ContinuousPolicy<T>

Models & Types · AiDotNet.ReinforcementLearning.Policies

Policy for continuous action spaces using a neural network to output Gaussian parameters.

Inheritance: ModelBase<T, Vector<T>, Vector<T>> → PolicyBase<T> → ContinuousPolicy<T>

Implements: IGeneratedParameterRegistrar<T>

Example

// Create a Gaussian continuous policy for unbounded continuous actions
var network = new NeuralNetwork<double>();
var policy = new ContinuousPolicy<double>(network, actionSize: 2, new EpsilonGreedyExploration<double>());
            
// Sample an action from the Gaussian distribution
var state = new Vector<double>(new double[] { 0.5, -0.3, 1.0, 0.2 });
var action = policy.SelectAction(state, training: true);

Constructors

Constructor Summary
ContinuousPolicy() Initializes a new instance with default settings.
ContinuousPolicy(ContinuousPolicyOptions<T> options) Builds the policy from options: a dense ReLU network of the options’ hidden widths from StateSize inputs to a mean and a log standard deviation for each action dimension (2 × ActionSize outputs).
ContinuousPolicy(INeuralNetwork<T> policyNetwork, ContinuousPolicyOptions<T> options) Initializes a new instance of the ContinuousPolicy class from its options: action size, exploration strategy, tanh squashing and Seed.
ContinuousPolicy(INeuralNetwork<T> policyNetwork, int actionSize, IExplorationStrategy<T> explorationStrategy, bool useTanhSquashing, Random random)

Methods

Method Summary
IReadOnlyList<INeuralNetwork<T>> GetNetworks()
T ComputeLogProb(Vector<T> state, Vector<T> action)
Vector<T> SelectAction(Vector<T> state, bool training)
void Reset()