Candidate Point Generators

Functions for generating candidate points from which the batch proposer selects. Different generators trade off coverage, speed, and adaptivity.

Details

All generator functions return a generator function with signature:

function(search_space, n)

The generator function returns a data.table of n candidate configurations.

Available Generators

• candidate_generator_lhs(): Latin Hypercube Sampling (space-filling)

• candidate_generator_random(): Pure random sampling

• candidate_generator_sobol(): Sobol quasi-random sequence

• candidate_generator_grid(): Regular grid (for small spaces)

• candidate_generator_local(): Local sampling around best points

• candidate_generator_mixed(): Combination of global and local sampling