Let's say there is a single objective optimization problem SP (e.g. a linear program) and an optimization algorithm SA for it (e.g. simplex). Then applying SA to SP yields an optimal solution s=SA(SP).

Now let's assume there is a multi objective optimization problem MP (e.g. a linear program with two linear objectives) but there is no access to a multi objective optimization algorithm to compute the (Pareto-optimal) solutions (front).

One option is of course to linearise the MP to turn it into a SP, basically the result is a weighted sum (in the example of two terms) but this obviously does not result in all Pareto-optimal solutions but it totally depends on the chosen weights.

Is there any way to iteratively apply the single objective algorithm SA to the multi objective MP to mimic a multi objective optimization algorithm MA?


1 Answer 1


Actually, using a weighted sum does find all Pareto-optimal solutions.

Let your two objective functions be $f_1,f_2$. Define a single objective function

$$f(x) = \alpha f_1(x) + (1-\alpha) f_2(x),$$

where $\alpha$ is a constant. If you sweep $\alpha$ over all possible values in $[0,1]$ and, for each such $\alpha$, solve the single-objective optimization problem (minimizing $f$), then the set of solutions you obtain includes all Pareto-optimal solutions to the original multi-objective optimization problem.


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.