One-shot Clustering for Policy Evaluation in Federated TD Learning with LFA
Analysed a clustering algorithm for TD(0) learning in federated reinforcement learning.
anshu-arora-pranay-sharma-avishek-ghosh
Professional Summary
I am interested in machine learning and robotics, with a focus on reinforcement learning, optimization, and data-efficient learning algorithms for embodied agents. My goal is to develop systems that can generalize, adapt, and make effective decisions under real-world constraints. I am particularly excited by problems at the intersection of learning, control, and autonomous interaction.
Education
MS Electrical Engineering
2026-09
2028-06
Stanford University
B.Tech Electrical Engineering
2022-10
2026-05
IIT Bombay
Interests
Reinforcement Learning
Optimization
Learning algorithms
Signal Processing
Computational Neuroscience
My Research
My research interest lies at the intersection of machine learning, optimization and robotics. I am broadly interested in developing efficient and adaptive learning algorithms that are both theoretically grounded and practically scalable, particularly for decision-making and autonomous systems operating in real-world environments.
I have worked on deriving theoretical guarantees for clustering in federated reinforcement learning, studied policy optimization methods such as TRPO and PPO, and analyzed backdoor attacks in federated learning to better understand robust distributed training. In signal processing, I have designed algorithms for sub-Nyquist sampling and perfect signal reconstruction using adaptive time encoding frameworks.
Looking forward, I hope to work on problems that combine rigorous mathematical foundations with real-world implementation. I am especially excited by the challenge of building intelligent systems that can learn efficiently, adapt reliably, and operate robustly in dynamic environments.
Featured Publications
•
•
1 min read
Backdoor Attacks on Federated Learning
Literature review on advancements in robust federated machine learning.
anshu-arora-malay-kedia-pranay-sharma
Variable-Bias Integrate-and-Fire Time Encoding Machines
Adaptive time-encoding framework for efficient signal sampling and reconstruction.
anshu-arora-satish-mulleti
Technical Design Report of Matsya 6D, Autonomous Underwater Vehicle
Insights into working of software, electrical and mechanical subsystems to create Matsya 6D.
co-author
Recent Publications
(2026).
Tracing the Roots: Trust Region Policy Optimization.
Unpublished.
(2026).
Adaptive Non-Uniform Sampling of Bandlimited Signals via Algorithm–Encoder Co-Design.
Under Review.
(2025).
One-shot Clustering for Policy Evaluation in Federated TD Learning with LFA.
Unpublished.
(2025).
Backdoor Attacks on Federated Learning.
Unpublished.
(2025).
GenAI and PhD.
Unpublished.