File size: 9,554 Bytes
a1190da
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
#!/usr/bin/env python3
"""
Create visualizations for Model Scaling Study.
Generates publication-ready charts and tables.
"""

import json
import numpy as np
import pandas as pd
import matplotlib.pyplot as plt
import seaborn as sns
from pathlib import Path

# Set style for publication-quality figures
sns.set_style("whitegrid")
plt.rcParams['figure.figsize'] = (12, 8)
plt.rcParams['font.size'] = 12
plt.rcParams['axes.labelsize'] = 14
plt.rcParams['axes.titlesize'] = 16
plt.rcParams['xtick.labelsize'] = 12
plt.rcParams['ytick.labelsize'] = 12
plt.rcParams['legend.fontsize'] = 12

# Create output directory
output_dir = Path('visualizations')
output_dir.mkdir(exist_ok=True)

print("="*80)
print("CREATING VISUALIZATIONS FOR MODEL SCALING STUDY")
print("="*80)
print()

# Load data
print("Loading data...")

# Quality results
quality_data = {
    'Base': {'valid_rate': 0.994, 'diversity': 0.978, 'unique': 489, 'samples': 500},
    'Medium': {'valid_rate': 0.992, 'diversity': 0.988, 'unique': 494, 'samples': 500},
    'Large': {'valid_rate': 1.000, 'diversity': 0.986, 'unique': 493, 'samples': 500}
}

# Nguyen benchmark summary
with open('results_nguyen_benchmarks/summary.json') as f:
    nguyen_data = json.load(f)

# Extract Nguyen stats by model
nguyen_stats = {}
for model in ['base', 'medium', 'large']:
    model_results = [r for r in nguyen_data['results'] if r['model'] == model]
    nguyen_stats[model.capitalize()] = {
        'avg_valid_rate': np.mean([r['valid_rate'] for r in model_results]),
        'avg_best_r2': np.mean([r['best_r2'] for r in model_results]),
        'max_r2': max([r['best_r2'] for r in model_results]),
        'benchmarks_gt_099': sum([1 for r in model_results if r['best_r2'] > 0.99])
    }

print("Data loaded successfully!")
print()

# ============================================================================
# Figure 1: Valid Rate Comparison (Quality + Benchmarks)
# ============================================================================
print("Creating Figure 1: Valid Rate Comparison...")

fig, (ax1, ax2) = plt.subplots(1, 2, figsize=(14, 6))

models = ['Base', 'Medium', 'Large']
colors = ['#3498db', '#e74c3c', '#2ecc71']

# Quality valid rates
quality_valid = [quality_data[m]['valid_rate'] * 100 for m in models]
bars1 = ax1.bar(models, quality_valid, color=colors, alpha=0.8, edgecolor='black', linewidth=1.5)
ax1.set_ylabel('Valid Expression Rate (%)', fontsize=14, fontweight='bold')
ax1.set_title('Quality Evaluation\n(500 samples per model)', fontsize=16, fontweight='bold')
ax1.set_ylim([95, 101])
ax1.axhline(y=100, color='green', linestyle='--', linewidth=2, label='Perfect (100%)')
ax1.legend()

# Add value labels on bars
for bar, val in zip(bars1, quality_valid):
    height = bar.get_height()
    ax1.text(bar.get_x() + bar.get_width()/2., height + 0.3,
             f'{val:.1f}%', ha='center', va='bottom', fontsize=12, fontweight='bold')

# Benchmark valid rates
benchmark_valid = [nguyen_stats[m]['avg_valid_rate'] * 100 for m in models]
bars2 = ax2.bar(models, benchmark_valid, color=colors, alpha=0.8, edgecolor='black', linewidth=1.5)
ax2.set_ylabel('Valid Expression Rate (%)', fontsize=14, fontweight='bold')
ax2.set_title('Nguyen Benchmarks\n(36 experiments, 3,600 expressions)', fontsize=16, fontweight='bold')
ax2.set_ylim([0, 100])

# Add value labels on bars
for bar, val in zip(bars2, benchmark_valid):
    height = bar.get_height()
    ax2.text(bar.get_x() + bar.get_width()/2., height + 2,
             f'{val:.1f}%', ha='center', va='bottom', fontsize=12, fontweight='bold')

plt.tight_layout()
plt.savefig(output_dir / 'fig1_valid_rate_comparison.png', dpi=300, bbox_inches='tight')
print(f"  Saved: {output_dir / 'fig1_valid_rate_comparison.png'}")
plt.close()

# ============================================================================
# Figure 2: R² Performance on Nguyen Benchmarks
# ============================================================================
print("Creating Figure 2: R² Performance...")

fig, ax = plt.subplots(figsize=(12, 6))

x = np.arange(len(models))
width = 0.25

avg_r2 = [nguyen_stats[m]['avg_best_r2'] for m in models]
max_r2 = [nguyen_stats[m]['max_r2'] for m in models]

bars1 = ax.bar(x - width/2, avg_r2, width, label='Average Best R²',
               color='#3498db', alpha=0.8, edgecolor='black', linewidth=1.5)
bars2 = ax.bar(x + width/2, max_r2, width, label='Maximum R²',
               color='#e74c3c', alpha=0.8, edgecolor='black', linewidth=1.5)

ax.set_ylabel('R² Score', fontsize=14, fontweight='bold')
ax.set_title('Symbolic Regression Performance (Nguyen Benchmarks)', fontsize=16, fontweight='bold')
ax.set_xticks(x)
ax.set_xticklabels(models)
ax.legend(fontsize=12)
ax.set_ylim([0.85, 1.05])
ax.axhline(y=1.0, color='green', linestyle='--', linewidth=2, alpha=0.5, label='Perfect Fit')
ax.grid(axis='y', alpha=0.3)

