Harsha Kokel

Hello, I am Harsha Kokel, a Ph.D. student working with Prof. Sriraam Natarajan and the StARLinG Lab at University of Texas at Dallas. My research focuses on leveraging domain knowledge to improve generalization in various different AI/ML approaches.

I am interested in knowledge-based systems that learn not only from data but also from the knowledge about the domain. This additional knowledge could either be the structure of the domain like a graphical model, or some qualitative information like monotonicty or synergy, in sequential decision making it could also be action schemas (a planner). Since I am interested in involving human experts in the learning process, I am also interested in explainable AI.

Publications

A Probabilistic Approach to Extract Qualitative Knowledge for Early Prediction of Gestational Diabetes,
Athresh Karanam, Alexander L. Hayes, Harsha Kokel, David M. Haas, Predrag Radivojac, Sriraam Natarajan, In AIME 2021.
paper | blog
RePReL: Integrating Relational Planning and Reinforcement Learning for Effective Abstraction,
Harsha Kokel, Arjun Manoharan, Sriraam Natarajan, Balaraman Ravindran, Prasad Tadepalli, In ICAPS 2021.
paper | blog | code | supplemental
Graph Sparsification via Meta-Learning,^{* contributed talk (4/22 accepted paper)}
Guihong Wan, Harsha Kokel, In Deep Learning for Graphs (DLG) Workshop at AAAI 2021.
paper | slide | video
A Unified Framework for Knowledge Intensive Gradient Boosting: Leveraging Human Experts for Noisy Sparse Domain,
Harsha Kokel, Phillip Odom, Shuo Yang, Sriraam Natarajan, In AAAI 2020.
paper | blog | code | supplemental | poster | slide | DOI