NetworkDismantling · thomas0299 · Mar 27, 2025 · Mar 27, 2025
diff --git a/README.md b/README.md
@@ -141,7 +141,7 @@ You can find them in their sub-folders.
 #### All algorithms except FINDER and GDM
    
 ```bash
-conda create -n dismantling python=3.9 boost boost-cpp graph-tool dill tqdm numpy scipy pandas seaborn matplotlib -c anaconda -c conda-forge
+conda create -n dismantling python=3.9 boost boost-cpp graph-tool dill tqdm numpy scipy pandas seaborn matplotlib parse pyyaml -c anaconda -c conda-forge
 
 conda activate dismantling
 ```
@@ -154,7 +154,7 @@ Please refer to the [PyTorch installation matrix](https://pytorch.org/get-starte
 
 Just as an example, if your host machine supports CUDA (11.8), the command should look like the following:
 ```bash
-conda create --name gdm boost boost-cpp graph-tool dill tqdm numpy scipy pandas seaborn matplotlib python pytorch torchvision torchaudio pytorch-cuda=11.8 pyg -c pyg -c pytorch -c nvidia -c conda-forge -c anaconda
+conda create --name gdm boost boost-cpp graph-tool dill tqdm numpy scipy pandas seaborn matplotlib python parse pyyaml pytorch torchvision torchaudio pytorch-cuda=11.8 pyg -c pyg -c pytorch -c nvidia -c conda-forge -c anaconda
 
 conda activate gdm
 ```

diff --git a/network_dismantling/CoreGDM/core_network_dismantler.py b/network_dismantling/CoreGDM/core_network_dismantler.py
@@ -312,7 +312,7 @@ def test(args, model, early_stopping_dict: dict = None, networks_provider=None,
             "rem_num": np.inf,
         })
 
-        removals, prediction_time, dismantle_time = dismantler(network=network.copy(),
+        removals, prediction_time, dismantle_time, _ = dismantler(network=network.copy(),
                                                                node_generator=predictor,
                                                                generator_args=generator_args,
                                                                stop_condition=stop_condition,

diff --git a/...dismantling/FINDER_ND/python_interface.py → ...ismantling/FINDER_ND/_python_interface.py b/...dismantling/FINDER_ND/python_interface.py → ...ismantling/FINDER_ND/_python_interface.py
diff --git a/network_dismantling/GDM/network_dismantler.py b/network_dismantling/GDM/network_dismantler.py
@@ -205,7 +205,7 @@ def test(args, model, networks_provider, print_model=True, logger=logging.getLog
                          f"Aiming to reach LCC size {stop_condition} ({stop_condition * 100 / network_size:.3f}%)"
                          )
 
-            removals, prediction_time, dismantle_time = dismantler(network, predictor, generator_args, stop_condition)
+            removals, prediction_time, dismantle_time, _ = dismantler(network, predictor, generator_args, stop_condition)
 
             peak_slcc = max(removals, key=itemgetter(4))
 

diff --git a/network_dismantling/GDM/python_interface.py b/network_dismantling/GDM/python_interface.py
@@ -376,6 +376,7 @@ def _GDM(
                 # df=new_df_runs,
                 test_networks={f"{network_name}": network},
                 logger=logger,
+                threshold=args.threshold,
             )
         except FileNotFoundError as e:
             raise RuntimeError(

diff --git a/network_dismantling/GDM/reinsert.py b/network_dismantling/GDM/reinsert.py
@@ -16,6 +16,7 @@
 #   You should have received a copy of the GNU General Public License
 #   along with GDM.  If not, see <http://www.gnu.org/licenses/>.
 
