data: add Z=128 pipeline results and comparison plots
Run FER validation at Z=128 with normalized min-sum (alpha=0.875). Best alpha found via sweep: 0.875 (threshold 2.90 photons/slot). Z=128 matrix achieves girth=8 vs girth=6 at Z=32. Add Z=128 vs Z=32 FER comparison plot. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
This commit is contained in:
318
model/plot_de_results.py
Normal file
318
model/plot_de_results.py
Normal file
@@ -0,0 +1,318 @@
|
||||
#!/usr/bin/env python3
|
||||
"""Generate presentation plots from density evolution results."""
|
||||
|
||||
import json
|
||||
import numpy as np
|
||||
import matplotlib
|
||||
matplotlib.use('Agg')
|
||||
import matplotlib.pyplot as plt
|
||||
from pathlib import Path
|
||||
|
||||
RESULTS_PATH = Path(__file__).parent.parent / 'data' / 'de_results.json'
|
||||
RESULTS_Z128_PATH = Path(__file__).parent.parent / 'data' / 'de_results_z128.json'
|
||||
OUT_DIR = Path(__file__).parent.parent / 'data' / 'plots'
|
||||
|
||||
# Presentation-friendly style
|
||||
plt.rcParams.update({
|
||||
'font.size': 13,
|
||||
'axes.titlesize': 15,
|
||||
'axes.labelsize': 14,
|
||||
'legend.fontsize': 11,
|
||||
'xtick.labelsize': 12,
|
||||
'ytick.labelsize': 12,
|
||||
'figure.dpi': 150,
|
||||
'savefig.bbox': 'tight',
|
||||
'savefig.pad_inches': 0.15,
|
||||
})
|
||||
|
||||
COLORS = {
|
||||
'original': '#d62728', # red
|
||||
'peg_ring': '#1f77b4', # blue
|
||||
'optimized': '#2ca02c', # green
|
||||
}
|
||||
|
||||
LABELS = {
|
||||
'original': 'Original staircase [7,2,2,2,2,2,2,1]',
|
||||
'peg_ring': 'PEG ring [7,3,3,3,2,2,2,2]',
|
||||
'optimized': 'DE-optimized [7,4,4,4,4,3,3,3]',
|
||||
}
|
||||
|
||||
MARKERS = {
|
||||
'original': 's',
|
||||
'peg_ring': '^',
|
||||
'optimized': 'o',
|
||||
}
|
||||
|
||||
|
||||
def load_results():
|
||||
with open(RESULTS_PATH) as f:
|
||||
return json.load(f)
|
||||
|
||||
|
||||
def plot_fer_comparison(data):
|
||||
"""FER vs lambda_s for all three matrices."""
|
||||
fig, ax = plt.subplots(figsize=(8, 5.5))
|
||||
|
||||
fer_data = data['fer_comparison']
|
||||
lam_s = fer_data['lam_s_points']
|
||||
|
||||
for key in ['original', 'peg_ring', 'optimized']:
|
||||
fer_vals = [fer_data[key][str(l)]['fer'] for l in lam_s]
|
||||
# Replace 0 with small value for log scale
|
||||
fer_plot = [max(f, 2e-3) for f in fer_vals]
|
||||
ax.semilogy(lam_s, fer_plot,
|
||||
color=COLORS[key], marker=MARKERS[key], markersize=8,
|
||||
linewidth=2, label=LABELS[key],
|
||||
markeredgecolor='white', markeredgewidth=0.8)
|
||||
|
||||
ax.set_xlabel(r'Signal photons/slot ($\lambda_s$)')
|
||||
ax.set_ylabel('Frame Error Rate (FER)')
|
||||
ax.set_title('FER Comparison: Rate-1/8 QC-LDPC (n=256, k=32)')
|
||||
ax.legend(loc='upper right', framealpha=0.9)
|
||||
ax.set_xlim(1.5, 10.5)
|
||||
ax.set_ylim(1e-3, 1.0)
|
||||
ax.grid(True, alpha=0.3, which='both')
|
||||
ax.set_xticks([2, 3, 4, 5, 7, 10])
|
||||
ax.set_xticklabels(['2', '3', '4', '5', '7', '10'])
|
||||
|
||||
# Annotate the improvement
|
||||
ax.annotate('13x fewer\nframe errors',
|
||||
xy=(5, 0.01), xytext=(6.5, 0.06),
|
||||
fontsize=11, color=COLORS['optimized'],
|
||||
arrowprops=dict(arrowstyle='->', color=COLORS['optimized'], lw=1.5),
|
||||
ha='center')
|
||||
|
||||
fig.savefig(OUT_DIR / 'fer_comparison.png')
|
||||
plt.close(fig)
|
||||
print(f' Saved fer_comparison.png')
|
||||
|
||||
|
||||
def plot_threshold_bars(data):
|
||||
"""Bar chart of DE thresholds."""
