Create stability_visualizer.py

stability_visualizer.py

@@ -0,0 +1,134 @@
+import csv
+import argparse
+import matplotlib.pyplot as plt
+from matplotlib.lines import Line2D
+
+K = 1.0
+BOUNDARY_EPS = 0.02
+
+
+def load_csv(path):
+    with open(path, newline="", encoding="utf-8") as f:
+        return list(csv.DictReader(f))
+
+
+def index_by_id(rows):
+    index = {}
+    for row in rows:
+        sid = str(row.get("scenario_id", "")).strip()
+        if not sid:
+            raise ValueError("Blank or missing scenario_id detected")
+        if sid in index:
+            raise ValueError(f"Duplicate scenario_id detected: {sid}")
+        index[sid] = row
+    return index
+
+
+def parse_binary(value, field_name, sid):
+    try:
+        parsed = int(value)
+    except ValueError as e:
+        raise ValueError(f"Invalid {field_name} for {sid}: {value}") from e
+    if parsed not in (0, 1):
+        raise ValueError(f"Invalid {field_name} for {sid}: {parsed}")
+    return parsed
+
+
+def compute_surfaces(pressure, buffer, lag, coupling):
+    s1 = buffer - (pressure * coupling) - (K * lag)
+    s2 = buffer - (pressure * (coupling ** 2)) - (K * lag)
+    s3 = buffer - (pressure * coupling) - (K * (lag ** 2))
+    return {
+        "baseline_surface": s1,
+        "coupling_surface": s2,
+        "lag_surface": s3,
+    }
+
+
+def plot_projection(pred_path, truth_path):
+    preds = load_csv(pred_path)
+    truth = load_csv(truth_path)
+
+    pred_map = index_by_id(preds)
+    truth_map = index_by_id(truth)
+
+    if set(pred_map.keys()) != set(truth_map.keys()):
+        raise ValueError("scenario_id mismatch between prediction and truth files")
+
+    plt.figure(figsize=(10, 8))
+
+    for sid in sorted(truth_map):
+        truth_row = truth_map[sid]
+        pred_row = pred_map[sid]
+
+        pressure = float(truth_row["pressure"])
+        buffer = float(truth_row["buffer"])
+        lag = float(truth_row["lag"])
+        coupling = float(truth_row["coupling"])
+
+        pred = parse_binary(pred_row["prediction"], "prediction", sid)
+        label = parse_binary(truth_row["label_stable"], "label_stable", sid)
+
+        surfaces = compute_surfaces(pressure, buffer, lag, coupling)
+        manifold_margin = min(surfaces.values())
+        active_surface = min(surfaces, key=surfaces.get)
+
+        x = pressure * coupling
+        y = buffer
+
+        if abs(manifold_margin) <= BOUNDARY_EPS:
+            color = "orange"
+            marker = "s"
+        elif manifold_margin < 0:
+            color = "red"
+            marker = "x"
+        else:
+            color = "green"
+            marker = "o"
+
+        if label == 0 and pred == 1:
+            color = "purple"
+            marker = "*"
+        elif label == 1 and pred == 0:
+            color = "blue"
+            marker = "D"
+
+        size = 110
+        edgecolor = "black" if active_surface == "coupling_surface" else "none"
+        linewidth = 1.2 if active_surface == "coupling_surface" else 0.0
+
+        plt.scatter(
+            x,
+            y,
+            c=color,
+            marker=marker,
+            s=size,
+            edgecolors=edgecolor,
+            linewidths=linewidth,
+        )
+
+    plt.xlabel("Pressure × Coupling")
+    plt.ylabel("Buffer Capacity")
+    plt.title("Clarus Stability Manifold Projection (2D View)")
+    plt.grid(True)
+    plt.tight_layout()
+
+    legend_elements = [
+        Line2D([0], [0], marker="o", color="w", label="Stable region", markerfacecolor="green", markersize=10),
+        Line2D([0], [0], marker="x", color="red", label="Collapse region", linestyle="None", markersize=10),
+        Line2D([0], [0], marker="s", color="w", label="Near boundary", markerfacecolor="orange", markersize=10),
+        Line2D([0], [0], marker="*", color="w", label="False rescue", markerfacecolor="purple", markersize=12),
+        Line2D([0], [0], marker="D", color="w", label="False collapse", markerfacecolor="blue", markersize=10),
+        Line2D([0], [0], marker="o", color="black", label="Coupling surface active", markerfacecolor="white", markersize=9),
+    ]
+    plt.legend(handles=legend_elements)
+    plt.show()
+
+
+if __name__ == "__main__":
+    parser = argparse.ArgumentParser()
+    parser.add_argument("--pred", required=True)
+    parser.add_argument("--truth", required=True)
+    args = parser.parse_args()
+
+    plot_projection(args.pred, args.truth)