ReliabilityResult
@Serializable
@SerialName(value = "ReliabilityResult")
Per-bin reliability snapshot for a binary probabilistic classifier. Bins are equal-width over [0, 1] indexed by predicted probability. Underlying sums use weights so soft labels and importance-weighted streams compose correctly.
Constructors
ReliabilityResult
constructor(numBins: Int, sumProbability: DoubleArray, sumOutcome: DoubleArray, totalWeights: DoubleArray)(source)
Properties
meanProbability
Mean predicted probability per bin (NaN where the bin is empty).
numBins
outcomeRate
Empirical outcome rate per bin (NaN where the bin is empty).
sumOutcome
Sum w_i*outcome_i per bin.
sumProbability
Sum w_i*prob_i per bin.
totalWeights
Sum w_i per bin.
Functions
equals
expectedCalibrationError
Expected Calibration Error: weighted mean over bins of |meanProbability[i] - outcomeRate[i]|. Empty bins contribute 0.