Purushottam Kar
 |
Purushottam Kar, Ph.D. |
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.
News
[24 May 2023] [New] Our work giving a more accurate way to train large language models for extreme classification applications will appear at KDD 2023. A preprint and code will be released [here] soon.
[20 May 2023] [New] Our work offering personalized retrieval at the scale of millions of users and millions of items will appear at SIGIR 2023. A preprint and code will be released [here] soon.
[20 January 2023] [New] Congratulations to Sharath H. Padmanabha, Fahad Shaikh, Mayank Bansal, Debanjan Chatterjee and Preeti Singh receiving the Best Paper Award at the 16th Innovations in Software Engineering Conference (ISEC 2023) for a paper that makes innovative use of ranking and recommendation techniques to help students who are learning to code by automatically suggesting tips on program repair when they make coding mistakes.
[15 January 2023] [New] Congratulations to Amit Chandak for receiving a Best Research Paper Nomination at the 10th ACM IKDD Conference on Data Sciences (CODS 2023) for a paper on techniques for creating light-weight yet expressive deep ensembles using first order perturbation methods.
[21 November 2022] [New] Our work on the first provably robust algorithms for learning generalized linear models will appear at AAAI 2023. A preprint and code will be released [here] soon.
[21 November 2022] [New] Our work extending the state of the art on light-weight algorithms for pedagogical code repair to help conduct large programming courses such as ESC101 will appear at ISEC 2023. A preprint and code will be released [here] soon.
[21 November 2022] [New] Our work on the design and deployment of a searchable question bank to assist instrutctors and tutors of large programming courses like ESC101 will appear at ISEC 2023. A preprint and code will be released [here] soon.
[28 October 2022] [New] Our work on efficiently creating deep ensembles using computationally inexpensive gradient perturbations will appear at CODS 2023. A preprint and code will be released [here] soon.
[26 October 2022] [New] A survey paper on our interdisciplinary work on building tools to help conduct large programming course using ML/AI and PL tools will appear in a forthcoming issue of the CACM. A link to the paper is available [here].
[22 October 2022] [New] Our work on training large language models with hundreds of millions of parameters on extreme classification problems with millions of categories will appear at WSDM 2023. A preprint and code will be released [here] soon.
[03 March 2022] Our work on multi-modal (textual + visual) learning with millions of categories will appear at CVPR 2022. A preprint and code will be released [here] soon.
[21 January 2022] Our work on a lazy algorithm for accelerated GCN training will appear at ICLR 2022. A preprint is available [here] and code will be released soon.
[01 January 2022] Congratulations to Debojyoti Dey and Bhaskar Mukhoty for receiving a Best Research Paper Nomination at the 9th ACM IKDD Conference on Data Sciences (CODS 2022) for their paper on techniques for optimizing functions that offer only local convexity around their global optimum.
[31 October 2021] Congratulations to Fahad Shaikh and Sharath HP for winning the SIIC Student Innovation Award at the 54th IITK Convocation for thesis work developing the PRIORITY system to aid the conduct of large programming courses such as ESC101 (500+ students).
[31 October 2021] Congratulations to Ayush Nagal for winning the Cadence Gold Medal at the 54th IITK Convocation for thesis work developing robust machine learning-based calibration algorithms for low-cost air quality monitoring sensors.
[15 October 2021] Our work on Siamese Networks for Extreme Classification has been given the Greatness Award under the Engineering Excellence category by Bing Ads, Microsoft. Congratulations to all my collaborators who made this possible.
[03 October 2021] Our work on provably global optimization for locally convex functions arising in generalized linear modelling tasks will appear at CODS 2022. Code for our method and a preprint are available [here].
[27 July 2021] Our work developing the first algorithm for non-parametric regression with adversarially corrupted data and provable breakdown points under general conditions will appear at ECML 2021 and the Machine Learning journal. Code for our method and a preprint are available [here].
[12 May 2021] Our work on Siamese networks for deep extreme classification with 100 million labels will appear at ICML 2021. Code for our method and a preprint are available [here].
[16 January 2021] Our work on deep extreme classification with label graphs, that improves upon our previous work that uses label features alone, will appear at The Web Conference 2021 (formerly known as WWW). Code for our method and a preprint are available [here].
[04 January 2021] Our work on robust calibration of low-cost air quality monitoring sensors will appear in the journal Atmospheric Measurement Techniques. Code for our method and a link to the paper are available [here].
[16 December 2020] I delivered a keynote on deep learning for extreme-classification at the Amazon Research Days 2020 event. Slides for my talk are available [here].
[16 October 2020] Our work on deep extreme classification with label features with applications to large-scale product-to-product recommendation will appear at the WSDM 2021 conference. Code for our method and a preprint are linked [here].
[02 August 2020] Our work on Bayesian optimization techniques for designing effective lockdown and other non-pharmaceutical interventions to check the spread of pandemics such as CoViD-19 will appear in the Transactions of the INAE. Code for our method and a preprint are available [here].
[27 May 2020] Our work on accelerated program repair with applications to AI-based e-tutors for introductory programming courses will appear at AIED 2020. Code for our method and a preprint are available [here].
[28 June 2019] Congratulations to Neeraj Kumar for winning the Best Software Award at the 52nd IITK Convocation for thesis work developing the SphinX pegagogy helper system for conducting large courses.
[28 June 2019] Congratulations to Ankit Jalan for winning the Cadence Silver Medal at the 52nd IITK Convocation for thesis work on advances in dimensionality reduction in extreme classification.
[28 June 2019] Congratulations to Sayash Kapoor for winning the Motorola Gold Medal at the 52nd IITK Convocation for all-round performance including research advances on robust multi-armed bandits.