DQNAgent<T>
Models & Types · AiDotNet.ReinforcementLearning.Agents.DQN
Deep Q-Network (DQN) agent for reinforcement learning.
Inheritance: ReinforcementLearningAgentBase<T> → DeepReinforcementLearningAgentBase<T> → DQNAgent<T>
Implements: IGeneratedParameterRegistrar<T>, IGradientComputable<T, Vector<T>, Vector<T>>
For Beginners
DQN learns to play games (or solve problems) by learning how valuable each action is in each situation. It uses a neural network to estimate these “Q-values” - essentially, expected future rewards.
The agent:
- Sees the current state (like game screen)
- Evaluates each possible action using its Q-network
- Picks the action with highest Q-value (with some random exploration)
- Learns from past experiences stored in memory
Famous for: Learning to play Atari games from pixels (DeepMind, 2015)
How It Works
DQN is a landmark algorithm that combined Q-learning with deep neural networks, enabling RL to scale to high-dimensional state spaces. It introduced two key innovations:
- Experience Replay: Breaks temporal correlations by training on random past experiences
- Target Network: Provides stable Q-value targets by using a slowly-updating copy
Reference: Mnih, V., et al. (2015). “Human-level control through deep reinforcement learning.” Nature.
Example
// Create a Deep Q-Network agent for discrete action spaces
var options = new DQNOptions<double> { StateSize = 4, ActionSize = 2, LearningRate = 0.001 };
var agent = new DQNAgent<double>(options);
// Select an action for the current game state
var state = new Vector<double>(new double[] { 0.5, -0.3, 1.0, 0.2 });
var action = agent.SelectAction(state);
Constructors
| Constructor | Summary |
|---|---|
DQNAgent() |
Initializes a new instance with default settings. |
DQNAgent(DQNOptions<T> options) |
Initializes a new instance of the DQNAgent class. |
Properties
| Property | Summary |
|---|---|
int FeatureCount { get; } |
Methods
| Method | Summary |
|---|---|
Dictionary<string, T> GetMetrics() |
|
ModelMetadata<T> GetModelMetadata() |
|
ModelOptions GetOptions() |
|
T Train() |
|
Vector<T> ComputeGradients(Vector<T> input, Vector<T> target, ILossFunction<T> lossFunction) |
Computes gradients of the loss with respect to this agent’s parameters, without updating them. |
Vector<T> SelectAction(Vector<T> state, bool training) |
|
byte[] Serialize() |
Auto-generated common model serialization surface. |
void ApplyGradients(Vector<T> gradients, T learningRate) |
|
void Deserialize(byte[] data) |
Auto-generated common model deserialization surface. |
void LoadModel(string filepath) |
|
void SaveModel(string filepath) |
|
void StoreExperience(Vector<T> state, Vector<T> action, T reward, Vector<T> nextState, bool done) |