feat(awa): add 3d performance metrics and enhanced visualization

Author: MichalRedm

scripts/experiment_awa.py

@@ -185,6 +185,9 @@ def run_awa_experiment():
     ref_metrics_3d, _ = ecm.generate_pareto_front_3d(
         num_points_per_dim=15, **data_kwargs
     )
+    ref_front_3d = np.column_stack(
+        [ref_metrics_3d["Return"], ref_metrics_3d["Risk"], ref_metrics_3d["Diversification"]]
+    )
 
     print("  Evaluating MOEA/D (3D)...")
     moead_3d = MOEADOptimizer(problem=prob_3d)
@@ -195,6 +198,19 @@ def run_awa_experiment():
         **moead_kwargs,
         **data_kwargs,
     )
+    front_moead_3d = np.column_stack(
+        [
+            metrics_moead_3d["Return"],
+            metrics_moead_3d["Risk"],
+            metrics_moead_3d["Diversification"],
+        ]
+    )
+
+    m_dict_3d_1 = {
+        "IGD": calculate_igd(front_moead_3d, ref_front_3d),
+        "Spacing": calculate_spacing(front_moead_3d),
+        "Spread": calculate_spread(front_moead_3d, ref_front_3d),
+    }
 
     print("  Evaluating MOEA/D-AWA (3D)...")
     awa_3d = MOEADAWAOptimizer(problem=prob_3d)
@@ -206,6 +222,26 @@ def run_awa_experiment():
         **moead_kwargs,
         **data_kwargs,
     )
+    front_awa_3d = np.column_stack(
+        [
+            metrics_awa_3d["Return"],
+            metrics_awa_3d["Risk"],
+            metrics_awa_3d["Diversification"],
+        ]
+    )
+
+    m_dict_3d_2 = {
+        "IGD": calculate_igd(front_awa_3d, ref_front_3d),
+        "Spacing": calculate_spacing(front_awa_3d),
+        "Spread": calculate_spread(front_awa_3d, ref_front_3d),
+    }
+
+    print(
+        f"  3D MOEA/D     - IGD: {m_dict_3d_1['IGD']:.6f}, Spacing: {m_dict_3d_1['Spacing']:.6f}, Spread: {m_dict_3d_1['Spread']:.6f}"
+    )
+    print(
+        f"  3D MOEA/D-AWA - IGD: {m_dict_3d_2['IGD']:.6f}, Spacing: {m_dict_3d_2['Spacing']:.6f}, Spread: {m_dict_3d_2['Spread']:.6f}"
+    )
 
     from src.visualization import plot_two_variants_comparison_3d
 
@@ -215,7 +251,9 @@ def run_awa_experiment():
         metrics_awa_3d,
         name1="Standard MOEA/D",
         name2="MOEA/D-AWA",
-        save_path=os.path.join(run_folder, "exp4_3d_comparison.gif"),
+        metrics_dict1=m_dict_3d_1,
+        metrics_dict2=m_dict_3d_2,
+        save_path=os.path.join(run_folder, "awa_comparison_3d.gif"),
     )
 
     print(f"AWA Experiments finished. Results in {run_folder}")

src/visualization.py

@@ -776,82 +776,112 @@ def plot_two_variants_comparison_3d(
     metrics2,
     name1="Standard MOEA/D",
     name2="Variant",
+    metrics_dict1=None,
+    metrics_dict2=None,
     save_path=None,
 ):
     """
-    Compares two algorithm variants against ECM in two separate 3D subplots.
+    Compares two algorithm variants against ECM in two separate 3D subplots,
+    including performance metrics in the legend.
     """
-    fig = plt.figure(figsize=(18, 8))
+    fig = plt.figure(figsize=(20, 9))
     ax1 = fig.add_subplot(121, projection="3d")
     ax2 = fig.add_subplot(122, projection="3d")
+    
+    # Define color palette
+    c_ref = "#333333"
+    c1 = "#1f77b4"
+    c2 = "#ff7f0e"
+
+    def format_label(name, m_dict):
+        if not m_dict:
+            return name
+        m_str = "\n".join([f"{k}: {v:.4f}" for k, v in m_dict.items()])
+        return f"{name}\n{m_str}"
 
