Skip to main content
10 events
when toggle format what by license comment
May 12, 2015 at 9:40 comment added VividD Any chance to buy more RAM?
May 4, 2015 at 13:51 comment added Sasha @BenjaminDickman As I have already stated in the above comment on Apr. 25, the direct use of quadrature integration of $f(w)$ is bound to fail, because the integrand is discontinuous. I have implemented and attempted to run a hybrid-numeric approach, where $w_1$ is integrated out symbolically for fixed numeric remaining $w_i$. My attempted run of this crashed consuming over 40GB of ram. I have not attempted to rerun this until now. Therefore I did not have any good news updates to post.
May 2, 2015 at 17:12 comment added Benjamin Dickman ... Still running???
Apr 5, 2014 at 22:37 comment added Eckhard Still running?${}$
Apr 25, 2013 at 9:56 comment added achille hui @Sasha What a pity, I also want to know more digits. In your way of computing the numerical integral, will it be easier to compute the contribution from the 7 permutations mentioned in leonbloy's update one by one? I suspect the lower probability for a permutation, the more problematic it will be for adaptive quadratures.
Apr 25, 2013 at 4:27 comment added Sasha Yes, it still does. The function $f(w_1,\ldots,w_5)$ is discontinuous, therefore adaptive numerical quadrature methods run into trouble endlessly subdividing close the boundary. The call I quoted finished in 22 hours, but converged to only 2 significant digits of precision, which is less than what Monte-Carlo gave. I rerun the computation with different settings, which will hopefully improve the matter. Ultimately, I think one would need to resolve to hybrid symbolic-numeric approach, like integrate wrt last variable symbolically.
Apr 25, 2013 at 4:20 comment added Benjamin Dickman It's still running?
Apr 24, 2013 at 23:59 history bounty awarded CommunityBot
S Apr 23, 2013 at 20:11 history answered Sasha CC BY-SA 3.0
S Apr 23, 2013 at 20:11 history made wiki Post Made Community Wiki by Sasha