NGBoostRegressionOptions<T>
Options & Configuration · AiDotNet.Models.Options
Configuration options for NGBoost (Natural Gradient Boosting) regression models.
Inheritance: ModelOptions → DecisionTreeOptions → NGBoostRegressionOptions<T>
For Beginners
NGBoost is like regular gradient boosting, but instead of predicting a single number, it predicts a full probability distribution. This tells you not just “what the prediction is” but also “how confident the model is.”
For example, instead of predicting “house price = $300,000”, NGBoost might predict “house price is normally distributed with mean $300,000 and standard deviation $50,000.” This uncertainty information is valuable for decision-making.
How It Works
NGBoost is a probabilistic gradient boosting algorithm that outputs full probability distributions instead of point predictions. It uses natural gradients (gradients preconditioned by the Fisher Information Matrix) for more stable optimization.
Constructors
| Constructor | Summary |
|---|---|
NGBoostRegressionOptions() |
Properties
| Property | Summary |
|---|---|
IScoringRule<T> ScoringRule { get; set; } |
Gets or sets the scoring rule used for optimization. |
NGBoostDistributionType DistributionType { get; set; } |
Gets or sets the type of distribution to fit. |
bool UseNaturalGradient { get; set; } |
Gets or sets whether to use natural gradients. |
bool Verbose { get; set; } |
Gets or sets whether to verbose output during training. |
double ColumnSubsampleRatio { get; set; } |
Gets or sets whether to apply column subsampling. |
double LearningRate { get; set; } |
Gets or sets the learning rate (shrinkage factor). |
double SubsampleRatio { get; set; } |
Gets or sets the subsampling ratio for each iteration. |
int NumberOfIterations { get; set; } |
Gets or sets the number of boosting iterations (trees). |
int VerboseEval { get; set; } |
Gets or sets how often to print progress (every N iterations). |
int? EarlyStoppingRounds { get; set; } |
Gets or sets the number of early stopping rounds. |