normal#

scikitplot.random.normal(self, loc=0.0, scale=1.0, size=None)#

Normal distribution (Box-Muller transform).

Parameters:

locfloat, default=0.0: Mean
scalefloat, default=1.0: Standard deviation
sizeint or tuple, optional: Output shape

Returns:

float or ndarray: Normal samples

Notes

Uses the Box-Muller transform:: z = sqrt(-2 * log(u1)) * cos(2 * pi * u2)

where u1, u2 are i.i.d. Uniform(0, 1).

u1 is clamped away from zero to prevent log(0) = -inf. Since KissBitGenerator.random() maps [0, 2^64-1] onto [0.0, 1.0) and zero is a valid raw value, u1 == 0.0 is possible (probability 1/2^64).

The size != None path is fully vectorized in NumPy’s C layer. No Python list or scalar boxing is performed, which avoids the GCC -Wstringop-overflow false positives triggered by Python 3.13t’s free-threaded Py_INCREF atomic operations.

Examples

>>> gen = KissGenerator(42)
>>> gen.normal(0, 1, size=1000)
array([...])