DynaQOptions<T>
Options & Configuration · AiDotNet.Models.Options
Configuration options for Dyna-Q reinforcement learning agents.
Inheritance: ModelOptions → ReinforcementLearningOptions<T> → DynaQOptions<T>
For Beginners
Dyna-Q is like an agent that not only learns from real experience but also “imagines” additional experiences using a model of the environment. The planning steps parameter controls how many imagined experiences the agent uses per real step.
How It Works
Dyna-Q combines model-free Q-learning with a learned environment model to perform additional planning steps, accelerating learning from limited real experience.
Constructors
| Constructor | Summary |
|---|---|
DynaQOptions() |
Initializes a new instance with default values. |
DynaQOptions(DynaQOptions<T> other) |
Initializes a new instance by copying values from another instance. |
Properties
| Property | Summary |
|---|---|
int ActionSize { get; set; } |
Number of discrete actions available to the agent. |
int PlanningSteps { get; set; } |
Number of planning steps per real environment step. |
int StateSize { get; set; } |
Dimension of the environment state vector. |