Workshop on developing Reinforcement Learning
summary
Workshop walks through concrete code and outcomes: Q‑learning SARSA DQN implemented from scratch, then gymnasium environments (grid world, FrozenLake, pendulum, cartpole) and stable‑baselines and machine used for continuous tasks. Algo‑trading demo uses S&P 2010–2019, 80/20 split, four features, and shows a short run 14,000 paper profit on training but is unstable and omits transaction costs and risk management. Hyperparameter tuning and heavy compute turn out to be the real work.
AI-written summary · about
chapters (16)
- 0:00Reinforcement learning intro
- 1:07Fundamentals of reinforcement learning
- 9:42Model free vs model based RL
- 15:26Q learning and TD methods
- 20:00Deep reinforcement learning with neural networks
- 25:25Policy based RL architectures
- 31:31RL frameworks and tools
- 34:18Gymnasium for simulation
- 35:44Reinforcement learning in finance applications
- 42:01Key conclusions on reinforcement learning for finance
- 50:09Demo: Grid world Q learning
- 53:20Demo: Frozen lake comparisons
- 1:01:03Demo: Pendulum DDPG SAC
- 1:13:45Demo: Hyperparameter tuning
- 1:22:20Demo: Algo trading system
- 1:25:20Q&A
related talks
01h 31mReinforcement Learning for Finance
02h 42m[Full Workshop] Reinforcement Learning, Kernels, Reasoning, Quantization & Agents
56mAn Introduction to Reinforcement Learning - Snake Your Way Out of a Paper Bag
01h 21mIntroduction to reinforcement learning using Python & OpenAI Gym
28mDeep Reinforcement Learning: theory, intuition, code
01h 17mTraining intelligent game agents using deep reinforcement learning
33mBreakthroughs and the Future of (Deep) Reinforcement Learning
26mHacking Reinforcement Learning