RL Bootcamp 2025 - Day 1

Reinforcement learning (RL) remains a driving force behind modern AI—powering robotics, autonomous systems, game-playing, and complex decision-making. Mastering RL opens doors across research and industry, from reliable control to scalable, data-efficient learning. Yet real-world RL still faces hurdles: framing problems as meaningful MDPs, stabilizing training, evaluating and reproducing results, choosing the right algorithms, and deploying to hardware with safety and efficiency. The 2025 Reinforcement Learning Bootcamp is designed to lower these barriers and turn RL into a practical tool you can use with confidence. What’s new in 2025 (Sept 17–19): Keynotes (speakers confirmed): • Prof. Sergey Levine (UC Berkeley, online) — Recent Advances in Deep Reinforcement Learning • Prof. Peter Auer (University of Leoben) — Multi-Armed Bandits and Exploration Strategies • Dr. Samuele Tosatto (University of Innsbruck) — Where are all the intelligent robots? A quest for efficiency in reinforcement learning Core lectures: Intro to RL & Fundamentals; deep dives on Policy Gradients & Actor-Critics. Hands-on track: A guided, end-to-end “Implementing PPO from Scratch” workshop (from setup to training and evaluation). Advanced & applied sessions: World models, model-based RL, multi-agent RL, and advanced policy optimization (PPO/TRPO/SAC). Community & practice: Networking, group photo, and a closing Salzburg visit. Across three full days, you’ll blend theory with practice—configuring modern RL environments, implementing baselines, stress-testing stability and reproducibility, and discussing evaluation, deployment, and safety. Whether you’re a student, researcher, or practitioner, you’ll leave with practical insights, working code, and a clearer path to real-world RL impact. Note: The schedule is tentative and subject to change as we finalize workshop details; speakers are confirmed.