Timeline for Compare C++ Standard Library's Mersenne Twister with Mathematica's Mersenne Twister
Current License: CC BY-SA 3.0
8 events
| when toggle format | what | by | license | comment | |
|---|---|---|---|---|---|
| Apr 13, 2017 at 12:55 | history | edited | CommunityBot |
replaced http://mathematica.stackexchange.com/ with https://mathematica.stackexchange.com/
|
|
| Feb 14, 2013 at 4:20 | comment | added | Oleksandr R. | @Stefan you missed my point. MT19937 generates pseudorandom strings of 32 bits at a time. How that output is used by Mathematica to produce a result drawn from a particular distribution is completely unspecified and could easily account for the difference. I highly doubt that it is giving you the raw output from the underlying RNG. | |
| Feb 13, 2013 at 21:12 | comment | added | Szabolcs | @Stefan To put it another way: if you need a quick solution right now, I suggest you use C++'s RNG as I described. If you really want to know if you can set parameters, the best way to find out is asking [email protected]. This will take some time though and there may not be such parameters at all. Or if there are, you may not have access to all of them, e.g. word length. If they reply you, please post the reply here as an answer. | |
| Feb 13, 2013 at 21:09 | comment | added | Stefan | @OleksandrR. hmm...MT19937 is a pseudo-random generator. You enter a specific seed value and it is guaranteed that the distribution will always be identical...if not, well, flawed. I don't say, they've used it in a unusual way. Maybe it is not a MT19937 at all, which would make me sad. Btw. I've already exported the distribution to a csv file days before, in order to compare the results of my algorithms with those from Mathematica to ensure I'm correct. Don't you think that you should be allowed to have influence on parameters within the RNG framework? Well, I think so. | |
| Feb 13, 2013 at 19:56 | comment | added | Oleksandr R. | @Stefan they could even be using the same parameters--if the method for producing particular values from the RNG output differs, the results will still be different. I would think this is a much more likely explanation than that they decided to use the Mersenne twister in an unusual way. | |
| Feb 13, 2013 at 19:49 | comment | added | Stefan | Interesting. LibraryLink was news to me, but using the same parameter sets plus the same seed hast to provide the same answers. On the standard library side these values and the imementation are open. I'm not interested how, at wolfram, the guys did implement this. I was asking because I'd like to verify some algorithms I wrote in C++, in Mathematica to proof they work correct. They use a different parameter set. Why? Can I change it as a user? | |
| Feb 13, 2013 at 19:45 | history | edited | Szabolcs | CC BY-SA 3.0 |
added 100 characters in body
|
| Feb 13, 2013 at 19:39 | history | answered | Szabolcs | CC BY-SA 3.0 |