Pinned Loading

1. Federated-MARL-Gym-Environment Federated-MARL-Gym-Environment Public

   𝙵𝚎𝚍𝚎𝚛𝚊𝚝𝚎𝚍 𝙼𝙰𝚁𝙻-𝙶𝚢𝚖 : We introduce a custom multi-agent reinforcement learning environment built with Gymnasium and Pygame, designed for evaluating federated RL (FRL) algorithms. The environment mod…

   Jupyter Notebook

2. Robust-Asynchronous-Q-Learning-with-Markovian-Data Robust-Asynchronous-Q-Learning-with-Markovian-Data Public

   Accepted at NeuRIPS 2025-Reliable ML Workshop🎉. 𝚁𝚘𝚋𝚞𝚜𝚝 𝙰𝚜𝚢𝚗𝚌-𝚀/𝙰𝚜𝚢𝚗𝚌-𝚁𝙰𝚀/𝙼: The first provably robust variants of asynchronous Q-learning that tolerates adversarially corrupted rewards. Our algorit…

   1

3. Adversarially-Robust-TD-Learning-with-Markovian-Data Adversarially-Robust-TD-Learning-with-Markovian-Data Public

   Accepted at AISTATS 25🎉. Policy evaluation estimates long-term returns in RL. Temporal Difference (TD) learning, a classic method with finite-time guarantees, assumes well-behaved rewards. But what…

   MATLAB

4. Robust-Q-Learning-under-Corrupted-Rewards Robust-Q-Learning-under-Corrupted-Rewards Public

   Accepted at IEEE CDC 24 🎉. We analyze Q-learning's robustness against strong-contamination attacks that disrupt reward signals. Our robust Q-learning algorithm employs historical data to create res…

   Jupyter Notebook 1 1

5. Robust-Federated-Q-Learning-with-Almost-No-communication Robust-Federated-Q-Learning-with-Almost-No-communication Public

   𝚁𝚘𝚋𝚞𝚜𝚝 𝙵𝚎𝚍-𝚀 : A federated Q-learning algorithm that stays reliable even when a small fraction of agents are adversarial. It blends model-based/model-free updates with median-of-means to ensure (i)…

   Jupyter Notebook

6. Q-Learning-over-Static-and-Time-Varying-Networks Q-Learning-over-Static-and-Time-Varying-Networks Public

   𝚅𝚁𝙳𝚀 : We propose and analyze a new algorithm that achieves collaborative speedups in sample complexity for Q-learning over static and time-varying networks.

   Jupyter Notebook 1