First I'd suggest you to create an m-file to evaluate your error vector.
% Example for that function:
function Error = CostFunction(Unknowns)
% Unknowns can be a vector, so you can extract it like this:
a = Unknowns(1);
b = Unknowns(2);
% Compute the Error here.
% End of m-file
Now in your main m-file you can define the initial conditions for your unknowns:
Unknowns = [1, 0, 4]; % 3 variables in this example
options = optimset(' .... ', ' .... '); % your solver options goes here
UnknownsLB = [0, -5, 2]; % lower bound definition
UnknownsUB = [2, 5, 10]; % upper bound definition
x = lsqnonlin(@CostFunction,Unknowns,UnknownsLB,UnknownsUB,options) % solve it!
