yProv4BO/plot.py at main · HPCI-Lab/yProv4BO · GitHub

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
import json
import glob
import numpy as np
import matplotlib
matplotlib.use('Agg')
import matplotlib.pyplot as plt
from matplotlib.collections import LineCollection
from pathlib import Path

FIGURES_DIR = Path("figures")

# ── Configuration ───────────────────────────────────────────────────────────
GPU_CONFIG = 'run_20260507_144824'
STOP_ITERS = [15, 38, 100]
COLORS_STAGES = ['#1db992', '#f3a31d', '#ef565a']

# ── Data Loading & Recalculation ────────────────────────────────────────────
records = []
for f in glob.glob(f'deepcam_results/{GPU_CONFIG}/detailed_logs/*.json'):
    if '._' not in f:
        with open(f) as fh:
            j = json.load(fh)
            if j["objectives"]["energy_kWh"] < 100:
                records.append(j)

records = sorted(records, key=lambda x: (x['iteration'], x['candidate']))

raw_ious = []
raw_energies = []
cum_energies = []
best_ious = []
iterations = []
current_cum_en = 0
current_best_iou = -np.inf

# Process all records for scatter and step-plot logic[cite: 3]
records = records[:102]
for r in records:
    # Wh conversion as seen in the screenshot
    energy_wh = r['objectives']['energy_kWh'] * 1000
    current_cum_en += energy_wh
    current_best_iou = max(current_best_iou, r['objectives']['iou_validation'])

    r['cum_en_recalc'] = current_cum_en
    r['best_iou_recalc'] = current_best_iou

    raw_ious.append(r['objectives']['iou_validation'])
    raw_energies.append(current_cum_en)
    cum_energies.append(current_cum_en)
    best_ious.append(current_best_iou)
    iterations.append(r['iteration'])

total_energy = cum_energies[-1]

# ── Gradient Step Plot Logic ────────────────────────────────────────────────
x_steps, y_steps, iter_steps = [], [], []
for i in range(len(cum_energies) - 1):
    # Horizontal/Vertical segments for step effect[cite: 3]
    x_steps.extend([cum_energies[i], cum_energies[i+1], cum_energies[i+1]])
    y_steps.extend([best_ious[i], best_ious[i], best_ious[i+1]])
    iter_steps.extend([iterations[i], iterations[i], iterations[i+1]])

points = np.array([x_steps, y_steps]).T.reshape(-1, 1, 2)
segments = np.concatenate([points[:-1], points[1:]], axis=1)

# ── Plotting ────────────────────────────────────────────────────────────────
fig, ax = plt.subplots(figsize=(8, 9), facecolor='white')
cmap = plt.get_cmap('viridis')
norm = plt.Normalize(0, 100)

# 1. Background Scatter: All individual results
sc = ax.scatter(raw_energies, raw_ious, c=COLORS_STAGES[-1],
                alpha=0.2, s=25, zorder=1, edgecolors='none')

# 2. Gradient Step Line[cite: 3]
lc = LineCollection(segments, cmap=cmap, norm=norm, linewidth=3, zorder=3)
lc.set_array(np.array(iter_steps))
ax.add_collection(lc)

# Colorbar for BO Iterations[cite: 1, 3]
cbar = fig.colorbar(lc, ax=ax, fraction=0.046, pad=0.04)
cbar.set_label('BO Iteration', rotation=90, labelpad=15, fontsize=12)

# 3. Milestones, Vertical Lines, and Text Labels[cite: 1]
for i, stop_it in enumerate(STOP_ITERS):
    target_recs = [r for r in records if r['iteration'] == stop_it]
    if not target_recs: continue

    m = target_recs[-1]
    en_val, iou_val = m['cum_en_recalc'], m['best_iou_recalc']
    pct = (en_val / total_energy) * 100

    # Vertical dash lines[cite: 1]
    ax.axvline(x=en_val, color=COLORS_STAGES[i], linestyle='--', alpha=0.5, zorder=2)

    # Large marker dots[cite: 1]
    label = f"S{i+1}: stop iter {stop_it} - {en_val:.0f} Wh ({pct:.0f}%)"
    ax.plot(en_val, iou_val, 'o', color=COLORS_STAGES[i], markersize=12,
            label=label, zorder=5, markeredgecolor='white')

    # Floating label for IoU value
    ax.text(en_val, iou_val + 0.01, f'{iou_val:.3f}', color=COLORS_STAGES[i],
            fontweight='bold', ha='center', fontsize=10, zorder=6)

# Aesthetics & Labels
ax.spines['top'].set_visible(False)
ax.spines['right'].set_visible(False)
ax.spines['left'].set_linewidth(2)
ax.spines['bottom'].set_linewidth(2)

plt.title('Evolution of the Pareto Frontier (Convergence Analysis)',
          fontsize=14, fontweight='bold', pad=20)
plt.xlabel('Cumulative Campaign Energy (Wh)', fontsize=12)
plt.ylabel('IoU (Validation)', fontsize=12)
plt.xlim(0, total_energy * 1.05)
plt.ylim(min(raw_ious) - 0.02, max(best_ious) + 0.05)

# Legend placement below the plot[cite: 1]
plt.legend(loc='upper center', bbox_to_anchor=(0.5, -0.12), frameon=False, fontsize=10)

plt.tight_layout()
plt.savefig(FIGURES_DIR / 'pareto_gradient_scatter_final.png', dpi=300, bbox_inches='tight')