HyperbandOptimizer<T, TInput, TOutput>

Models & Types · AiDotNet.HyperparameterOptimization

Implements Hyperband optimization for hyperparameter tuning with early stopping.

Inheritance: HyperparameterOptimizerBase<T, TInput, TOutput> → HyperbandOptimizer<T, TInput, TOutput>

How It Works

For Beginners: Hyperband is a smart resource allocation strategy that:

  • Trains many configurations with minimal resources initially
  • Progressively eliminates poorly performing configurations
  • Allocates more resources to promising configurations
  • Uses “successive halving” to efficiently explore the search space

Key concepts:

  • Resource (R): Training budget (e.g., epochs, iterations, data samples)
  • Configuration: A specific set of hyperparameter values
  • Bracket: A group of configurations competing via successive halving
  • Successive Halving: Repeatedly train, evaluate, and keep top half

Hyperband runs multiple brackets with different exploration/exploitation trade-offs, combining the best properties of random search with aggressive early stopping.

Constructors

Constructor Summary
HyperbandOptimizer(bool maximize, int maxResource, int reductionFactor, int minResource, int? seed) Initializes a new instance of the HyperbandOptimizer class.

Properties

Property Summary
int NumBrackets { get; } Gets the number of brackets in this Hyperband configuration.

Methods

Method Summary
Dictionary<string, object> SuggestNext(HyperparameterTrial<T> trial) Suggests the next hyperparameter configuration to try.
HyperparameterOptimizationResult<T> Optimize(Func<Dictionary<string, object>, T> objectiveFunction, HyperparameterSearchSpace searchSpace, int nTrials) Searches for the best hyperparameter configuration using Hyperband.
List<BracketInfo> GetBracketInfo() Gets detailed information about the bracket structure.
int GetTotalConfigurationCount() Gets the total number of configurations that would be evaluated in a full Hyperband run.