GA optimization: presenting bounds as linear constraints

GA generates different results if I use linear constraints to describe the feasible search space instead of using bounds. In addition, it looks like:

• Substituting bounds with linear constraints makes the optimization process faster.
• The results obtained by using bounds rather than equivalent linear constraints were significantly improved!? (better objective function value - with minimization problem).

Could anyone please share with me your experience regarding this problem? much appreciated!

Answer 1

[Matt J]

Yes, that's to be expected. Because a region specified by bounds has explicit boundaries, GA can select an initial population confined to that region and force future generations to lie within that region. It cannot do that with more general constraints.