GradientBanditOptions<T>

Options & Configuration · AiDotNet.Models.Options

Configuration options for the gradient bandit, which learns action preferences with a softmax policy.

Inheritance: ModelOptions → ReinforcementLearningOptions<T> → GradientBanditOptions<T>

Constructors

Constructor Summary
GradientBanditOptions()

Properties

Property Summary
bool UseBaseline { get; set; } Whether to subtract the average reward as a baseline from each reward before updating preferences.
double Alpha { get; set; } Step size for updating the action preferences.
int NumArms { get; set; } Number of arms (actions) the bandit chooses among.