4 reports · 2013–2017

Reinforcement learning

Agents that learn from reward — value-based methods, policy gradients and self-play.

Fig. 02 / Field notesLearn by acting
An agent follows a route through a grid world; actions and observations connect the policy to its environment.
An agent connects observations, decisions and feedback.

Playing Atari with Deep Reinforcement Learning

Reinforcement learning · 7 min read · Intermediate

The Deep Q-Network learned Atari games at superhuman level from raw pixels with experience replay — deep reinforcement learning's founding result.

reinforcement-learningdqndeep-learning Read

Mastering the Game of Go without Human Knowledge

Reinforcement learning · 8 min read · Advanced

AlphaGo Zero mastered Go from pure self-play with no human games, beating the version that beat Lee Sedol — a milestone for reinforcement learning.

alphagoreinforcement-learningself-play Read

PPO: Proximal Policy Optimization

Reinforcement learning · 5 min read · Intermediate

PPO stabilizes policy gradients with a clipped objective, becoming the default RL algorithm behind robotics, games and RLHF fine-tuning.

ppopolicy-gradientreinforcement-learning Read