Update app.py

app.py

@@ -36,6 +36,8 @@ REPO_CONFIG = {
 }
 
 TARGET_COLS = ["error_Z_global", "error_X_global", "error_Y_global"]
+IDEAL_COLS = ["ideal_expval_Z_global", "ideal_expval_X_global", "ideal_expval_Y_global"]
+NOISY_COLS = ["noisy_expval_Z_global", "noisy_expval_X_global", "noisy_expval_Y_global"]
 
 NON_FEATURE_COLS = {
     "sample_id", "sample_seed", "circuit_hash", "split", "circuit_qasm",
@@ -43,13 +45,10 @@ NON_FEATURE_COLS = {
     "noise_type", "noise_prob", "observable_bases", "observable_mode", "shots",
     "gpu_requested", "gpu_available", "backend_device", "precision_mode",
     "circuit_signature", "noise_label",
-    "ideal_expval_Z_global", "noisy_expval_Z_global",
-    "ideal_expval_X_global", "noisy_expval_X_global",
-    "ideal_expval_Y_global", "noisy_expval_Y_global",
+    *IDEAL_COLS, *NOISY_COLS, *TARGET_COLS,
     "sign_ideal_Z_global", "sign_noisy_Z_global",
     "sign_ideal_X_global", "sign_noisy_X_global",
     "sign_ideal_Y_global", "sign_noisy_Y_global",
-    *TARGET_COLS,
 }
 
 SOFT_EXCLUDE_PATTERNS = ["ideal_", "noisy_", "sign_ideal_", "sign_noisy_"]
@@ -59,9 +58,7 @@ _ASSET_CACHE: Dict[str, pd.DataFrame] = {}
 # ========================= HELPERS =========================
 
 def load_guide_content() -> str:
-    """
-    Read the GUIDE.md file from the root directory.
-    """
+    """Read the GUIDE.md file from the root directory."""
     try:
         with open("GUIDE.md", "r", encoding="utf-8") as f:
             return f.read()
@@ -184,10 +181,18 @@ def default_feature_selection(features: List[str]) -> List[str]:
     selected = [f for f in preferred if f in features]
     return selected[:10] if selected else features[:10]
 
-def make_regression_figure(y_true: np.ndarray, y_pred: np.ndarray, basis: str) -> plt.Figure:
-    """Generate diagnostic regression plots."""
-    fig, axs = plt.subplots(1, 2, figsize=(14, 6))
-    axs[0].scatter(y_true, y_pred, alpha=0.6, s=15)
+def make_regression_figure(
+    y_true: np.ndarray, 
+    y_pred: np.ndarray, 
+    ideal_vals: np.ndarray, 
+    noisy_vals: np.ndarray, 
+    basis: str
+) -> plt.Figure:
+    """Generate diagnostic regression plots including physics emulation."""
+    fig, axs = plt.subplots(1, 3, figsize=(20, 6))
+
+    # 1. Error Prediction (Predicted vs True)
+    axs[0].scatter(y_true, y_pred, alpha=0.6, s=15, color='#3498db')
     min_v, max_v = min(y_true.min(), y_pred.min()), max(y_true.max(), y_pred.max())
     axs[0].plot([min_v, max_v], [min_v, max_v], 'r--', lw=2)
     axs[0].set_xlabel("True Error")
@@ -195,14 +200,27 @@ def make_regression_figure(y_true: np.ndarray, y_pred: np.ndarray, basis: str) -
     axs[0].set_title(f"{basis} Error: Predicted vs True")
     axs[0].grid(True, alpha=0.3)
 
+    # 2. Residual Distribution
     residuals = y_true - y_pred
-    axs[1].hist(residuals, bins=50, alpha=0.7, color="skyblue", edgecolor="black")
+    axs[1].hist(residuals, bins=50, alpha=0.7, color="#2ecc71", edgecolor="black")
     axs[1].axvline(0, color="red", linestyle="--")
     axs[1].set_xlabel("Residual")
     axs[1].set_ylabel("Count")
-    axs[1].set_title(f"{basis} Residual Distribution")
+    axs[1].set_title(f"{basis} Error Residuals")
     axs[1].grid(True, alpha=0.3)
 
