Function minimization example - using FloatingPoint or Binary genotype.
- Description
- Tries to find the (global) minimum of a continuous multidimensional function. The example uses either FloatingPoint or Binary genotype. Functions are implemented in Black-Box Optimization Benchmarking package available at http://coco.gforge.inria.fr. The location of the optimum is drawn uniformly from [-4, 4] in all dimensions for all functions. The value of the optimum of all functions is set to zero in this example (in original BBOB package the value is in [-100, 100]).
The example produces two executables (separate VS projects):
- funcmin: runs the selected algorithm on a selected function
- ecf_experiment: runs the selected algorithm on a range of functions in a BBOB-defined experimentation scheme (see exampleexperiment.cpp)
To run the ecf_experiment executable:
- define functions and function instances in exampleexperiment.cpp
- define all the other parameters (the algorithm, output folder, dimension, termination criteria) in the configuration file (see parameters.txt)
- Parameters (in the config file)
- "coco.function": defines the minimization function to be used (default: 1); 1-24 noiseless functions, 101-130 noisy
- "coco.enableoutput": should COCO framework output results for post-processing (disabled by default)
- "coco.folder": destination folder for COCO output; must be created beforehand (see /bbob/README.txt)
- with FloatingPoint genotype, a number of dimensions and the interval [lbound, ubound] must be set
- with Binary genotype, a number of dimensions, the interval [lbound, ubound] and precision (number of digits after the decimal point) must be set
- References
-
1.18.0