ProbabilityCalibratorOptions
Options & Configuration · AiDotNet.Models.Options
Configuration options for probability calibrator.
For Beginners
Many models give probability-like outputs, but these aren’t always true probabilities. For example:
- A decision tree might output 0.9 for all “yes” predictions (overconfident)
- A neural network might output values between 0.3-0.7 only (underconfident)
- SVM outputs are not probabilities at all without calibration
Calibration transforms these outputs into reliable probabilities:
- If you see 1000 predictions of “60% chance”, about 600 should be correct
- This is crucial for decision-making (medical diagnoses, financial risk, etc.)
When to use:
- Whenever you need actual probability estimates (not just rankings)
- When combining predictions from different models
- For threshold-based decisions where probability values matter
How It Works
Probability calibration ensures that predicted probabilities are reliable. When a calibrated model says “70% chance”, you should expect the event to occur about 70% of the time across many such predictions.
Constructors
| Constructor | Summary |
|---|---|
ProbabilityCalibratorOptions() |
Properties
| Property | Summary |
|---|---|
ProbabilityCalibrationMethod CalibratorMethod { get; set; } |
Gets or sets the calibration method. |
bool UseCrossValidation { get; set; } |
Gets or sets whether to use cross-validation for calibration. |
double LearningRate { get; set; } |
Gets or sets the learning rate for gradient-based methods. |
double Regularization { get; set; } |
Gets or sets the regularization strength. |
double Tolerance { get; set; } |
Gets or sets the convergence tolerance. |
int MaxIterations { get; set; } |
Gets or sets the maximum number of iterations for optimization-based methods. |
int NumBins { get; set; } |
Gets or sets the number of bins for histogram-based methods. |
int NumFolds { get; set; } |
Gets or sets the number of cross-validation folds. |
int? Seed { get; set; } |
Gets or sets the random seed for reproducibility. |