+    # 3. Physics Emulation (Ideal vs Noisy Expectation Values)
+    pred_noisy_vals = ideal_vals + y_pred
+    
+    axs[2].scatter(ideal_vals, noisy_vals, alpha=0.4, s=15, label="Actual Noisy (Simulated)", color="#95a5a6")
+    axs[2].scatter(ideal_vals, pred_noisy_vals, alpha=0.6, s=15, label="Predicted Noisy (ML)", color="#e74c3c")
+    axs[2].plot([-1, 1], [-1, 1], 'k--', lw=1, alpha=0.7, label="No Noise Limit")
+    axs[2].set_xlabel("Ideal Expectation Value")
+    axs[2].set_ylabel("Noisy Expectation Value")
+    axs[2].set_title(f"Physics Emulation: {basis} Basis Shift")
+    axs[2].legend()
+    axs[2].grid(True, alpha=0.3)
+
     fig.tight_layout()
     return fig
 
@@ -218,7 +236,7 @@ def train_regressor(
         return None, "### ❌ Please select at least one feature.", None, None
 
     df = load_single_dataset()
-    required_cols = feature_columns + TARGET_COLS
+    required_cols = feature_columns + TARGET_COLS + IDEAL_COLS + NOISY_COLS
     train_df = df.dropna(subset=required_cols).copy()
 
     if len(train_df) < 50:
@@ -230,9 +248,15 @@ def train_regressor(
     seed = int(random_state)
     depth = int(max_depth) if max_depth and int(max_depth) > 0 else None
 
-    X_train, X_test, y_train, y_test = train_test_split(
-        X, y, test_size=test_size, random_state=seed
-    )
+    # Track indices to extract ideal and noisy arrays for the test set later
+    indices = np.arange(len(train_df))
+    idx_train, idx_test = train_test_split(indices, test_size=test_size, random_state=seed)
+
+    X_train, X_test = X.iloc[idx_train], X.iloc[idx_test]
+    y_train, y_test = y.iloc[idx_train], y.iloc[idx_test]
+    
+    ideal_test = train_df[IDEAL_COLS].iloc[idx_test].values
+    noisy_test = train_df[NOISY_COLS].iloc[idx_test].values
 
     model = Pipeline([
         ("imputer", SimpleImputer(strategy="median")),
@@ -264,9 +288,10 @@ def train_regressor(
         f"**Y-error** — MAE: {mae[2]:.5f} | RMSE: {rmse[2]:.5f} | R²: {r2[2]:.4f}\n"
     )
 
-    fig_z = make_regression_figure(y_test.iloc[:, 0].values, y_pred[:, 0], "Z")
-    fig_x = make_regression_figure(y_test.iloc[:, 1].values, y_pred[:, 1], "X")
-    fig_y = make_regression_figure(y_test.iloc[:, 2].values, y_pred[:, 2], "Y")
+    # Generate figures passing ideal and true noisy data
+    fig_z = make_regression_figure(y_test.iloc[:, 0].values, y_pred[:, 0], ideal_test[:, 0], noisy_test[:, 0], "Z")
+    fig_x = make_regression_figure(y_test.iloc[:, 1].values, y_pred[:, 1], ideal_test[:, 1], noisy_test[:, 1], "X")
+    fig_y = make_regression_figure(y_test.iloc[:, 2].values, y_pred[:, 2], ideal_test[:, 2], noisy_test[:, 2], "Y")
 
     return fig_z, metrics_text, fig_x, fig_y
 
@@ -363,9 +388,9 @@ with gr.Blocks(title=APP_TITLE) as demo:
             run_btn = gr.Button("🚀 Train Multi-Output Regressor", variant="primary")
 
             with gr.Row():
-                plot_z = gr.Plot(label="Z Error")
-                plot_x = gr.Plot(label="X Error")
-            plot_y = gr.Plot(label="Y Error")
+                plot_z = gr.Plot(label="Z Error Metrics")
+                plot_x = gr.Plot(label="X Error Metrics")
+            plot_y = gr.Plot(label="Y Error Metrics")
             metrics = gr.Markdown()
 
         with gr.TabItem("📖 Guide"):