bug fix

experiments/mu_G/debug.py

@@ -0,0 +1,16 @@
+from pathlib import Path
+import sys
+
+import matplotlib.pyplot as plt
+import pandas as pd
+
+sys.path.append(str(Path(__file__).parents[2]))
+
+from src.utils import plot_results
+
+df = pd.read_csv("mu_G_results.csv")
+fig = plot_results(df, ["mu", "G_free_energy"], [True, False])
+fig.tight_layout()
+# plt.savefig("mu_G_results.png")
+plt.show()
+

experiments/mu_G/plot.py

@@ -0,0 +1,16 @@
+from pathlib import Path
+import sys
+
+import matplotlib.pyplot as plt
+import pandas as pd
+
+sys.path.append(str(Path(__file__).parents[2]))
+
+from src.utils import plot_results
+
+df = pd.read_csv("mu_G_results.csv")
+fig = plot_results(df, ["mu", "G_free_energy"], [True, False])
+fig.tight_layout()
+plt.savefig("mu_G_results.png")
+plt.show()
+

experiments/mu_gap/plot.py

@@ -0,0 +1,16 @@
+from pathlib import Path
+import sys
+
+import matplotlib.pyplot as plt
+import pandas as pd
+
+sys.path.append(str(Path(__file__).parents[2]))
+
+from src.utils import plot_results
+
+df = pd.read_csv("mu_gap_results.csv")
+fig = plot_results(df, ["mu", "gap"], [True, False])
+fig.tight_layout()
+plt.savefig("mu_gap_results.png")
+plt.show()
+

src/utils.py

@@ -21,14 +21,14 @@ def plot_results(df, targets, target_bool):
     fig, ax = plt.subplots(2, 2, figsize=(20,10))
     ax = ax.flatten()
     global_values = torch.tensor(df[targets].values)
-    for mask in target_bool:
+    for i, mask in enumerate(target_bool):
         if not mask:
-            global_values[:, mask] *= -1
+            global_values[:, i] *= -1
     global_pareto_idx = is_non_dominated(global_values)
     global_pareto_front = global_values[global_pareto_idx]
-    for mask in target_bool:
+    for i, mask in enumerate(target_bool):
         if not mask:
-            global_pareto_front[:, mask] *= -1
+            global_pareto_front[:, i] *= -1
     ref_point = global_values.min(0)[0]
     hv = DominatedPartitioning(ref_point=ref_point, Y=global_values)
     global_hv = hv.compute_hypervolume().item()
@@ -40,9 +40,9 @@ def plot_results(df, targets, target_bool):
         for iter_no in range(max_iterations + 1):
             current_data = df[df[f'iteration_{acq}'] <= iter_no]
             local_values = torch.tensor(current_data[targets].values)
-            for mask in target_bool:
+            for i, mask in enumerate(target_bool):
                 if not mask:
-                    local_values[:, mask] *= -1
+                    local_values[:, i] *= -1
             hv = DominatedPartitioning(ref_point=ref_point, Y=local_values)
             hv_iterations[acq].append(hv.compute_hypervolume().item())
     ax[0].plot(np.array(hv_iterations["qEHVI"]) / global_hv * 100, marker='o', label="qEHVI")
@@ -57,14 +57,14 @@ def plot_results(df, targets, target_bool):
     max_iterations = int(df['iteration_qEHVI'].max())
     current_data = df[df['iteration_qEHVI'] <= max_iterations]
     qEHVI_values = torch.tensor(current_data[targets].values)
-    for mask in target_bool:
+    for i, mask in enumerate(target_bool):
         if not mask:
-            qEHVI_values[:, mask] *= -1
+            qEHVI_values[:, i] *= -1
     qEHVI_pareto_idx = is_non_dominated(qEHVI_values)
     qEHVI_pareto_front = qEHVI_values[qEHVI_pareto_idx]
-    for mask in target_bool:
+    for i, mask in enumerate(target_bool):
         if not mask:
-            qEHVI_pareto_front[:, mask] *= -1
+            qEHVI_pareto_front[:, i] *= -1
     ax[1].scatter(qEHVI_pareto_front[:, 0], qEHVI_pareto_front[:, 1],
         color='red', label="qEHVI Pareto Front", marker="*", s=100)
     ax[1].scatter(df[targets[0]], df[targets[1]], alpha=0.3)
@@ -79,14 +79,14 @@ def plot_results(df, targets, target_bool):
     max_iterations = int(df['iteration_qNEHVI'].max())
     current_data = df[df['iteration_qNEHVI'] <= max_iterations]
     qNEHVI_values = torch.tensor(current_data[targets].values)
-    for mask in target_bool:
+    for i, mask in enumerate(target_bool):
         if not mask:
-            qNEHVI_values[:, mask] *= -1
+            qNEHVI_values[:, i] *= -1
     qNEHVI_pareto_idx = is_non_dominated(qNEHVI_values)
     qNEHVI_pareto_front = qNEHVI_values[qNEHVI_pareto_idx]
-    for mask in target_bool:
+    for i, mask in enumerate(target_bool):
         if not mask:
-            qNEHVI_pareto_front[:, mask] *= -1
+            qNEHVI_pareto_front[:, i] *= -1
     ax[2].scatter(qNEHVI_pareto_front[:, 0], qNEHVI_pareto_front[:, 1],
         color='red', label="qNEHVI Pareto Front", marker="*", s=100)
     ax[2].scatter(df[targets[0]], df[targets[1]], alpha=0.3)
@@ -101,14 +101,14 @@ def plot_results(df, targets, target_bool):
     max_iterations = int(df['iteration_random'].max())
     current_data = df[df['iteration_random'] <= max_iterations]
     random_values = torch.tensor(current_data[targets].values)
-    for mask in target_bool:
+    for i, mask in enumerate(target_bool):
         if not mask:
-            random_values[:, mask] *= -1
+            random_values[:, i] *= -1
     random_pareto_idx = is_non_dominated(random_values)
     random_pareto_front = random_values[random_pareto_idx]
-    for mask in target_bool:
+    for i, mask in enumerate(target_bool):
         if not mask:
-            random_pareto_front[:, mask] *= -1
+            random_pareto_front[:, i] *= -1
     ax[3].scatter(random_pareto_front[:, 0], random_pareto_front[:, 1],
         color='red', label="Random Pareto Front", marker="*", s=100)
     ax[3].scatter(df[targets[0]], df[targets[1]], alpha=0.3)