Purushottam Kar

Purushottam Kar 

Purushottam Kar, Ph.D.
Postdoctoral Research Fellow
Machine Learning and Optimization Group
Microsoft Research India

Welcome to my homepage. You will find my contact details and some articles and talks of mine on this site.
Please use the links on the left hand menu to navigate.


  • [15 July 2015] Check out our recent works - the first work develops scalable methods for Robust regression using iterative hard thresholding that can efficiently scale to huge dimensionalities and offer much better performance than state-of-the-art methods. The second work deals with beating the state of the art in Extreme multi-label learning involving millions of labels using locally non-linear embeddings. Preprints available here

  • [15 July 2015] Our work on scalable iterative hard thresholding methods for M-estimation problems (NIPS 2014) will be presented at the International Symposium on Mathematical Programming (ISMP 2015), July 12-17, 2015.

  • [12 July 2015] I will be joining the CSE Department at IIT Kanpur as an Assistant Professor this fall. This webpage will continue to function as my primary homepage.

  • [27 Jun 2015] I have been given a reviewer award by the International Conference on Machine Learning (ICML) for my reviewing work for the 2015 edition of the conference.

  • [26 Jun 2015] Delivered an introductory tutorial on statistical learning theory and concentration at the MSR Summer School on Machine Learning at IISc. Tutorial video and slides would be available here.

  • [25 Apr 2015] Two papers to appear at ICML 2015. One paper deals with improved stochastic optimization techniques for two specific classes of structured loss functions. This supersedes our earlier NIPS 2014 paper for these classes of loss functions. The second paper deals with maximizing precision at the top for bipartite ranking problems.

  • [16 Dec 2014] My Ph.D. thesis has been given the IUPRAI Doctoral Dissertation Award for the year 2014.

  • [9 Sep 2014] Two papers to appear at NIPS 2014 - download preprints here. One paper deals with high dimensional statistical estimation using projected gradient methods. The second paper deals with online and stochastic methods for structured loss functions such as ranking losses, F-measure etc.