My research interests lie in the fields of Artificial Intelligence and Machine Learning and their application to healthcare problems. More specifically, I am interested in the areas of Relational Learning, Reinforcement Learning, Graphical Models, and Planning. Please read more about our projects and team on our team webpage.
Till 2013, I was a faculty member at Translational Science Institute of Wake Forest School of Medicine. I was a Post-Doc earlier at the Department of Computer Science in the University of Wisconsin Madison, working with Professors Jude Shavlik and David Page.
Note to incoming students: If you are interested in working with me, please register for a course that I teach. I do not hire students before they take my course. Also, I do not have any internship/short-term positions. Please do not contact me if you want to work with me for less than a year.
Our paper on learning Poisson Sum Product Networks has been accepted at AAAI. Pdf coming soon.
Our paper on Markov Logic Networks for extracting adverse drug events from text will appear in KAIS journal. This work explores the use of MLNs for medical information extraction, more precisely, adverse drug events. We present a NLP based pipeline for extracting articles from PubMed.
3 papers accepted at ILP. Advice giving + transfer learning (Phillip), Boosting MLNs using Tuffy (Marcin) and SRL for Relation Extraction (Dileep). Pdfs coming soon.
Phillip's ECML submission on Actively Interacting with Experts: A Probabilistic Logic Approach, has been accepted. The paper extends the active advice seeking framework to relational setting where a group of regions of the feature space can be naturally represented and presented to the user for queries. This allows for rich interactions between the learning algorithm and the human expert. Human is seen as more than a mere labeller by our relational learning method.
Our collaborative paper between the data and design teams of health informatics on Identifying Rare Diseases from Behavioural Data: A Machine Learning Approach is accepted at IEEE CHASE 2016. This paper explores the role of machine learning in predicting the occurrences of rare diseases based on social interactions.
Phillip's Active advice-seeking for Inverse Reinforcement Learning accepted at AAMAS 2016. The paper takes a new look at IRL and introduces seamless interactions with human experts. Called active advice seeking, this work, inspired by active learning, seeks advice from human experts based on the uncertainty in demonstrations.
Mayukh's paper on Scaling Relational Learning and Inference via Approximate Counting is accepted at SDM 2016. Counting is the central task for probabilistic learning and reasoning in large domains. However, it is non-trivial and this paper takes a graph-theoretic approach to counting. A nice message passing algorithm based on belief propagation is presented for this approach.
Learning Relational Continuous-Time Bayesian Networks paper by Shuo Yang accepted at AAAI 2016. This paper extends the CTBN formulation to relational domains. It also presents a powerful learning method based on boosting that allows for CTBNs to be applied to large, noisy domains such as EHRs.
I am the co-chair of the AAAI Student Outreach Activities. If you are a student attending AAAI 2016, you should check the website out.
Our DARPA grant on Cognitively Coherent Human-Computer Communication is funded.
I was awarded the Trustees Award for Teaching Excellence at IU.
I co-organized the UAI 2015 workshop on Statistical Relational AI (STAR AI '15).
Thanks to Turvo Inc for the generous gift.
Thanks to XEROX PARC for the faculty award.