ExplorationScheduleConfig<T>
Options & Configuration · AiDotNet.Configuration
Configuration for exploration schedule (epsilon decay for epsilon-greedy).
How It Works
For Beginners: The agent needs to explore early in training (try random actions) but exploit more later (use learned policy). This schedule controls that transition.
Constructors
| Constructor | Summary |
|---|---|
ExplorationScheduleConfig() |
Creates a new instance with default values. |
Properties
| Property | Summary |
|---|---|
ExplorationDecayType DecayType { get; set; } |
Type of decay schedule. |
T FinalEpsilon { get; set; } |
Final exploration rate (0.01 = mostly learned policy). |
T InitialEpsilon { get; set; } |
Initial exploration rate (1.0 = fully random). |
int DecaySteps { get; set; } |
Number of steps over which to decay from initial to final epsilon. |