Dear Waldek, just to make sure:
> Oh, I think I made a stupid mistake. I defined > > mon(n,k) == (random k * x + random k * y)^n > > which makes the coefficients not of size log k, but rather of size > > log(binomial(n,j)* k^j). Stupid me. I just ran the test with mon(n,k) == reduce(+, [random k * x^i*y^(n-i) for i in 0..n]) instead, and get the "expected" O(k^2) output: (6) -> [test(kappa, 500, 5) for kappa in 10..100 by 10] [[10,1],[10,0],[10,1],[10,0],[10,1]] [[20,2],[20,2],[20,2],[20,2],[20,2]] [[30,4],[30,4],[30,3],[30,3],[30,5]] [[40,7],[40,7],[40,7],[40,7],[40,6]] [[50,11],[50,10],[50,11],[50,10],[50,11]] [[60,18],[60,15],[60,14],[60,15],[60,15]] [[70,19],[70,20],[70,20],[70,20],[70,22]] I also modified William Sit's test (using Expand), and obtained Out[11]= {{10, 0.272878}, {10, 0.273776}, {10, 0.273362}, {20, 0.902124}, > {20, 0.899902}, {20, 0.899222}, {30, 1.920647}, {30, 1.922753}, > {30, 1.916153}, {40, 3.617779}, {40, 3.708420}, {40, 3.608447}, > {50, 6.507658}, {50, 6.525488}, {50, 6.509765}, {60, 10.755069}, > {60, 10.809554}, {60, 10.868204}, {70, 16.384399}, {70, 16.461227}, > {70, 16.337861}} (on a much faster machine) which corresponds to Axiom's performance. So, it remains to check, what situation is realistic for my problem, and whether I can do anything about it... Many thanks, Martin _______________________________________________ Axiom-developer mailing list Axiom-developer@nongnu.org http://lists.nongnu.org/mailman/listinfo/axiom-developer