Reinforcement Learning Redmond

Reinforcement Learning Redmond

Projects

Overview

The Reinforcement Learning group in Redmond is passionate about both RL theory and applications. The group focuses on foundational advances in Model-based Learning, Memory, Effective Exploration, and Counterfactual Learning. Our application domains include Interactive Fiction (Text-based) games, Xbox and Microsoft Studios games, and indoor agriculture.

We collaborate closely with the wider reinforcement learning effort at MSR including groups in Cambridge, Montreal, and New York.

A recent overview of our research: BreakthroughsRL2019.pptx

Representative publications:

People

Core Group Members

Affiliates

Open Source

Jericho – A fast learning environment for Agents to play Interactive Fiction games.
NAIL – A general Interactive Fiction game playing agent. Winner of the 2018 CIG Text Adventure AI Competition.
Frigatebird – Enabling unmanned aerial vehicles to travel long distances by soaring.