But sorry, we've made a mistake...

The f function is actually more challenging that the above one, it should be:

f:= x -> exp(3*x)/((1+exp(x))^5);

Could you please take a look at this again Rob?


If we try using the "series" function, we got the following:

Error, division by zero series

If we trying using the "minimax" function, we got the following:

> minimax(f(x), x = -0.3e-1 .. 0.3e-1, 10);

Error, (in numapprox:-remez) error curve fails to oscillate sufficiently; try different degrees

What could be the problems? Thanks a lot!

Thanks! I knew Cholesky decomposition is easy and it is not a mess when translated into C...

However I do want M^(0.5)... which is a mess...

Robert: I cannot seem to understand how to use your algorithm for obtaining my 3x3 square root matrix?

Also, since it's going to be used in numerical program, we probably cannot assume the eigenvalues are distinct... anything can happen for that matrix.

Of course, we know it's real...

Any thoughts? Thanks!