AI-woordenlijst
Het complete woordenboek van kunstmatige intelligentie
Min-Max Scaling
Linear rescaling of values into a predefined interval, typically [0,1] or [-1,1]. Presents data by subtracting the minimum and then dividing by the range of the distribution.
Johnson Transformation
System of three families of transformations (SB, SL, SU) covering all possible continuous distributions. Automatically selects the appropriate family and estimates parameters to normalize data.
Variance Stabilizing Transformation
Application of a function whose derivative is inversely proportional to the standard deviation of the variable. Aims to make variance independent of the mean, essential for parametric models.
Freeman-Tukey Transformation
Specific transformation for count data combining square roots of x and x+1. Particularly effective for stabilizing variance of Poisson distributions with small values.
Anscombe Transformation
Square root function with optimized additive correction for data following a Poisson distribution. Transforms x into square root of (x + 3/8) to approximate a normal distribution.
Wilson-Hilferty Transformation
Cubic root transformation approximating a chi-square distribution by a normal distribution. Applies the power 1/3 with correction to reduce skewness of chi-square distributions.
Lambert W Transformation
Application of the Lambert W function to symmetrize skewed Gaussian distributions. Parameterizes the transformation by skewness and kurtosis to reverse Gaussian deformations.
Fisher Transformation
Hyperbolic tangent transformation of correlation coefficients to stabilize their variance. Converts the bounded distribution of correlations [-1,1] into an approximately normal distribution on R.
Robust Scaling
Scaling method using the median and interquartile range instead of the mean and standard deviation. Resistant to extreme values that would affect standard transformations based on moments.