+import re
 import logging
 from ast import literal_eval
 from errno import ENOSPC
@@ -292,9 +293,9 @@ def main(
 
                 run.drop(run_columns, inplace=True, errors="ignore")
                 # Get the removals
-                removals = run.pop("removals")
-                removals = literal_eval(removals)
-
+                original_removals = run.pop("removals")
+                original_removals = re.sub(r'np\.float64\(([^)]+)\)', r'\1', original_removals)
+                original_removals = literal_eval(original_removals)
                 # Remove the columns that are not needed
                 run.drop(run_columns,
                          inplace=True,
@@ -316,18 +317,18 @@ def main(
 
                 if threshold is None:
                     threshold = run.get("threshold",
-                                        removals[-1][3],
+                                        original_removals[-1][3],
                                         )
 
                 stop_condition = int(np.ceil(threshold * network.num_vertices()))
                 generator_args = {
-                    "removals": list(map(itemgetter(1), removals)),
+                    "removals": list(map(itemgetter(1), original_removals)),
                     "stop_condition": stop_condition,
                     "network_name": network_name,
                     "logger": logger,
                 }
 
-                removals, _, _ = dismantler(
+                removals, _, _, last_lcc_size = dismantler(
                     network=network.copy(),
                     predictor=predictor,
                     generator_args=generator_args,
@@ -336,22 +337,6 @@ def main(
                     logger=logger,
                 )
 
-                peak_slcc = max(removals, key=itemgetter(4))
-
-                _run = {
-                    "network": network_name,
-                    "removals": removals,
-                    "slcc_peak_at": peak_slcc[0],
-                    "lcc_size_at_peak": peak_slcc[3],
-                    "slcc_size_at_peak": peak_slcc[4],
-                    "r_auc": simpson(list(r[3] for r in removals), dx=1),
-                    "rem_num": len(removals),
-                    "threshold": threshold,
-                }
-
-                for key, value in _run.items():
-                    run[key] = value
-
                 # Check if something is wrong with the removals
                 if removals[-1][2] == 0:
                     # for removal in removals:
@@ -368,10 +353,29 @@ def main(
                             last_valid_index = i
                        else:
                             break
-
                     logger.error(f"Last valid index: {last_valid_index}: {removals[last_valid_index]}")
-                    raise RuntimeError(f"Had to remove too many nodes ({len(removals)})")
+                    removals = removals[:last_valid_index+1]
+                    # raise RuntimeError(f"Had to remove too many nodes ({len(removals)})")
+
+                if last_lcc_size > stop_condition: # careful size of LCC not actual %, stop condition is number of nodes not threshold
+                    raise RuntimeError(f"Reinsertion did not ensure a proper dismantling with {last_lcc_size=} > {stop_condition=}.")
+
+                peak_slcc = max(removals, key=itemgetter(4))
+
+                _run = {
+                    "network": network_name,
+                    "removals": removals,
+                    "slcc_peak_at": peak_slcc[0],
+                    "lcc_size_at_peak": peak_slcc[3],
+                    "slcc_size_at_peak": peak_slcc[4],
+                    "r_auc": simpson(list(r[3] for r in removals), dx=1),
+                        "rem_num": len(removals),
+                    "threshold": threshold,
+                }
 
+                for key, value in _run.items():
+                    run[key] = value
+
                 all_runs.append(run)
 
                 run_df = pd.DataFrame(data=[run], columns=network_df.columns)

diff --git a/network_dismantling/GDM/training_data_extractor.py b/network_dismantling/GDM/training_data_extractor.py
@@ -55,7 +55,7 @@ def training_data_extractor(g, threshold=None,
         degree = g.get_out_degrees(vertices)
 
         logger.debug("Computing Max degree")
-        max_degree = np.max(degree)
+        max_degree = np.max(degree) if len(degree) > 0 else 1
 
         logger.debug("Normalizing the degree")
         degree = np.divide(degree, max_degree)

diff --git a/network_dismantling/common/dismantlers.py b/network_dismantling/common/dismantlers.py
@@ -85,7 +85,7 @@ def threshold_dismantler(
             logger.debug("EARLY STOPPING")
             break
 
-    return removals, None, None
+    return removals, None, None, None
 
 
 # TODO REMOVE THIS FROM THE REVIEW. IT IS NOT USED!
@@ -183,7 +183,7 @@ def kcore_lcc_threshold_dismantler(
             # print("EARLY STOPPING")
             break
 
