feat(matplotlib): implement calibration-curve

plots/calibration-curve/implementations/matplotlib.py

@@ -0,0 +1,118 @@
+""" pyplots.ai
+calibration-curve: Calibration Curve
+Library: matplotlib 3.10.8 | Python 3.13.11
+Quality: 93/100 | Created: 2025-12-26
+"""
+
+import matplotlib.pyplot as plt
+import numpy as np
+
+
+# Data: Simulate predictions from classifiers with different calibration properties
+np.random.seed(42)
+n_samples = 2000
+n_bins = 10
+
+# Generate ground truth - imbalanced to be realistic (35% positive rate)
+y_true = np.random.binomial(1, 0.35, n_samples)
+
+# Well-calibrated model: predictions closely match true probability
+# Using logistic transformation with moderate noise
+logits_calibrated = 1.2 * (y_true * 2 - 1) + np.random.normal(0, 1.0, n_samples)
+y_prob_calibrated = 1 / (1 + np.exp(-logits_calibrated))
+
+# Overconfident model: pushes predictions toward 0 and 1 (sigmoid with steeper slope)
+logits_over = 2.0 * (y_true * 2 - 1) + np.random.normal(0, 0.5, n_samples)
+y_prob_overconfident = 1 / (1 + np.exp(-logits_over))
+
+# Underconfident model: predictions clustered toward 0.5 (flatter sigmoid)
+logits_under = 0.5 * (y_true * 2 - 1) + np.random.normal(0, 0.8, n_samples)
+y_prob_underconfident = 1 / (1 + np.exp(-logits_under))
+
+# Calculate calibration curves for each model
+bin_edges = np.linspace(0, 1, n_bins + 1)
+
+# Well-calibrated model calibration curve
+bin_idx_cal = np.digitize(y_prob_calibrated, bin_edges[1:-1])
+prob_true_cal = [np.mean(y_true[bin_idx_cal == i]) for i in range(n_bins) if np.sum(bin_idx_cal == i) > 0]
+prob_pred_cal = [np.mean(y_prob_calibrated[bin_idx_cal == i]) for i in range(n_bins) if np.sum(bin_idx_cal == i) > 0]
+
+# Overconfident model calibration curve
+bin_idx_over = np.digitize(y_prob_overconfident, bin_edges[1:-1])
+prob_true_over = [np.mean(y_true[bin_idx_over == i]) for i in range(n_bins) if np.sum(bin_idx_over == i) > 0]
+prob_pred_over = [
+    np.mean(y_prob_overconfident[bin_idx_over == i]) for i in range(n_bins) if np.sum(bin_idx_over == i) > 0
+]
+
+# Underconfident model calibration curve
+bin_idx_under = np.digitize(y_prob_underconfident, bin_edges[1:-1])
+prob_true_under = [np.mean(y_true[bin_idx_under == i]) for i in range(n_bins) if np.sum(bin_idx_under == i) > 0]
+prob_pred_under = [
+    np.mean(y_prob_underconfident[bin_idx_under == i]) for i in range(n_bins) if np.sum(bin_idx_under == i) > 0
+]
+
+# Calculate Brier scores (mean squared error of probability predictions)
+brier_cal = np.mean((y_prob_calibrated - y_true) ** 2)
+brier_over = np.mean((y_prob_overconfident - y_true) ** 2)
+brier_under = np.mean((y_prob_underconfident - y_true) ** 2)
+
+# Create figure with two subplots: calibration curve and histogram
+fig, (ax1, ax2) = plt.subplots(2, 1, figsize=(16, 9), gridspec_kw={"height_ratios": [3, 1]})
+
+# Primary colors from style guide
+python_blue = "#306998"
+python_yellow = "#FFD43B"
+third_color = "#E377C2"  # Colorblind-safe pink/magenta
+
+# Plot calibration curves
+ax1.plot([0, 1], [0, 1], "k--", linewidth=2, label="Perfect Calibration", alpha=0.7)
+ax1.plot(
+    prob_pred_cal,
+    prob_true_cal,
+    "o-",
+    color=python_blue,
+    linewidth=3,
+    markersize=12,
+    label=f"Well-Calibrated (Brier: {brier_cal:.3f})",
+)
+ax1.plot(
+    prob_pred_over,
+    prob_true_over,
+    "s-",
+    color=python_yellow,
+    linewidth=3,
+    markersize=12,
+    label=f"Overconfident (Brier: {brier_over:.3f})",
+)
+ax1.plot(
+    prob_pred_under,
+    prob_true_under,
+    "^-",
+    color=third_color,
+    linewidth=3,
+    markersize=12,
+    label=f"Underconfident (Brier: {brier_under:.3f})",
+)
+
+# Style calibration plot
+ax1.set_xlabel("Mean Predicted Probability", fontsize=20)
+ax1.set_ylabel("Fraction of Positives", fontsize=20)
+ax1.set_title("calibration-curve · matplotlib · pyplots.ai", fontsize=24)
+ax1.tick_params(axis="both", labelsize=16)
+ax1.legend(fontsize=16, loc="lower right")
+ax1.grid(True, alpha=0.3, linestyle="--")
+ax1.set_xlim(0, 1)
+ax1.set_ylim(0, 1)
+
+# Histogram of predicted probabilities
+ax2.hist(y_prob_calibrated, bins=20, alpha=0.6, color=python_blue, label="Well-Calibrated", edgecolor="white")
+ax2.hist(y_prob_overconfident, bins=20, alpha=0.6, color=python_yellow, label="Overconfident", edgecolor="white")
+ax2.hist(y_prob_underconfident, bins=20, alpha=0.6, color=third_color, label="Underconfident", edgecolor="white")
+ax2.set_xlabel("Predicted Probability", fontsize=20)
+ax2.set_ylabel("Count", fontsize=20)
+ax2.tick_params(axis="both", labelsize=16)
+ax2.legend(fontsize=14, loc="upper right")
+ax2.grid(True, alpha=0.3, linestyle="--")
+
+plt.tight_layout()
+plt.savefig("plot.png", dpi=300, bbox_inches="tight")

plots/calibration-curve/metadata/matplotlib.yaml

@@ -0,0 +1,26 @@
+library: matplotlib
+specification_id: calibration-curve
+created: '2025-12-26T19:38:04Z'
+updated: '2025-12-26T19:44:38Z'
+generated_by: claude-opus-4-5-20251101
+workflow_run: 20528201914
+issue: 0
+python_version: 3.13.11
+library_version: 3.10.8
+preview_url: https://storage.googleapis.com/pyplots-images/plots/calibration-curve/matplotlib/plot.png
+preview_thumb: https://storage.googleapis.com/pyplots-images/plots/calibration-curve/matplotlib/plot_thumb.png
+preview_html: null
+quality_score: 93
+review:
+  strengths:
+  - Excellent multi-model comparison showing well-calibrated, overconfident, and underconfident
+    classifiers
+  - Includes histogram subplot as suggested in spec for showing prediction distributions
+  - Brier scores integrated into legend for quick comparison
+  - Clean separation of calibration curve calculation logic
+  - Colorblind-friendly palette with distinct marker shapes for each model
+  - Professional layout with appropriate subplot height ratios
+  weaknesses:
+  - Axis labels lack units or additional context
+  - Could use more distinctive matplotlib features like fill_between for confidence
+    bands