discover_network hardcodes rng = np.random.default_rng(42) (discovery.py:146) and exposes no random_state. Results are bit-reproducible (good), but a user cannot obtain independent replicates to assess the stability of the shuffle test, which for a permutation-based method is important.
discover_networkhardcodesrng = np.random.default_rng(42)(discovery.py:146) and exposes norandom_state. Results are bit-reproducible (good), but a user cannot obtain independent replicates to assess the stability of the shuffle test, which for a permutation-based method is important.