It is very challenging optimization task with 256 variables.

The hirnyk solution was interrupted due to the limiting number of function evaluations (2000 for each particular search of GlobalSearch command). But why Global optimization ? It is very time consumed. I try local optimizer Search from DirectSearch package and obtain Obj~10^(-24); equality are bad though and the limiting number of function evaluations (10000) also is reached.  You can set evaluationlimit=500000 and run Search for night or even more.

with(Statistics):
CDF(Normal(m,sigma),x);

 Or, if you want Phi(x,m,sigma): 

with(Statistics):
Phi:=unapply(CDF(Normal(m,sigma),x),x,m,sigma);

 Or, if you want Phi(x) (m=0,sigma=1): 

with(Statistics):
Phi:=unapply(CDF(Normal(0,1),x),x);

with(ArrayTools):

A1:=Array([1,2,3,4,5,6,7]); 
A2:=Array([1,2,3.3,4.4,5.5,6.0,7.0]);  # two Arrays

d1:=A1-~A1;
d2:=A2-~A1;  #differences between two Arrays

IsZero(d1);
IsZero(d2);  # if true then Arrays have equal elements

Statistics:-Count command computes the size of one-dimensional Array, Vector and list.

with(Statistics):
Count([1,2,3,4]);
Count(<1,2,3,4>);
Count(Array([1,2,3,4]));

Axel,

Thank you.

I'm sorry, but I have not package variant for Classical interface. Because Maple actually does't  support Classical inerface it seems Standard interface has not alternatives.

The package was written on Maple 12; from Maple 13 there are some 3d plot option such as "color='variable name'",  "plotlist" and "viewpoint=circleleft". If you remove these options all must be okey, I hope.

There is package "The orthogonal series expansions package for Maple" in Application centre.

This package have command GramSchmidtL2() for orthonormalization of uni- and multi-variable functions.