General improvements

list_distances.py

@@ -3,12 +3,17 @@ import pandas as pd
 import numpy as np
 from mlpy import canberra_stability
 from itertools import combinations
+from pathlib import Path
 
 parser = argparse.ArgumentParser()
 parser.add_argument('--resultsdir', type=str, help='Results folder')
 parser.add_argument('--dataset', type=str, help='Dataset name')
-parser.add_argument('--model', type=str, default='randomForest', help='Model (default: %default)')
 parser.add_argument('--target', type=str, help='Clinical endpoint')
+parser.add_argument('--model', type=str, default='randomForest', help='Model (default: %(default)s)')
+parser.add_argument('--nf_min', type=int, default=10, help='Min #feat (default: %(default)s)')
+parser.add_argument('--nf_max', type=int, default=50, help='Max #feat (default: %(default)s)')
+parser.add_argument('--nf_step', type=int, default=10, help='Increase by these many feat (default: %(default)s)')
+parser.add_argument('--nf_rsnf', type=int, nargs='+', help='One or more #feat for rSNF')
 parser.add_argument('--layers', type=str, nargs='+', help='')
 args = parser.parse_args()
 
@@ -16,10 +21,22 @@ RESULTSDIR = args.resultsdir # top-level results directory
 DATASET = args.dataset # 'tcga_breast'
 TARGET = args.target # 'ER'
 MODEL = args.model
+NF_MIN = args.nf_min
+NF_MAX = args.nf_max
+NF_STEP = args.nf_step
+NF_RSNF = args.nf_rsnf
 LAYERS = args.layers
 N_LAYERS = len(LAYERS)
 MODE = 'rSNF'
 
+assert(
+    Path(RESULTSDIR, DATASET).expanduser().exists()
+), f"{RESULTSDIR}/{DATASET} not found"
+
+assert(
+    Path(RESULTSDIR, f"{DATASET}_SNFdap").expanduser().exists()
+), f"{RESULTSDIR}/{DATASET}_SNFdap not found"
+
 for k in range(2, N_LAYERS+1):
     for comb in combinations(LAYERS, k):
         layers_concat = '_'.join(comb)
@@ -30,7 +47,13 @@ for k in range(2, N_LAYERS+1):
         # prepare ranks for canberra_stability
         ranks = pd.concat([np.argsort(bordas[0]['FEATURE_ID']),
                            np.argsort(bordas[1]['FEATURE_ID'])], axis=1).transpose().values
-        for nf in np.arange(10, 60, 10):
+        for nf in np.arange(NF_MIN, NF_MAX + NF_STEP, NF_STEP):
+            cs = canberra_stability(ranks, nf)
+            print(f'{MODE} - {layers_concat} - stability({nf}) = {cs:.3f}')
+        # additional steps for NF_RSNF
+        print()
+        for nf in NF_RSNF:
             cs = canberra_stability(ranks, nf)
             print(f'{MODE} - {layers_concat} - stability({nf}) = {cs:.3f}')
         print()
+        print()