Projects

Collaborators: Sriraam Natarajan (UTD), Gautam Kunapuli (UTD) and David Page (Duke University)

Description: The aim of this NIH project is to predict and discover drug-dug interactions pairs. We focus on multi relational data consisting of set of drugs and their relationship with different enzymes, transporters and targets and capture both individual and neighborhood interactions.

Collaborators: Sriraam Natarajan (UTD) and Gautam Kunapuli (UTD)

Description: The aim of this project is to learn first order rules using relational trees as density estimators and then use Gaifman locality thereom to count number of satisfied groundings for these rules. These counts are propositional features for query relation and can be used foe link prediction or node classification.

Collaborators: Sriraam Natarajan (UTD)

Description: The aim of this project is to incorporate human as an ally in the adversarial setting of generative adversarial networks. We show that feature correlations as advice can help learning in data scare environments in deep learning.

Collaborators: Sriraam Natarajan (UTD)

Description: The aim of this project is to lift deep models such as graph convolutional networks (GCNs) to relational setting using statistical relational learning (SRL). We focus on learning a secondary graph from initial knowledge base using a SRL method and learning features from this secondary graph. These features are more richer than simple node based features used as the input to GCNs.