0
$\begingroup$

I was reading this question on CS stack exchange called How important is initial state for local search optimisation?

I would like to extend it with the following example:

I have been reading about optimisation techniques and noticed that if a random initial state is provided to a genetic algorithm, it may still find a good enough solution. However, a simulated annealing algorithm may be stuck and never get out of really bad initial solution.

Does it mean these two are not comparable if a bad initial solution is provided? Are there any other algorithms that are not that sensitive to an initial solution?

Improve this question
$\endgroup$
1
  • $\begingroup$ It's often more about the problem than about the algorithm (although usually both matter). E.g. in convex optimization, the initial solution doesn't matter. $\endgroup$
    – Dmitry
    Feb 27 at 19:16

1 Answer 1

Reset to default
0
$\begingroup$

The answer is 'it depends'. It depends on the particular problem you are optimizing. I doubt you are going to get a single answer that applies to everything. In general, reducing the dependence on the initial state is generally considered a good property of an optimization algorithm, but there is no clear way to evaluate this, and the dependence is likely to depend on the specific optimization task you are solving. It is very hard to make sweeping statements, because how well an optimization approach works depends heavily on the task.

Improve this answer
$\endgroup$

Your Answer

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

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