Prashanth L.A.
Associate Professor
SSB 314,
Computer Science and Engineering,
Indian Institute of Technology Madras
Chennai 600036
Email: prashla AT cse.iitm.ac.in
Tel: +91-44-22574377
Research
My research focuses on reinforcement learning (RL), multi-armed bandits, and zeroth-order stochastic optimization.
Current interests are in the development of algorithms and derivation of theoretical guarantees for risk-sensitive learning (RL and bandits), policy-gradient methods, and estimation/optimization of risk measures with sampling.
Selected Publications
Actor-Critic Algorithms for Risk-Sensitive MDPs, Neural Information Processing Systems, 2013 (full oral presentation).
Cumulative Prospect Theory Meets Reinforcement Learning: Prediction and Control, ICML 2016; related journal version, IEEE Transactions on Automatic Control, 2018.
A Wasserstein Distance Approach for Concentration of Empirical Risk Estimates, Journal of Machine Learning Research, 2022.
Generalized Simultaneous Perturbation-Based Gradient Search With Reduced Estimator Bias, IEEE Transactions on Automatic Control, 2025. Preprint.
Preference-centric Bandits: Optimality of Mixtures and Regret-efficient Algorithms, IEEE Transactions on Information Theory, 2026 (accepted). Preprint.
Reinforcement Learning for Exponential Utility: Algorithms and Convergence in Discounted MDPs, NeurIPS 2026 (accepted).
Books and Monographs
Gradient-Based Algorithms for Zeroth-Order Optimization (2025), with Shalabh Bhatnagar. Foundations and Trends in Optimization.
Online book / Author copy / CourseRisk-Sensitive Reinforcement Learning via Policy Gradient Search (2022), with Michael C. Fu. Foundations and Trends in Machine Learning.
Published copy / TutorialStochastic Recursive Algorithms for Optimization: Simultaneous Perturbation Methods (2013), with Shalabh Bhatnagar and H. L. Prasad. Springer.
Sample chapter / Tutorial
Click here for more book resources
Recent News
Oct-2026: Recognized as a Top Area Chair at NeurIPS-2026.
Oct-2026: A paper entitled Preference-centric Bandits: Optimality of Mixtures and Regret-efficient Algorithms accepted for publication in IEEE Transactions on Information Theory.
Sep-2026: A paper entitled Reinforcement Learning for Exponential Utility: Algorithms and Convergence in Discounted MDPs accepted for publication in NeurIPS (2026).
Sep-2026: Serving as an area chair for ICLR-2027 and on the senior program committee of AAAI-2027.
May-2026: Serving as an area chair for NeurIPS-2026.
Feb-2026: Serving as an area chair for ICML-2026 and RLC-2026.
Jan-2026: Teaching a course on topics in RL. For details, click here.
Nov-2025: A paper entitled Policy Newton methods for Distortion Riskmetrics accepted for publication in AAAI (2026).
Professional Service
Area Chair: ICLR 2027; NeurIPS 2026, ICML 2026, and RLC 2026.
Senior Program Committee: AAAI 2027.
Further service and reviewing recognitions
Prospective interns
I do not have open positions and you are encouraged to apply directly for the IITM summer fellowship programme. Details here.