Generate Gaussian-distributed random floats strictly within [a, b] using rejection sampling (no clipping)

gaussian_generator.py

import numpy as np


def gaussian_random_reject(a, b, mean=None, std=None, size=1):
    """
    Generate Gaussian-distributed random floats strictly within [a, b]
    using rejection sampling (no clipping).

    Parameters:
        a, b : float
            Range limits.
        mean : float (optional)
            Mean of the Gaussian. Defaults to midpoint of [a, b].
        std : float (optional)
            Standard deviation. Defaults to (b - a) / 6.
        size : int
            Number of samples to return.

    Returns:
        numpy.ndarray of shape (size,)
    """
    if mean is None:
        mean = (a + b) / 2
    if std is None:
        std = (b - a) / 6

    samples = []
    while len(samples) < size:
        # generate in batches for efficiency
        batch = np.random.normal(loc=mean, scale=std, size=size)
        # keep only those within [a, b]
        accepted = batch[(batch >= a) & (batch <= b)]
        samples.extend(accepted.tolist())

    return np.array(samples[:size])


# Example usage:
"""
vals = gaussian_random_reject(1, 10, size=5)
print(vals)
"""