Multi-Objective Optimization

Often we do not only want to optimize just cost or runtime, but both or other objectives instead. SMAC offers a multi-objective optimization interface to do exactly that. Right now, the algorithm used for this is ParEgo [Christescu & Knowles, 2015]. ParEgo weights and sums the individual objectives so that we can optimize a single scalar.

The costs returned by your target algorithm are stored as usual in the runhistory object, such that you can recover the Pareto front later on.

The basic recipe is as follows:

  1. Make sure that your target algorithm returns a cost dictionary containing the objective names as keys and the objective values as values, e.g. {'myobj1': 0.3, 'myobj2': 200}. Alternatively, you can simply return a list, e.g [0.3, 200].

  2. When instantiating SMAC pass the names of your objectives to the scenario object via the multi_objectives argument, e.g. multi_objectives = "myobj1, myobj2" or multi_objectives = ["myobj1", "myobj2"]. Please set run_obj = 'quality'.

  3. Now you can optionally pass a custom multi-objective algorithm class or further kwargs to the SMAC facade (via multi_objective_algorithm and/or multi_objective_kwargs). Per default, ParEgo is used as the multi-objective algorithm.

Warning

Multi-Objective Optimization does currently not support Intensifications like Hyperband or Successive Halving.

We show an example of how to use multi-objective with a nice Pareto front plot in our examples: Scalarized Multi-Objective Using ParEGO.