| |
| """Unit checks for the four agreed acquisition formulations.""" |
|
|
| import numpy as np |
|
|
| from run_experiment import RuntimePrediction, acquisition_values |
|
|
|
|
| def main() -> None: |
| mean = np.array([0.2, 0.4, 0.6, 0.8]) |
| score_sd = np.array([0.1, 0.1, 0.1, 0.1]) |
| runtime = RuntimePrediction( |
| mean=np.array([1.0, 2.0, 3.0, 4.0]), |
| sd=np.array([0.4, 0.3, 0.2, 0.1]), |
| conservative=np.array([1.4, 2.3, 3.2, 4.1]), |
| lower=1.0, |
| upper=5.0, |
| ) |
| rng = np.random.default_rng(0) |
| for method in ["direct_cost_uc", "runtime_uncertainty", "cost_tiers_5", "relaxed_cost_power_05", "log_cost"]: |
| utility, acquisition, selected_cost = acquisition_values( |
| method, mean, score_sd, runtime, np.arange(4), np.ones(7), rng, |
| ) |
| assert np.isfinite(utility).all() |
| assert np.isfinite(acquisition).all() |
| assert (acquisition > 0).all() |
| if method == "runtime_uncertainty": |
| assert np.array_equal(selected_cost, runtime.conservative) |
| else: |
| assert np.array_equal(selected_cost, runtime.mean) |
| print("formula tests: PASS") |
|
|
|
|
| if __name__ == "__main__": |
| main() |
|
|