|
||||
fig, ax = plt.subplots(figsize=(7, 5))
|
||||
|
||||
thresholds = {
|
||||
'Original\nstaircase': data['reference_thresholds']['Original staircase [7,2,2,2,2,2,2,1]'],
|
||||
'PEG\nring': data['reference_thresholds']['PEG ring [7,3,3,3,2,2,2,2]'],
|
||||
'DE-\noptimized': data['best_threshold'],
|
||||
}
|
||||
# Shannon limit for rate 1/8 Poisson channel ~ 0.47 photons/slot
|
||||
shannon_limit = 0.47
|
||||
|
||||
names = list(thresholds.keys())
|
||||
vals = list(thresholds.values())
|
||||
colors = [COLORS['original'], COLORS['peg_ring'], COLORS['optimized']]
|
||||
|
||||
bars = ax.bar(names, vals, color=colors, width=0.55, edgecolor='white', linewidth=1.5)
|
||||
|
||||
# Value labels on bars
|
||||
for bar, val in zip(bars, vals):
|
||||
ax.text(bar.get_x() + bar.get_width()/2, bar.get_height() + 0.1,
|
||||
f'{val:.1f}', ha='center', va='bottom', fontweight='bold', fontsize=14)
|
||||
|
||||
# Shannon limit line
|
||||
ax.axhline(y=shannon_limit, color='black', linestyle='--', linewidth=1.5, alpha=0.6)
|
||||
ax.text(2.42, shannon_limit + 0.12, f'Shannon limit ({shannon_limit})',
|
||||
fontsize=11, ha='right', va='bottom', style='italic', alpha=0.7)
|
||||
|
||||
ax.set_ylabel(r'DE Threshold ($\lambda_s^*$, photons/slot)')
|
||||
ax.set_title('Decoding Threshold Comparison')
|
||||
ax.set_ylim(0, 6.5)
|
||||
ax.grid(True, alpha=0.2, axis='y')
|
||||
|
||||
# Improvement annotation
|
||||
improvement_db = 10 * np.log10(thresholds['Original\nstaircase'] / thresholds['DE-\noptimized'])
|
||||
ax.annotate(f'{improvement_db:.1f} dB\nimprovement',
|
||||
xy=(2, thresholds['DE-\noptimized']),
|
||||
xytext=(2.35, 4.2),
|
||||
fontsize=12, fontweight='bold', color=COLORS['optimized'],
|
||||
arrowprops=dict(arrowstyle='->', color=COLORS['optimized'], lw=1.5),
|
||||
ha='center')
|
||||
|
||||
fig.savefig(OUT_DIR / 'threshold_comparison.png')
|
||||
plt.close(fig)
|
||||
print(f' Saved threshold_comparison.png')
|
||||
|
||||
|
||||
def plot_degree_distribution(data):
|
||||
"""VN degree distribution comparison."""
|
||||
fig, axes = plt.subplots(1, 3, figsize=(12, 4), sharey=True)
|
||||
|
||||
distributions = {
|
||||
'Original staircase': [7, 2, 2, 2, 2, 2, 2, 1],
|
||||
'PEG ring': [7, 3, 3, 3, 2, 2, 2, 2],
|
||||
'DE-optimized': data['best_degrees'],
|
||||
}
|
||||
color_keys = ['original', 'peg_ring', 'optimized']
|
||||
col_labels = [f'c{i}' for i in range(8)]
|
||||
|
||||
for ax, (name, degrees), ckey in zip(axes, distributions.items(), color_keys):
|
||||
colors_bar = [COLORS[ckey] if i > 0 else '#888888' for i in range(8)]
|
||||
ax.bar(col_labels, degrees, color=colors_bar, edgecolor='white', linewidth=1)
|
||||
ax.set_title(name, fontsize=12, fontweight='bold')
|
||||
ax.set_xlabel('Base matrix column')
|
||||
ax.set_ylim(0, 8)
|
||||
ax.grid(True, alpha=0.2, axis='y')
|
||||
|
||||
# Annotate average parity degree
|
||||
avg_parity = np.mean(degrees[1:])
|
||||
ax.text(4, 7.2, f'avg parity dv = {avg_parity:.1f}',
|
||||
ha='center', fontsize=10, style='italic')
|
||||
|
||||
axes[0].set_ylabel('Variable node degree')
|
||||
fig.suptitle('VN Degree Distributions (col 0 = info, cols 1-7 = parity)',
|
||||
fontsize=13, y=1.02)
|
||||
fig.tight_layout()
|
||||
|
||||
fig.savefig(OUT_DIR / 'degree_distributions.png')
|
||||
plt.close(fig)
|
||||
print(f' Saved degree_distributions.png')
|
||||
|
||||
|
||||
def plot_base_matrix_heatmap(data):
|
||||
"""Heatmap of the constructed base matrix."""
