BayesianOptimizer<T, TInput, TOutput>
Models & Types ยท AiDotNet.HyperparameterOptimization
Implements Bayesian optimization for hyperparameter tuning using Gaussian Process regression.
Inheritance: HyperparameterOptimizerBase<T, TInput, TOutput> โ BayesianOptimizer<T, TInput, TOutput>
How It Works
For Beginners: Bayesian optimization is a smart search strategy that learns from previous trials to decide what to try next. Unlike grid or random search, it:
- Builds a model of how hyperparameters affect performance
- Uses this model to focus on promising regions
- Balances exploration (trying new areas) with exploitation (refining good areas)
This makes it much more efficient than random search, often finding good hyperparameters in 10-20 trials instead of 100s.
Key components:
- Surrogate Model: A Gaussian Process that predicts performance for any hyperparameter combination
- Acquisition Function: Decides where to sample next based on predicted mean and uncertainty
- Sequential Optimization: Each trial informs the next, unlike parallel random search
Constructors
| Constructor | Summary |
|---|---|
BayesianOptimizer(bool maximize, AcquisitionFunctionType acquisitionFunction, int nInitialPoints, double explorationWeight, int? seed) |
Initializes a new instance of the BayesianOptimizer class. |
Methods
| Method | Summary |
|---|---|
Dictionary<string, object> SuggestNext(HyperparameterTrial<T> trial) |
Suggests the next hyperparameter configuration using the acquisition function. |
HyperparameterOptimizationResult<T> Optimize(Func<Dictionary<string, object>, T> objectiveFunction, HyperparameterSearchSpace searchSpace, int nTrials) |
Searches for the best hyperparameter configuration using Bayesian optimization. |