-    return removals, None, None
+    return removals, None, None, None
 
 
 def lcc_threshold_dismantler(
@@ -270,7 +270,7 @@ def lcc_threshold_dismantler(
             generator.close()
             break
 
-    return removals, None, None
+    return removals, None, None, None
 
 
 def lcc_peak_dismantler(
@@ -380,7 +380,7 @@ def lcc_peak_dismantler(
 
     generator.close()
 
-    return removals, None, None  # prediction_time, dismantle_time
+    return removals, None, None, None  # prediction_time, dismantle_time
 
 
 def enqueued(original_function=None,
@@ -454,7 +454,7 @@ def wrapper(
             generator_args["sorting_function"] = function
 
             # kwargs["generator_args"] = generator_args
-            removals, prediction_time, dismantle_time = dismantler(
+            removals, prediction_time, dismantle_time, _ = dismantler(
                 network=network,
                 predictor=predictor,
                 generator_args=generator_args,

diff --git a/network_dismantling/common/external_dismantlers/lcc_threshold_dismantler.py b/network_dismantling/common/external_dismantlers/lcc_threshold_dismantler.py
@@ -118,7 +118,7 @@ def _threshold_dismantler(
 
     del predictions_dict
 
-    return removals, prediction_time, dismantle_time
+    return removals, prediction_time, dismantle_time, lcc_size
 
 
 def lcc_threshold_dismantler(
@@ -227,7 +227,7 @@ def _iterative_threshold_dismantler(network, predictor, generator_args, stop_con
         f"{network_name}: iterative external dismantler returned in {dismantle_time}s"
     )
 
-    return removals, None, None
+    return removals, None, None, lcc_size
 
 
 def iterative_threshold_dismantler(network, predictor, generator_args, stop_condition):

diff --git a/network_dismantling/dismantler.py b/network_dismantling/dismantler.py
@@ -473,6 +473,9 @@ def main(args, logger=logging.getLogger("dummy")):
                             f"Dismantling {network_name} according to {display_name}. "
                             f"Aiming to LCC size {stop_condition} ({stop_condition / network_size:.3f})"
                         )
+
+                        if network_size > args.max_num_vertices:
+                            continue
                         # logger.debug(f"dismantling_method_kwargs: {dismantling_method_kwargs}")
 
                         generator_args["executor"] = executor
@@ -657,6 +660,14 @@ def get_df_columns():
         help="Number of jobs.",
     )
 
+    parser.add_argument(
+        "-mnv",
+        "--max_num_vertices",
+        type=int,
+        default=np.inf,
+        help="Filter for network size based on number of vertices.",
+    )
+
     # parser.add_argument(
     #     "-sa",
     #     "--simultaneous_access",

diff --git a/network_dismantling/heuristics/dismantler.py b/network_dismantling/heuristics/dismantler.py
@@ -224,7 +224,7 @@ def main(args):
                                                                                               "STATIC" if mode is True else "DYNAMIC") + " " + display_name,
                                                                                           stop_condition,
                                                                                           stop_condition / network_size))
-                removals, prediction_time, dismantle_time = dismantler(network=network.copy(),
+                removals, prediction_time, dismantle_time, _ = dismantler(network=network.copy(),
                                                                        predictor=generator,
                                                                        generator_args=generator_args,
                                                                        stop_condition=stop_condition,

diff --git a/network_dismantling/multiscale_entanglement/reinsertion/reinsert.py b/network_dismantling/multiscale_entanglement/reinsertion/reinsert.py
@@ -330,7 +330,7 @@ def main(
                     "logger": logger,
                 }
 
-                removals, _, _ = dismantler(
+                removals, _, _, _ = dismantler(
                     network=network.copy(),
                     predictor=predictor,
                     generator_args=generator_args,

diff --git a/network_dismantling/vertex_entanglement/reinsertion/reinsert.py b/network_dismantling/vertex_entanglement/reinsertion/reinsert.py
@@ -330,7 +330,7 @@ def main(
                     "logger": logger,
                 }
 
-                removals, _, _ = dismantler(
+                removals, _, _, _ = dismantler(
                     network=network.copy(),
                     predictor=predictor,
                     generator_args=generator_args,