Files Used During the Tuning Process

Each configuration file belongs to one SPOT project, if the same basename is used for corresponding files. SPOT uses simple text files as interfaces from the algorithm to the statistical tools.

1. The user has to provide the following files:

   (i)

   Region of interest (ROI) files specify the region over which the algorithm parameters are tuned. Categorical variables such as the recombination operator in ES, can be encoded as factors, e.g., ``intermediate recombination'' and ``discrete recombination.''

   (ii)

   Algorithm design (APD) files are used to specify parameters used by the algorithm, e.g., problem dimension, objective function, starting point, or initial seed.

   (iii)

   Configuration files (CONF) specify SPOT specific parameters, such as the prediction model or the initial design size.

   (iv)

   Meta files (META) specify parameters from the problem design. The algorithm is separately tuned for each setting of the meta parameters. For example, consider tuning a (1+1)-ES on the sphere function for varying problem dimensions, say $n=10, 20, \ldots, 100$. The problem dimension $n$ is a meta parameter in this case.

2. SPOT will generate the following files:

   (i)

   Design files (DES) specify algorithm designs. They are generated automatically by SPOT and will be read by the optimization algorithms.

   (ii)

   After the algorithm has been started with a parametrization from the algorithm design, the algorithm writes its results to the result file (RES). Result files provide the basis for many statistical evaluations/visualizations. They are read by SPOT to generate prediction models. Additional prediction models can easily be integrated into SPOT.

   (iii)

   During the meta mode, the algorithm is tuned on several problem instances. The best configuration from each problem instance is written to the final best solution file (FBS).