I am a postdoctoral Research Fellow working with the Machine Learning and Optimization group. Previously I was a doctoral student at the Indian Institute of Technology Kanpur where I was fortunate to have been jointly advised by Prof. Harish Karnick and Prof. Manindra Agrawal. I did my bachelor studies from the same institution as well.
I am interested in theoretical and applied aspects in machine learning - presently I am interested in topics such as online learning, multi-armed bandits and kernel learning techniques. I also maintain an avid interest in learning theory, data streaming algorithms, geometry and cognitive science.
I maintain a separate homepage containing details about my research and contact information at the following URL (note that this link resides in a different domain): [my homepage].
Details of some of my recent publications can be found below.
- Hsiang-Fu Yu, Prateek Jain, Purushottam Kar, and Inderjit S. Dhillon, Large-scale Multi-label Learning with Missing Labels, in 31st International Conference on Machine Learning, ICML, 2014
- Purushottam Kar, Bharath Sriperumbudur, Prateek Jain, and Harish Karnick, On the Generalization Ability of Online Learning Algorithms for Pairwise Loss Functions, in Proceedings of the 30th International Conference on Machine Learning, Journal of Machine Learning Research, 2013
- Purushottam Kar and Prateek Jain, Supervised Learning with Similarity Functions, in 26th Annual Conference on Neural Information Processing Systems (NIPS), Neural Information Processing Systems Foundation, 2012
Online Learning with Pairwise Loss Functions
MLSIG Seminar Series, Dept. of CSA, IISc, September 12, 2013.
A pre-Weekend Talk on Online Learning
TGIF Talk Series, Microsoft Research India, August 23, 2013.
Explicit Feature Methods for Accelerated Kernel Learning,
Machine Learning and Optimization Group, Microsoft Research India, August 14, 2013.
NIPS Reviewer Award for the year 2013
PC Member: CODS 2014
IEEE Transactions on Neural Networks and Learning Systems (NNLS)
Annual Conference on Neural Information Processing Systems (NIPS 2014)