The bbob Test Suite

The blackbox optimization benchmarking (bbob) test suite is COCO's standard test suite with 24 noiseless, single-objective and scalable test functions. Each function is provided in dimensions (2, 3, 5, 10, 20, 40) and in 15 instances, however also available for arbitrary dimensions and number of instances. Links to their definition with visualizations are provided in the table. ECDF of runtimes for 31 algorithms on the bbob suite
1 Separable Functions
f1Sphere Function
f2Separable Ellipsoidal Function
f3Rastrigin Function
f4Büche-Rastrigin Function
f5Linear Slope
2 Functions with low or moderate conditioning
f6Attractive Sector Function
f7Step Ellipsoidal Function
f8Rosenbrock Function, original
f9Rosenbrock Function, rotated
3 Functions with high conditioning and unimodal
f10Ellipsoidal Function
f11Discus Function
f12Bent Cigar Function
f13Sharp Ridge Function
f14Different Powers Function
4 Multi-modal functions with adequate global structure
f15Rastrigin Function
f16Weierstrass Function
f17Schaffer's F7 Function
f18Schaffer's F7 Function, moderately ill-conditioned
f19Composite Griewank-Rosenbrock Function F8F2
5 Multi-modal functions with weak global structure
f20Schwefel Function
f21Gallagher's Gaussian 101-me Peaks Function
f22Gallagher's Gaussian 21-hi Peaks Function
f23Katsuura Function
f24Lunacek bi-Rastrigin Function

Only f1 and f5 are purely convex quadratic (f1) or purely linear in the region of interest $[-5,5]^n$ (f5).

See also N. Hansen et al (2010). Comparing Results of 31 Algorithms from the Black-Box Optimization Benchmarking BBOB-2009. Workshop Proceedings of the GECCO Genetic and Evolutionary Computation Conference 2010, ACM.

A list of all so-far benchmarked algorithms on the bbob suite together with their links to papers describing the experiment can be found in our bbob data archive. Postprocessed data can be found here. For detailed explanations of how to use the functions in a COCO benchmarking experiment, please go to the COCO code page on Github.