poli.core.problem#

Implements an abstract problem.

Problems are standardized ways of running benchmarks. They are used to create black boxes and to run optimization algorithms.

At its core, a problem is a black box with a known setup. This setup includes the following information:

  • The initial evaluation(s) x0 and y0.

  • The fidelity of the black box (e.g., “high” or “low”).

  • The black box function and its information (e.g. whether it is noisy, continuous or discrete, etc.).

  • An evaluation budget, which is the maximum number of evaluations allowed.

Classes

Problem(black_box, x0[, data_package, ...])