Solver: Random Search (RNDSRCH)

Description:

Randomly sample solutions from the feasible region and use a fixed number of replications at each solution. The sampling distribution is specified inside each problem class in get_random_solution.

Modifications & Implementation:

The new random solutions maintain the type of each variable based on the sampling distributions that are discrete for integer decisions and otherwise continuous.

Scope:

• objective_type: single

• constraint_type: stochastic

• variable_type: mixed

• gradient_observations: not available

Solver Factors:

• crn_across_solns: Use CRN across solutions?

  • Default: True

• sample_size: Sample size per solution > 1.

  • Default: 10

References:

This solver is adapted from the article Chia, Y.L. and Glynn, P.W., (2013). Limit Theorems for Simulation-Based Optimization via Random Search. ACM Transactions on Modeling and Computer Simulation (TOMACS), 23(3), pp.1-18. (https://dl.acm.org/doi/abs/10.1145/2499913.2499915)