Curve fitting is a special case of a least squares problem were the parameters are the coefficients of a function f whose graph y=f(x) should pass through sample points, and were the objective function is the squared sum of residuals f(xi)-yi for observed points (x_i, y_i).