Thanks a lot, now it helps. I am sorry for polluting forum with questions. I am learning maple and I am sometimes confused with results. Tomas Pevny

Hi, thanks a lot for suggestions. I started with replacing repeated derivatives with taylor expansion, but I failed because I cannot get individual coefficients of Taylor expansion. I submitted following commands: f:= exp(-(y-1)^2/2)/(sqrt(2*Pi)) k := exp(-.472272*(x-y)^2); dF := int(f*k, y = -infinity .. infinity); dF := dF*exp(.472272*x^2) T := taylor(dF, x, 40) S:=convert(T, 'polynom') I expected that convert will get rid of O(x^40) term in T and than, I can use coeffs to access individual terms in the expansion. But it des not work this way. When I use convert, I loose terms x^0 to x^2 in the expansion. Can anyone help? I am lost. Thanks a lot, Tomas Pevny

Thanks a lot for the explanation. As I said, I am new to the maple and I am not use to the fact that the same algorithm will return me different results. Tomas

Thanks a lot. Setting digits to 20 works. Still I have to say that the behavior of maple is weird to me. Even there is numerical instability, if the algorithm is deterministic, it should returns same results, because the initial conditions are the same. Or do I miss something? Tomas