UncertaintyOptions
Options & Configuration · AiDotNet.Evaluation.Options
Configuration options for uncertainty quantification.
For Beginners
Uncertainty tells you how confident the model is:
- Aleatoric uncertainty: Inherent noise in the data (can’t be reduced)
- Epistemic uncertainty: Model’s lack of knowledge (can be reduced with more data)
A good uncertainty estimate lets you flag predictions the model isn’t sure about, which might need human review.
How It Works
Uncertainty quantification measures how confident a model is in its predictions. This is crucial for high-stakes applications where knowing “I don’t know” is valuable.
Constructors
| Constructor | Summary |
|---|---|
UncertaintyOptions() |
Properties
| Property | Summary |
|---|---|
PredictionIntervalMethod? IntervalMethod { get; set; } |
Method for prediction interval estimation. |
UncertaintyCalibrationMethod? CalibrationMethod { get; set; } |
Uncertainty calibration method. |
UncertaintyDecompositionMethod? DecompositionMethod { get; set; } |
Method for uncertainty decomposition. |
bool? CalibrateUncertainty { get; set; } |
Whether to compute calibrated uncertainty. |
bool? CheckQuantileCrossing { get; set; } |
Whether to check for quantile crossing. |
bool? ComputeCWC { get; set; } |
Whether to compute CWC (Coverage Width Criterion). |
bool? ComputeMPIW { get; set; } |
Whether to compute MPIW (Mean Prediction Interval Width). |
bool? ComputeMutualInformation { get; set; } |
Whether to compute mutual information (epistemic uncertainty). |
bool? ComputePICP { get; set; } |
Whether to compute PICP (Prediction Interval Coverage Probability). |
bool? ComputePredictionIntervals { get; set; } |
Whether to compute prediction intervals. |
bool? ComputePredictiveEntropy { get; set; } |
Whether to compute predictive entropy. |
bool? ComputeSharpness { get; set; } |
Whether to compute sharpness score. |
bool? ComputeUncertaintyReliability { get; set; } |
Whether to compute reliability diagrams for uncertainty. |
bool? DecomposeUncertainty { get; set; } |
Whether to decompose uncertainty into aleatoric/epistemic. |
bool? IdentifyHighUncertaintySamples { get; set; } |
Whether to identify high-uncertainty samples. |
double? CWCPenaltyFactor { get; set; } |
Penalty factor for CWC calculation. |
double? HighUncertaintyThreshold { get; set; } |
Threshold for high uncertainty (percentile). |
double? MCDropoutRate { get; set; } |
Dropout rate for MC Dropout. |
double? PredictionIntervalCoverage { get; set; } |
Coverage level for prediction intervals. |
double[] Quantiles { get; set; } |
Quantiles to compute for quantile regression. |
int? EnsembleSize { get; set; } |
Number of ensemble members for ensemble-based uncertainty. |
int? MCDropoutSamples { get; set; } |
Number of Monte Carlo dropout samples. |
int? RandomSeed { get; set; } |
Random seed for reproducibility. |
int? ReliabilityBins { get; set; } |
Number of bins for uncertainty reliability diagram. |