Sanmit Narvekar
sanmit at cs dot utexas dot edu
Office: GDC 3.424E

I defended my PhD from the Learning Agents Research Group at the Department of Computer Science at the University of Texas at Austin, where I was advised by Peter Stone. My dissertation research focused on curriculum learning -- the automated design of a sequence of training tasks that enable autonomous agents to learn faster or better. I was also a member of the UT Austin Villa Standard Platform League team, where I worked primarily on the vision system.

I received my BS in computer science from California State University, Los Angeles, where I worked on using probabilistic automata (e.g. HMMs, etc.) for anomaly detection in computer agents with Valentino Crespi. I also worked part time at the Jet Propulsion Laboratory, where I used multispectral and synthetic aperture radar data to detect natural disasters, and interned at Google with Eugene Ie and Craig Boutilier to apply RL to recommendation systems.

I am currently a Research Scientist at Waymo, where I work on ML models for behavior prediction in self-driving cars.

PhD Thesis  /  Google Scholar

Teaching

Fall 2015: TA for CS 344M: Autonomous Multiagent Systems
Fall 2016: TA for CS 394R: Reinforcement Learning: Theory and Practice

Publications

Preprints

RecSim: A Configurable Simulation Platform for Recommender Systems
Eugene Ie, Chih-wei Hsu, Martin Mladenov, Vihan Jain, Sanmit Narvekar, Jing Wang, Rui Wu, and Craig Boutilier
arXiv preprint arXiv:1909.04847, September 2019
[pdf] [bib] [code]


Journals

Curriculum Learning for Reinforcement Learning Domains: A Framework and Survey
Sanmit Narvekar, Bei Peng, Matteo Leonetti, Jivko Sinapov, Matthew E. Taylor, and Peter Stone
Journal of Machine Learning Research, 21(181):1-50, 2020
[pdf] [bib]


Refereed Conferences

Capturing Skill State in Curriculum Learning for Human Skill Acquisition.
Keya Ghonasgi, Reuth Mirsky, Sanmit Narvekar, Bharath Masetty, Adrian M. Haith, Peter Stone, and Ashish D. Deshpande
Proceedings of the International Conference on Intelligent Robots and Systems (IROS), Prague, Czech Republic, Sep 27 - Oct 1, 2021
[pdf] [bib]

SlateQ: A Tractable Decomposition for Reinforcement Learning with Recommendation Sets
Eugene Ie, Vihan Jain, Jing Wang, Sanmit Narvekar, Ritesh Agarwal, Rui Wu, Heng-Tze Cheng, Tushar Chandra, and Craig Boutilier
Proceedings of the 28th International Joint Conference on Artificial Intelligence (IJCAI), Macao, China, August 10-16, 2019
[pdf] [bib] [extended]

Learning Curriculum Policies for Reinforcement Learning
Sanmit Narvekar and Peter Stone
Proceedings of the 18th International Conference on Autonomous Agents and Multi-Agent Systems (AAMAS), Montreal, Canada, May 13-17, 2019
(An earlier version of this work appeared at the Continual Learning Workshop at NeurIPS 2018)
[pdf] [bib] [slides]

Autonomous Task Sequencing for Customized Curriculum Design in Reinforcement Learning
Sanmit Narvekar, Jivko Sinapov, and Peter Stone
Proceedings of the 26th International Joint Conference on Artificial Intelligence (IJCAI), Melbourne, Australia, August 19-25, 2017
(An earlier version of this work appeared at RLDM 2017)
[pdf] [bib] [slides]

Source Task Creation for Curriculum Learning
Sanmit Narvekar, Jivko Sinapov, Matteo Leonetti, and Peter Stone
Proceedings of the 15th International Conference on Autonomous Agents and Multi-Agent Systems (AAMAS), Singapore, May 9-13, 2016
[pdf] [bib] [slides]

Learning Inter-Task Transferability in the Absence of Target Task Samples
Jivko Sinapov, Sanmit Narvekar, Matteo Leonetti, and Peter Stone
Proceedings of the 14th International Conference on Autonomous Agents and Multi-Agent Systems (AAMAS), Istanbul, Turkey, May 4-8, 2015
[pdf] [bib]


Refereed Workshops and Symposia

Leveraging Reinforcement Learning for Human Motor Skill Acquisition
Keya Ghonasgi, Reuth Mirsky, Bharath Masetty, Sanmit Narvekar, Adrian Haith, Peter Stone, and Ashish Deshpande
Social AI for Human-Robot Interactions of Human-Care Service Robots Workshop at the International Conference on Intelligent Robots and Systems (IROS), Las Vegas, Nevada, October 25-29, 2020
[pdf] [bib]

Generalizing Curricula for Reinforcement Learning
Sanmit Narvekar and Peter Stone
4th Lifelong Learning Workshop at the International Conference on Machine Learning (ICML), Vienna, Austria, July 12-18, 2020
[pdf] [bib] [slides] [video]


Book Chapters

Towards a real-time, low-resource, end-to-end object detection pipeline for robot soccer.
Sai Kiran Narayanaswami, Mauricio Tec, Ishan Durugkar, Siddharth Desai, Bharath Masetty, Sanmit Narvekar, and Peter Stone
RoboCup 2022: Robot Soccer World Cup XXV, Lecture Notes in Artificial Intelligence, Springer, 2022
[pdf] [bib]

Fast and Precise Black and White Ball Detection for RoboCup Soccer
Jacob Menashe, Josh Kelle, Katie Genter, Josiah Hanna, Elad Liebman, Sanmit Narvekar, Ruohan Zhang, and Peter Stone
RoboCup 2017: Robot Soccer World Cup XXI, Lecture Notes in Artificial Intelligence, Springer, 2017
[pdf] [bib] [slides]


Magazines and Other

UT Austin Villa: Project-Driven Research in AI and Robotics
Katie Genter, Patrick MacAlpine, Jacob Menashe, Josiah Hanna, Elad Liebman, Sanmit Narvekar, Rouhan Zhang, and Peter Stone
IEEE Intelligent Systems, Expert Opinion, March 2016

Website template credits