add script to evaluate the amount of outlier runs depending on training parameters

mu_map/random_search/eval/outlier.py

+from mu_map.random_search.eval.params import parameter_groups
+from mu_map.random_search.eval.util import filter_by_params, load_data, remove_outliers
+
+data = load_data("cgan_random_search/")
+n_total = len(data)
+
+outlier_runs = pd.read_csv("cgan_random_search/outliers.csv")
+outlier_runs = outlier_runs[outlier_runs["outlier"]]
+outlier_runs = list(outlier_runs["run"])
+
+data_outlier = dict(filter(lambda i: i[0] in outlier_runs, data.items()))
+n_outlier = len(data_outlier)
+
+print(f"There are {n_outlier} outlier runs in {n_total} total runs which is {100 * n_outlier / n_total:.2f}%")
+print()
+print(f"By Parameters:")
+for param_label, param_groups in parameter_groups.items():
+    param_label = " ".join(map(lambda _str: _str[0].upper() + _str[1:], param_label.split("_")))
+    print(f" - {param_label}")
+
+    for label, value in param_groups.groups.items():
+        n_outlier = len(filter_by_params(data_outlier, value, param_groups.keys))
+        n_total = len(filter_by_params(data, value, param_groups.keys))
+        print(f"   - {label:>12}: {str(n_outlier):>2}/{str(n_total):>2} = {100 * n_outlier / n_total:.2f}%")
+    print()
+