DQN class

DQN wrapper with optional target network and configurable replay buffer

Enhancements:

• optional target network with periodic hard updates
• configurable discount (gamma) and batch size
• simple (FIFO) replay buffer with capacity
• serialization to/from Map/JSON

Constructors

DQN({required int nActions, required List<int> netLayers, int memoryCapacity = 1000, double epsilon = 0.1, double gamma = 0.99, int targetUpdateSteps = 100, bool useTarget = true, int? seed})

Properties

epsilon ↔ double

getter/setter pair

gamma ↔ double

getter/setter pair

hashCode → int

The hash code for this object.

no setter, inherited

memoryCapacity → int

final

nActions → int

final

network → ANN

final

runtimeType → Type

A representation of the runtime type of the object.

no setter, inherited

targetUpdateSteps → int

final

Methods

chooseAction(List<double> state) → int

noSuchMethod(Invocation invocation) → dynamic

Invoked when a nonexistent method or property is accessed.

inherited

remember(List<double> state, int action, double reward, List<double> nextState) → void

replay({int batchSize = 16}) → void

toJson() → String

toMap() → Map<String, dynamic>

toString() → String

A string representation of this object.

inherited

Operators

operator ==(Object other) → bool

The equality operator.

inherited

Static Methods

fromJson(String s, {int? seed}) → DQN

fromMap(Map<String, dynamic> m, {int? seed}) → DQN