Statistical Quality and Reproducibility of Pseudorandom Number Generators in Machine Learning Technologies

Machine learning (ML) frameworks rely heavily on pseudorandom number generators (PRNGs) for tasks such as data shuffling, weight initialization, dropout, and optimization. Yet, the statistical quality and reproducibility of these generators—particularly when integrated into frameworks like PyTorch, TensorFlow, and NumPy—are underexplored. In this paper, we compare the statistical quality of PRNGs used in ML frameworks (Mersenne Twister, PCG, and Philox) against their original C implementations. Using the rigorous TestU01 BigCrush test suite, we evaluate 896 independent random streams for each generator. Our results challenge claims of statistical robustness, revealing that even generators labelled "crush-resistant" (e.g., PCG, Philox) may fail certain statistical tests. Surprisingly, we can observe some differences in failure profiles between the native and framework-integrated versions of the same algorithm, highlighting some implementation differences that may exist. Mersenne Twister implementation in Pytorch and Numpy does not have the exact same failure profile as the original implementation in C. In addition, this is also the case for the TensorFlow implementation of Philox.