Professor Pujari works primarily in AI, Data Mining, and Combinatorial Algorithms. Some of the highlights are the following
Shape from Silhouettes– During 80’s, the problem of recovery of 3D shape (and structure) from multiple images was one of primary areas of research in Computer Vision. He proposed a novel method of recovering shapes from silhouettes and provided a very efficient algorithm. One of the novelties was to have cameras with changing direction of views which was later identified as Active Vision. His work was one of first proposals in Active Vision. (published in CVGIP, PRL
Knowledge Compilation and Nonmonotonic Reasoning– He proposed a novel method of knowledge compilation by computing prime implicates of propositional logic (published in Journal of Automated Reasoning)
Temporal Constraint Satisfaction Problem– This is yet another area of his significant contribution. His proposal, INDU-Interval Duration Network, initiated many new research activities in TCSP. His proposal of PIDN was presented in IJCAI (Published mostly in AI Conferences, IJCAI, Australian AI, ICTAI, PRICAI etc.)
Virus and Intrusion Detection– His proposal of ML approach to virus detection and masquerade detections is one of the first work of use of text-mining for IDS and are well cited.
Energy Efficient Target Coverage Problem– He developed several theoretical results on this subject. The problem is still unsolved and hoping to address this problem sometime. (Published in ICDCN, IEEE Sensors)
Recommender System and Matrix Factorization– This is the most recent area of his activities. His group’s novel interpretation of Maximum Margin Matrix Factorization yielded several important results. The group demonstrated this in several areas such as in Recommender System, in Multi-label Classification and in Transfer Learning.
Others- In addition to these areas, he takes interest in Scientific approaches preservation of Miniature Painting, Cyber-Physical Systems and AI for literature.