def run_exp(scenario, remove_small, or_level, current_sites, remove_timeout,
exclude_sites, _rnd):
config.OR_LEVEL = config.OR_LEVEL if or_level is None else or_level
config.REMOVE_SMALL = (config.REMOVE_SMALL
if remove_small is None else remove_small)
config.REMOVE_TIMEOUT = (config.REMOVE_TIMEOUT
if remove_timeout is None else remove_timeout)
scenario_obj = Scenario(scenario, exclude_sites=exclude_sites)
traces = scenario_obj.get_traces(current_sites=current_sites)
result = analyse.simulated_original(traces)
X, y, d = scenario_obj.get_features_cumul(current_sites)
X = preprocessing.MinMaxScaler().fit_transform(X) # scaling is idempotent
clf = result.clf.estimator
y_pred = model_selection.cross_val_predict(clf,
X,
y,
cv=10,
n_jobs=config.JOBS_NUM)
_add_as_artifact(y.tolist(), "y_true")
_add_as_artifact(y_pred.tolist(), "y_prediction")
_add_as_artifact(d, "y_domains")
return {
'C': result.clf.estimator.C,
'gamma': result.clf.estimator.gamma,
'sites': traces.keys(),
'score': result.best_score_,
'type': "cumul",
'C_gamma_result': _format_results(result.results),
# todo: remove duplicates (or_level...), after checking that they match
'outlier_removal': scenario_obj.trace_args,
'size_increase': scenario_obj.size_increase(),
'time_increase': scenario_obj.time_increase()
}
