I am currently implementing a machine learning method, which makes use of random numbers in C++. I am using boost and a seeded mersenne twister to initialize the RNG. Curiously, when I run seeded instances the RNGs diverge after a certain time of producing identical random numbers. I am stumped as to how this is possible since the RNGs should be deterministic. When running the RNG code excluded from the other program code I obtain the desired identical (pseudo-)randomness. Has anyone an idea as to what could be causing the divergence or an idea how to debug this?
Aucun commentaire:
Enregistrer un commentaire