# What is DQN? Definition and meaning

A reinforcement-learning algorithm that approximates the action-value function with a deep network, using replay memory and a target network.

## Definition

**DQN**: A reinforcement-learning algorithm that approximates the action-value function with a deep network, using replay memory and a target network.

In short: Deep Q-network: learning action values with a neural network.

## When it is used

Discrete-action control problems such as dispatching and scheduling decisions.

## Learn more

The full guide to **[DQN](/tools/dqn)** explains how it works, when to use it and how it compares with the alternatives, with examples. The [glossary](/glossary) lists every term on the site from A to Z.

Previous term: [DPO](/glossary/dpo) · Next term: [Dropout](/glossary/dropout)
