File size: 5,733 Bytes
32fd782 | 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 | """
Visualize AFRES v2 results.
"""
import json
import numpy as np
import matplotlib
matplotlib.use('Agg')
import matplotlib.pyplot as plt
def load(path="/app/afres_v2_results.json"):
with open(path) as f:
return json.load(f)
def plot_search_history(results, save="/app/v2_search_history.png"):
hist = results.get("search_history", [])
if not hist:
return
iters = [h["iteration"] for h in hist]
fitnesses = [h.get("fitness", 0) for h in hist]
maes = [h.get("mean_mae", 0) for h in hist]
ics = [h.get("mean_ic", 0) for h in hist]
fig, (ax1, ax2, ax3) = plt.subplots(1, 3, figsize=(18, 5))
ax1.scatter(iters, fitnesses, c='steelblue', alpha=0.6, s=30)
ax1.plot(sorted(set(iters)), [max(f for i, f in zip(iters, fitnesses) if i == ii)
for ii in sorted(set(iters))], 'r-', linewidth=2)
ax1.set_xlabel("Iteration")
ax1.set_ylabel("Fitness (Mean IC)")
ax1.set_title("Rubric Search: Fitness Landscape")
ax1.grid(True, alpha=0.3)
ax2.scatter(iters, ics, c='green', alpha=0.6, s=30)
ax2.set_xlabel("Iteration")
ax2.set_ylabel("Mean IC")
ax2.set_title("Mean IC per Evaluated Rubric")
ax2.grid(True, alpha=0.3)
ax3.scatter(iters, maes, c='orange', alpha=0.6, s=30)
ax3.set_xlabel("Iteration")
ax3.set_ylabel("Mean MAE")
ax3.set_title("Mean Regression MAE per Evaluated Rubric")
ax3.grid(True, alpha=0.3)
plt.tight_layout()
plt.savefig(save, dpi=150)
plt.close()
print(f"Saved {save}")
def plot_rubric_composition(results, save="/app/v2_rubric_composition.png"):
rubric = results.get("best_rubric", {})
items = rubric.get("items", [])
if not items:
return
fig, ax = plt.subplots(figsize=(10, 6))
labels = [i["id"] for i in items]
n_features = [len(i.get("feature_focus", [])) for i in items]
n_ops = [len(i.get("preferred_ops", [])) for i in items]
x = np.arange(len(labels))
width = 0.35
ax.bar(x - width/2, n_features, width, label='# Features', color='steelblue')
ax.bar(x + width/2, n_ops, width, label='# Operators', color='coral')
ax.set_xticks(x)
ax.set_xticklabels(labels, rotation=45, ha='right')
ax.set_ylabel("Count")
ax.set_title("Best Rubric: Constraint Composition")
ax.legend()
ax.grid(True, alpha=0.3, axis='y')
plt.tight_layout()
plt.savefig(save, dpi=150)
plt.close()
print(f"Saved {save}")
def plot_factor_ic_distribution(results, save="/app/v2_ic_distribution.png"):
"""Plot IC distribution of all evaluated factors."""
hist = results.get("search_history", [])
all_ics = []
for h in hist:
# We don't have per-factor data in history; get from top_factor approximation
pass
# Use the fact that mean_ic gives us approximate distribution
ics = [h.get("mean_ic", 0) for h in hist if h.get("mean_ic", 0) > 0]
if len(ics) < 2:
return
fig, ax = plt.subplots(figsize=(10, 5))
ax.hist(ics, bins=20, color='steelblue', edgecolor='black', alpha=0.7)
ax.axvline(np.mean(ics), color='red', linestyle='--', linewidth=2, label=f'Mean={np.mean(ics):.4f}')
ax.set_xlabel("Mean IC of Rubric-Generated Factors")
ax.set_ylabel("Frequency")
ax.set_title("Distribution of Factor Quality Across Rubric Evaluations")
ax.legend()
ax.grid(True, alpha=0.3)
plt.tight_layout()
plt.savefig(save, dpi=150)
plt.close()
print(f"Saved {save}")
def create_report(results, save="/app/v2_report.txt"):
lines = []
lines.append("=" * 70)
lines.append("AFRES v2 – Rubric-Discovery Report")
lines.append("=" * 70)
lines.append("")
rubric = results.get("best_rubric", {})
items = rubric.get("items", [])
lines.append(f"BEST DISCOVERED RUBRIC ({len(items)} item(s))")
lines.append("-" * 50)
for item in items:
lines.append(f" [{item['id']}] {item['description']}")
lines.append(f" features: {item.get('feature_focus', [])}")
lines.append(f" operators: {item.get('preferred_ops', [])}")
lines.append(f" horizon: {item.get('time_horizon_hint', '?')} "
f"complexity: {item.get('complexity_hint', '?')}")
lines.append("")
lines.append("-" * 50)
lines.append(f"Best fitness (mean IC): {results.get('best_fitness', 0):.4f}")
lines.append(f"Total factors generated: {results.get('total_factors_generated', 0)}")
lines.append(f"Valid factors: {results.get('valid_factors', 0)}")
lines.append(f"Mean IC (all valid): {results.get('mean_ic', 0):.4f}")
lines.append(f"Mean MAE (all valid): {results.get('mean_mae', 0):.4f}")
lines.append(f"Mean Sharpe (all valid): {results.get('mean_sharpe', 0):.2f}")
lines.append("")
top = results.get("top_factor", {})
lines.append("TOP FACTOR")
lines.append("-" * 50)
lines.append(f" ID: {top.get('id', '?')}")
lines.append(f" Expression: {top.get('expression', '?')}")
lines.append(f" IC: {top.get('ic', 0):+.4f}")
lines.append(f" MAE: {top.get('mae', 0):.4f}")
lines.append(f" R²: {top.get('r2', 0):.4f}")
lines.append(f" Sharpe: {top.get('sharpe', 0):.2f}")
lines.append(f" Returns: {top.get('returns', 0):.2f}")
lines.append("")
lines.append("=" * 70)
report = "\n".join(lines)
with open(save, "w") as f:
f.write(report)
print(f"Saved {save}")
return report
def main():
results = load()
plot_search_history(results)
plot_rubric_composition(results)
plot_factor_ic_distribution(results)
report = create_report(results)
print("\n" + report)
if __name__ == "__main__":
main()
|