- Mark as New
- Bookmark
- Subscribe
- Mute
- Subscribe to RSS Feed
- Permalink
- Report Inappropriate Content
Hello,
I've written an application that uses random numbers to generate an average (given sufficient sample size it converges) and am running into some strange problems with Sobol and quasi-random numbers. If I use VSL_BRNG_SOBOL and seed it with 2*ndims (taking MV gaussian draws), the applications works correctly when I take an odd number of draws (i.e. 3, 5, etc.). However, when I take an even number (i.e. 2, 4, etc.), I get results that are not correct!
Substituting VSL_BRNG_MCG59 and seeing with a random int to get pseudo-random numbers works correctly every time. What am I doing wrong or what is wrong with Sobol?
Thanks!
I've written an application that uses random numbers to generate an average (given sufficient sample size it converges) and am running into some strange problems with Sobol and quasi-random numbers. If I use VSL_BRNG_SOBOL and seed it with 2*ndims (taking MV gaussian draws), the applications works correctly when I take an odd number of draws (i.e. 3, 5, etc.). However, when I take an even number (i.e. 2, 4, etc.), I get results that are not correct!
Substituting VSL_BRNG_MCG59 and seeing with a random int to get pseudo-random numbers works correctly every time. What am I doing wrong or what is wrong with Sobol?
Thanks!
Link Copied
1 Reply
- Mark as New
- Bookmark
- Subscribe
- Mute
- Subscribe to RSS Feed
- Permalink
- Report Inappropriate Content
Thank you for your question. It could be nice if you providea short test case which demonstrates the situation youhave described. Also, some other detailswould be useful: what version of MKL you use, and what method (BOXMULLER or BOXMULLER 2) is used for generation of the random vectors. Thank you.
Reply
Topic Options
- Subscribe to RSS Feed
- Mark Topic as New
- Mark Topic as Read
- Float this Topic for Current User
- Bookmark
- Subscribe
- Printer Friendly Page