Q-learning

A reinforcement learning method that learns the expected value of each action in each state, so an agent can choose the action with the best long-term reward.

The “Q” stands for quality: Q(state, action) estimates future reward. The agent updates these estimates from experience using the Bellman equation. DeepMind’s Deep Q-Network (DQN) combined Q-learning with deep neural networks and learned to play Atari games from pixels.

See the DQN paper explained.

#Glossary #Reinforcement Learning