DQN
A deep Q-network (DQN) is a reinforcement-learning method that uses a neural network to approximate the state-action value function in Q-learning.
Applied Reinforcement Learning
Deep Q-network (DQN) is a reinforcement-learning algorithm that uses a neural network to approximate action values for Q-learning.
- How it is used
- Uses DQN for sequential decisions under uncertainty when tabular values cannot feasibly represent the state space.
- How deep it goes
- Develops the instability rationale, stabilization mechanisms, and a worked implementation with network updates and a training loop.
- Examples
- Works through simulated radiotherapy planning and identifies inventory ordering, energy dispatch, dynamic pricing, trading, and recommendation applications.
- What you can do after
- Builds a DQN agent from scratch for a simulated treatment environment, including replay, action selection, network updates, and a curriculum.
Evolutionary Deep Learning: Genetic algorithms and neural networks
Deep Q-learning is reinforcement learning that uses a deep neural network to approximate action-value estimates instead of storing them only in a Q-table.
- How it is used
- Applies deep learning to reinforcement-learning environments where neural networks replace Q-tables or policies.
- How deep it goes
- Explains the Q-learning target, network architecture, exploration, replay, and training loop, but does not derive the method mathematically.
- Examples
- Uses CartPole, Acrobot, MountainCar, LunarLander, Atari games, Gym notebooks, training plots, and replay-code exercises.
- What you can do after
- Build it from scratch in Keras, train it on Gym environments, inspect rewards and losses, and experiment with environments and hyperparameters.
Read how Evolutionary Deep Learning: Genetic algorithms and neural networks explains DQN →
Also covered in
What to know first
- actionAn action is a choice or operation selected or performed by an actor that can affect a system, environment, or process.
- stateState is information describing the condition or configuration of a system or computation at a particular time that may influence its behavior or subsequent transitions.
- reward signalA reward signal is a value or feedback indicating the desirability, benefit, or cost associated with an action, outcome, or behavior.
- convolutional neural networkA convolutional neural network is a neural-network architecture that processes structured inputs with convolutional operations, typically using learned local filters to capture patterns.
Related concepts
- adamuses · Adam is an adaptive stochastic-gradient optimization algorithm that updates parameters using exponential moving estimates of the first and second moments of gradients.
- deep neural networkuses · A deep neural network is a neural-network model with multiple layers of computational units that learn representations from data.
- experience replayuses · Experience replay is a reinforcement-learning technique in which an agent trains on batches sampled from stored interaction experiences.
- mean squared erroruses · Mean squared error is the arithmetic mean of the squared differences between corresponding predicted and reference numerical values.
- off-policy methodkind of · An off-policy method learns or evaluates a target policy using experience generated by a different behavior policy.
Questions about DQN
What is DQN?
A deep Q-network (DQN) is a reinforcement-learning method that uses a neural network to approximate the state-action value function in Q-learning.
Which Manning books explain DQN?
Applied Reinforcement Learning and Evolutionary Deep Learning: Genetic algorithms and neural networks.
What should I know before learning DQN?
Manning books that teach DQN usually introduce action, state, reward signal and convolutional neural network first.
What is the difference between DQN and continuous action space?
A deep Q-network (DQN) is a reinforcement-learning method that uses a neural network to approximate the state-action value function in Q-learning. A continuous action space is an action space in which actions can take real-valued values within specified bounds.
What is DQN a kind of?
DQN is a kind of off-policy method. An off-policy method learns or evaluates a target policy using experience generated by a different behavior policy.
What is DQN also called?
DQN is also called deep Q-learning or deep q-network.
These guides are generated from the text of Manning books by Manning's knowledge graph and link to the sections they draw on. The full books are available on liveBook and at manning.com.