     # Subplot 1: Variant 1 vs ECM
     if ref_metrics_3d is not None:
         ax1.scatter(
             ref_metrics_3d["Return"],
             ref_metrics_3d["Risk"],
             ref_metrics_3d["Diversification"],
-            c="black",
+            c=c_ref,
             marker="x",
-            s=20,
+            s=25,
             label="Reference (ECM)",
-            alpha=0.2,
+            alpha=0.25,
+            zorder=1,
         )
+    
+    label1 = format_label(name1, metrics_dict1)
     ax1.scatter(
         metrics1["Return"],
         metrics1["Risk"],
         metrics1["Diversification"],
         marker="o",
-        s=40,
-        label=name1,
-        alpha=0.6,
-        color="#1f77b4",
+        s=50,
+        label=label1,
+        alpha=0.7,
+        color=c1,
+        edgecolors="white",
+        linewidth=0.5,
+        zorder=2,
     )
-    ax1.set_title(f"{name1} vs ECM")
-    ax1.set_xlabel("Return")
-    ax1.set_ylabel("Risk")
-    ax1.set_zlabel("Diversification")
-    ax1.legend()
+    ax1.set_title(f"{name1} Performance (3D)", fontsize=14, pad=20)
+    ax1.set_xlabel("Return", fontsize=11)
+    ax1.set_ylabel("Risk", fontsize=11)
+    ax1.set_zlabel("Diversification", fontsize=11)
+    ax1.legend(loc="upper left", frameon=True, shadow=True, fontsize=10)
+    ax1.view_init(elev=25, azim=45)
 
     # Subplot 2: Variant 2 vs ECM
     if ref_metrics_3d is not None:
         ax2.scatter(
             ref_metrics_3d["Return"],
             ref_metrics_3d["Risk"],
             ref_metrics_3d["Diversification"],
-            c="black",
+            c=c_ref,
             marker="x",
-            s=20,
+            s=25,
             label="Reference (ECM)",
-            alpha=0.2,
+            alpha=0.25,
+            zorder=1,
         )
+    
+    label2 = format_label(name2, metrics_dict2)
     ax2.scatter(
         metrics2["Return"],
         metrics2["Risk"],
         metrics2["Diversification"],
         marker="o",
-        s=40,
-        label=name2,
-        alpha=0.6,
-        color="#ff7f0e",
+        s=50,
+        label=label2,
+        alpha=0.7,
+        color=c2,
+        edgecolors="white",
+        linewidth=0.5,
+        zorder=2,
     )
-    ax2.set_title(f"{name2} vs ECM")
-    ax2.set_xlabel("Return")
-    ax2.set_ylabel("Risk")
-    ax2.set_zlabel("Diversification")
-    ax2.legend()
-
-    # Animation function: Rotate both views
-    def update(frame):
-        ax1.view_init(elev=20, azim=frame)
-        ax2.view_init(elev=20, azim=frame)
-        return (fig,)
+    ax2.set_title(f"{name2} Performance (3D)", fontsize=14, pad=20)
+    ax2.set_xlabel("Return", fontsize=11)
+    ax2.set_ylabel("Risk", fontsize=11)
+    ax2.set_zlabel("Diversification", fontsize=11)
+    ax2.legend(loc="upper left", frameon=True, shadow=True, fontsize=10)
+    ax2.view_init(elev=25, azim=45)
 
-    ani = FuncAnimation(fig, update, frames=np.arange(0, 360, 4), interval=50)
+    plt.suptitle("3D Pareto Front Quality Comparison (AWA Optimization)", fontsize=16, y=0.98)
+    plt.tight_layout(rect=[0, 0.03, 1, 0.95])
 
     if save_path:
         if save_path.endswith(".gif"):
-            ani.save(save_path, writer="pillow", fps=15)
+            def update(frame):
+                ax1.view_init(elev=25, azim=frame)
+                ax2.view_init(elev=25, azim=frame)
+                return fig,
+
+            from matplotlib.animation import FuncAnimation
+            ani = FuncAnimation(fig, update, frames=np.arange(0, 360, 5), interval=100)
+            ani.save(save_path, writer="pillow", dpi=100)
         else:
-            plt.savefig(save_path)
+            plt.savefig(save_path, dpi=150, bbox_inches="tight")
     plt.close()