Dileep Kalathil

About

Dileep Kalathil is a leading researcher in reinforcement learning (RL), specializing in tackling the fundamental challenge of learning from sparse rewards. His work bridges the gap between theoretical algorithms and practical deployment, particularly in environments where feedback is limited or delayed. Kalathil's major contributions include developing methods that leverage offline demonstrations to guide RL agents, enabling efficient learning even when reward signals are minimal. His 2022 paper, "Reinforcement Learning with Sparse Rewards using Guidance from Offline Demonstration," has garnered 26 citations, highlighting its impact on the field. He has also advanced meta-RL by integrating demonstration data to accelerate adaptation in sparse-reward settings, and pioneered federated offline RL, where distributed agents collaboratively learn policies from small, heterogeneous datasets without sharing raw data. This work is critical for privacy-sensitive applications like healthcare and robotics. Kalathil's research is notable for its practical focus on data efficiency and real-world constraints, making him a key figure in the next generation of RL innovation.

Research Focus

Key Achievements

Top Papers

1. 1
   Reinforcement Learning with Sparse Rewards using Guidance from Offline Demonstration

   26 citations · 2022
2. 2
   Enhanced Meta Reinforcement Learning using Demonstrations in Sparse Reward Environments

   2 citations · 2022
3. 3
   Federated Ensemble-Directed Offline Reinforcement Learning

   2 citations · 2023

Key Collaborators

Contact & Links