Class RLAutoMLOptions<T>

Namespace

AiDotNet.Configuration

Assembly

AiDotNet.dll

Configuration options for running AutoML over reinforcement learning agents and hyperparameters.

public class RLAutoMLOptions<T>

Type Parameters

T

The numeric type used for calculations (e.g., float, double).

Inheritance

object

RLAutoMLOptions<T>

Inherited Members

object.Equals(object)

object.Equals(object, object)

object.GetHashCode()

object.GetType()

object.MemberwiseClone()

object.ReferenceEquals(object, object)

object.ToString()

Remarks

This options class is designed for use with AiModelBuilder when training RL agents. It supports the AiDotNet facade pattern by providing sensible defaults while allowing customization.

For Beginners: RL AutoML tries a few different RL agent settings, measures which one earns the most reward, and then trains the best configuration for your full training budget.

Properties

CandidateAgents

Gets or sets the allowed agent types for RL AutoML.

public List<RLAutoMLAgentType>? CandidateAgents { get; set; }

Property Value

List<RLAutoMLAgentType>

Remarks

If null or empty, AiDotNet selects a default set based on whether the environment action space is discrete or continuous.

EvaluationEpisodesPerTrial

Gets or sets the number of evaluation episodes to run per AutoML trial (no learning).

public int EvaluationEpisodesPerTrial { get; set; }

Property Value

int

Remarks

For Beginners: After a trial is trained, the agent is evaluated without exploration to measure performance.

MaxStepsPerEpisodeOverride

Gets or sets an optional maximum step count per episode override for AutoML trials.

public int? MaxStepsPerEpisodeOverride { get; set; }

Property Value

int?

Remarks

If null, the configured MaxStepsPerEpisode is used.

SearchSpaceOverrides

Gets or sets optional hyperparameter search-space overrides.

public Dictionary<string, ParameterRange> SearchSpaceOverrides { get; set; }

Property Value

Dictionary<string, ParameterRange>

Remarks

This dictionary is merged into the built-in search spaces. Unknown keys are ignored by agents that don't support them.

For Beginners: Use this if you want to constrain what AutoML is allowed to try (for example, a smaller learning rate range).

TrainingEpisodesPerTrial

Gets or sets the number of training episodes to run per AutoML trial.

public int TrainingEpisodesPerTrial { get; set; }

Property Value

int

Remarks

For Beginners: Each trial trains briefly to estimate how good that configuration is. Lower values are faster but noisier; higher values are slower but more reliable.