BetaPolicyOptions<T>
Options & Configuration · AiDotNet.ReinforcementLearning.Policies
Configuration options for Beta distribution policies: BetaPolicy’s
#ctor constructor builds its network from them.
Inheritance: ModelOptions → BetaPolicyOptions<T>
For Beginners
Use this when every action has a hard minimum and maximum (a throttle from 0 to 1, a valve from closed to open). The network predicts two shape numbers per action, and the policy draws actions that always land inside the allowed range.
How It Works
A Beta policy (Chou, Maturana and Scherer, 2017) samples each action dimension from a Beta(α, β) distribution on
[0, 1] and rescales it to [ActionMin, ActionMax]. Unlike a clipped Gaussian it puts no
probability mass outside the bounds, so bounded controls need no clipping bias.
Constructors
| Constructor | Summary |
|---|---|
BetaPolicyOptions() |
Properties
| Property | Summary |
|---|---|
IExplorationStrategy<T> ExplorationStrategy { get; set; } |
Gets or sets the exploration applied to sampled actions during training. |
ILossFunction<T> LossFunction { get; set; } |
Gets or sets the loss the policy network is trained with when it is fitted directly. |
double ActionMax { get; set; } |
Gets or sets the upper bound of every action dimension. |
double ActionMin { get; set; } |
Gets or sets the lower bound of every action dimension. |
int ActionSize { get; set; } |
Gets or sets the number of action dimensions. |
int StateSize { get; set; } |
Gets or sets the number of features in an observation (the network’s input width). |
int[] HiddenLayers { get; set; } |
Gets or sets the width of each hidden layer of the policy network, in order. |