor 30 North Texas High School Women For Their Outstanding Technical Accomplishments at the NCWIT Aspirations Awards
Prof. Prabhakaran Shares 3D Immersive Tele-Rehabilitation Research at Capitol Hill
Department of Computer Science