Reinforcement learning

Reinforcement Learning is a subfield of Artificial Intelligence (AI) that focuses on how software agents should take actions in an environment to maximize a notion of cumulative reward. It is one of three basic machine learning paradigms, alongside Supervised Learning and Unsupervised Learning.

Overview

Reinforcement Learning differs from standard Supervised Learning in that correct input/output pairs are never presented, and sub-optimal actions are not explicitly corrected. Instead, the focus is on finding a balance between exploration (of uncharted territory) and exploitation (of current knowledge).

The environment is typically stated in the form of a Markov Decision Process (MDP), as many reinforcement learning algorithms for this context utilize dynamic programming techniques. The main difference between the classical techniques and reinforcement learning algorithms is that the latter do not need knowledge about the MDP and they target large state spaces.

Elements of Reinforcement Learning

Reinforcement learning involves:

• A set of environment and agent states, S.
• A set of actions, A, of the agent.
• Rules of transitioning between states.
• Rules that determine the scalar immediate reward of a transition.
• Rules that describe what the agent observes.

Algorithms

Reinforcement learning algorithms include the Q-Learning, Deep Q Network (DQN), and Deep Deterministic Policy Gradient (DDPG) among others. These algorithms guide the agent to learn from the consequences of actions in order to maximize its reward.

Applications

Reinforcement learning can be used in various applications including robotics, game playing, resource management, and many more. It has been used to make significant breakthroughs in AI, such as AlphaGo, the first computer program to defeat a world champion in the complex game of Go.

See Also

• Machine Learning
• Artificial Intelligence
• Deep Learning
• Neural Networks

References

‎

‎