If you have continuous variables, then I think the "all different" restriction does not make sense. You can perturb an element by a very small amount and end up with a "different" value, but one which does not really change the value of an objective or constraint function, to within numerical tolerances.
As far as integer constraints, the Optimization Toolbox does not handle such constraints except for the bintprog function. The Global Optimization Toolbox ga function does handle integer constraints, and I believe that you could write a nonlinear constraint function that would attempt to enforce "all different" members of a population. For example, if components 1 through 6 are integer-valued, you could make a nonlinear inequality constraint function that is negative if these 6 elements have different values, and is positive if they have an identical value. For example:
function [c,ceq] = checkdifferent(x)
ceq = [];
ll = length(unique(x(1:6)));
c = -ll + 5.5; % negative when first 6 entries are different
Be warned, I have not tested this to see how well it works in practice.
Alan Weiss
MATLAB mathematical toolbox documentation