# Add value labels
for bar in bars1:
    height = bar.get_height()
    ax.text(bar.get_x() + bar.get_width()/2., height + 0.01,
            f'{height:.4f}', ha='center', va='bottom', fontsize=11, fontweight='bold')

for bar in bars2:
    height = bar.get_height()
    ax.text(bar.get_x() + bar.get_width()/2., height + 0.01,
            f'{height:.4f}', ha='center', va='bottom', fontsize=11, fontweight='bold')

plt.tight_layout()
plt.savefig(output_dir / 'fig2_r2_performance.png', dpi=300, bbox_inches='tight')
print(f"  Saved: {output_dir / 'fig2_r2_performance.png'}")
plt.close()

# ============================================================================
# Figure 3: Per-Benchmark Heatmap
# ============================================================================
print("Creating Figure 3: Per-Benchmark Heatmap...")

# Extract per-benchmark R² scores
benchmark_matrix = []
for bench in range(1, 13):
    row = []
    for model in ['base', 'medium', 'large']:
        result = [r for r in nguyen_data['results']
                 if r['model'] == model and r['benchmark'] == f'nguyen_{bench}']
        if result:
            row.append(result[0]['best_r2'])
        else:
            row.append(0)
    benchmark_matrix.append(row)

benchmark_matrix = np.array(benchmark_matrix)

fig, ax = plt.subplots(figsize=(10, 10))
im = ax.imshow(benchmark_matrix, cmap='RdYlGn', aspect='auto', vmin=0.5, vmax=1.0)

# Set ticks
ax.set_xticks(np.arange(3))
ax.set_yticks(np.arange(12))
ax.set_xticklabels(['Base\n(124M)', 'Medium\n(355M)', 'Large\n(774M)'], fontsize=12)
ax.set_yticklabels([f'Nguyen-{i+1}' for i in range(12)], fontsize=11)

# Add colorbar
cbar = plt.colorbar(im, ax=ax)
cbar.set_label('R² Score', rotation=270, labelpad=20, fontsize=14, fontweight='bold')

# Add text annotations
for i in range(12):
    for j in range(3):
        text = ax.text(j, i, f'{benchmark_matrix[i, j]:.3f}',
                      ha="center", va="center", color="black", fontsize=10, fontweight='bold')

ax.set_title('R² Scores by Model and Benchmark', fontsize=16, fontweight='bold', pad=20)
plt.tight_layout()
plt.savefig(output_dir / 'fig3_benchmark_heatmap.png', dpi=300, bbox_inches='tight')
print(f"  Saved: {output_dir / 'fig3_benchmark_heatmap.png'}")
plt.close()

# ============================================================================
# Figure 4: Scaling Progression (Valid Rate + R²)
# ============================================================================
print("Creating Figure 4: Scaling Progression...")

fig, (ax1, ax2) = plt.subplots(1, 2, figsize=(14, 6))

params = [124, 355, 774]  # Millions

# Valid rate progression (benchmarks)
ax1.plot(params, benchmark_valid, 'o-', color='#3498db', linewidth=3,
         markersize=12, label='Nguyen Valid Rate', markeredgecolor='black', markeredgewidth=2)
ax1.set_xlabel('Model Size (Million Parameters)', fontsize=14, fontweight='bold')
ax1.set_ylabel('Valid Expression Rate (%)', fontsize=14, fontweight='bold')
ax1.set_title('Valid Rate vs Model Size', fontsize=16, fontweight='bold')
ax1.grid(True, alpha=0.3)
ax1.legend(fontsize=12)

# Add value labels
for x, y in zip(params, benchmark_valid):
    ax1.text(x, y + 2, f'{y:.1f}%', ha='center', fontsize=11, fontweight='bold')

# R² progression
ax2.plot(params, avg_r2, 'o-', color='#e74c3c', linewidth=3,
         markersize=12, label='Average Best R²', markeredgecolor='black', markeredgewidth=2)
ax2.axhline(y=1.0, color='green', linestyle='--', linewidth=2, alpha=0.5, label='Perfect Fit')
ax2.set_xlabel('Model Size (Million Parameters)', fontsize=14, fontweight='bold')
ax2.set_ylabel('R² Score', fontsize=14, fontweight='bold')
ax2.set_title('R² vs Model Size', fontsize=16, fontweight='bold')
ax2.set_ylim([0.9, 1.02])
ax2.grid(True, alpha=0.3)
ax2.legend(fontsize=12)

# Add value labels
for x, y in zip(params, avg_r2):
    ax2.text(x, y + 0.005, f'{y:.4f}', ha='center', fontsize=11, fontweight='bold')

plt.tight_layout()
plt.savefig(output_dir / 'fig4_scaling_progression.png', dpi=300, bbox_inches='tight')
print(f"  Saved: {output_dir / 'fig4_scaling_progression.png'}")
plt.close()

print()
print("="*80)
print("ALL VISUALIZATIONS CREATED SUCCESSFULLY!")
print("="*80)
print()
print(f"Output directory: {output_dir.absolute()}")
print()
print("Generated files:")
print("  1. fig1_valid_rate_comparison.png - Quality vs Benchmark valid rates")
print("  2. fig2_r2_performance.png - R² scores comparison")
print("  3. fig3_benchmark_heatmap.png - Per-benchmark R² heatmap")
print("  4. fig4_scaling_progression.png - Scaling laws visualization")
print()
print("These figures are publication-ready (300 DPI, high resolution)")
print()