metaheuristic_designer.analysis.experiment_runner module#

run_experiment(problems, algorithms, max_evals, n_runs=30, base_seed=42, output_root='experiment_data')[source]#

Run a fair, reproducible comparison of algorithms on a set of problems.

Parameters:

problemslist of ObjectiveFunc: Benchmark functions to solve. IOHObjective instances will produce IOH-compatible log files automatically.
algorithmsdict of str -> callable: Keys are algorithm names; values are factories (objfunc, seed, budget) -> solver.
max_evalsint: Common evaluation budget per run.
n_runsint: Number of independent repetitions per (problem, algorithm) pair.
base_seedint: Master seed for reproducibility.
output_rootstr: Folder where IOH log files are written. Non-IOH problems are silently ignored.

Returns:

pd.DataFrame: Columns: algorithm, problem_name, fid, dimension, instance, run, best_objective.

Parameters: