When parameter tuning actually is parameter control
Wessing S., Preuss M., Rudolph G.
Abstract
In this paper, we show that sequential parameter optimization (SPO), a method that was designed for (offline) parameter tuning, can be successfully used as a controller for multistart approaches of evolutionary algorithms (EA). We demonstrate this by replacing the restart heuristic of the IPOP-CMA-ES with the SPO algorithm. Experiments on the BBOB 2010 test cases suggest that the performance is at least competitive while the approach provides more options, e.g. setting more than one parameter at once. Essentially, we argue that SPO is a generalization of the IPOP heuristic and that the distinction between tuning and control is - although often useful - an artificial one. Copyright 2011 ACM.
Keywords
CMA-ES; Parameter control; Parameter setting; Parameter tuning; SPO