|
||||
fig, ax = plt.subplots(figsize=(7, 5))
|
||||
|
||||
H = np.array(data['constructed_matrix'])
|
||||
m_base, n_base = H.shape
|
||||
|
||||
# Create display matrix: shift values where connected, NaN where not
|
||||
display = np.full_like(H, dtype=float, fill_value=np.nan)
|
||||
for r in range(m_base):
|
||||
for c in range(n_base):
|
||||
if H[r, c] >= 0:
|
||||
display[r, c] = H[r, c]
|
||||
|
||||
im = ax.imshow(display, cmap='YlGnBu', aspect='auto', vmin=0, vmax=31)
|
||||
|
||||
# Add text annotations
|
||||
for r in range(m_base):
|
||||
for c in range(n_base):
|
||||
if H[r, c] >= 0:
|
||||
ax.text(c, r, str(H[r, c]), ha='center', va='center',
|
||||
fontsize=11, fontweight='bold', color='black')
|
||||
else:
|
||||
ax.text(c, r, '-', ha='center', va='center',
|
||||
fontsize=11, color='#cccccc')
|
||||
|
||||
ax.set_xticks(range(n_base))
|
||||
ax.set_xticklabels([f'c{i}' for i in range(n_base)])
|
||||
ax.set_yticks(range(m_base))
|
||||
ax.set_yticklabels([f'r{i}' for i in range(m_base)])
|
||||
ax.set_xlabel('Column (c0=info, c1-c7=parity)')
|
||||
ax.set_ylabel('Row (check node group)')
|
||||
ax.set_title(f'Optimized Base Matrix (Z=32, girth={data["matrix_checks"]["girth"]})')
|
||||
|
||||
cbar = fig.colorbar(im, ax=ax, shrink=0.8, label='Circulant shift')
|
||||
|
||||
fig.savefig(OUT_DIR / 'base_matrix_heatmap.png')
|
||||
plt.close(fig)
|
||||
print(f' Saved base_matrix_heatmap.png')
|
||||
|
||||
|
||||
def load_z128_results():
|
||||
"""Load Z=128 results if available."""
|
||||
if RESULTS_Z128_PATH.exists():
|
||||
with open(RESULTS_Z128_PATH) as f:
|
||||
return json.load(f)
|
||||
return None
|
||||
|
||||
|
||||
def plot_z128_comparison(data_z32, data_z128):
|
||||
"""
|
||||
FER vs lambda_s comparison between Z=32 and Z=128 for the
|
||||
DE-optimized degree distribution [7,4,4,4,4,3,3,3].
|
||||
|
||||
Shows both lifting factors on the same axes, with the optimized
|
||||
matrix in each case plus the PEG ring reference.
|
||||
"""
|
||||
fig, ax = plt.subplots(figsize=(9, 6))
|
||||
|
||||
# Colors and styles for Z=32 vs Z=128
|
||||
z32_color = '#2ca02c' # green (matches optimized)
|
||||
z128_color = '#9467bd' # purple
|
||||
z32_peg_color = '#1f77b4' # blue (matches peg_ring)
|
||||
z128_peg_color = '#ff7f0e' # orange
|
||||
|
||||
# Z=32 optimized data (from de_results.json)
|
||||
fer_z32 = data_z32['fer_comparison']
|
||||
lam_s_z32 = fer_z32['lam_s_points']
|
||||
fer_z32_opt = [fer_z32['optimized'][str(l)]['fer'] for l in lam_s_z32]
|
||||
fer_z32_peg = [fer_z32['peg_ring'][str(l)]['fer'] for l in lam_s_z32]
|
||||
|
||||
# Z=128 data
|
||||
fer_z128 = data_z128['fer_results']
|
||||
lam_s_z128 = fer_z128['lam_s_points']
|
||||
fer_z128_opt = [fer_z128['optimized'][str(l)]['fer'] for l in lam_s_z128]
|
||||
fer_z128_peg = [fer_z128['peg_ring'][str(l)]['fer'] for l in lam_s_z128]
|
||||
|
||||
# Replace 0 with small value for log scale
|
||||
floor = 5e-3
|
||||
|
||||
# Plot Z=32 curves (solid lines)
|
||||
ax.semilogy(lam_s_z32, [max(f, floor) for f in fer_z32_opt],
|
||||
color=z32_color, marker='o', markersize=8, linewidth=2,
|
||||
label='Optimized Z=32 (n=256, offset MS)',
|
||||
markeredgecolor='white', markeredgewidth=0.8)
|
||||
ax.semilogy(lam_s_z32, [max(f, floor) for f in fer_z32_peg],
|
||||
color=z32_peg_color, marker='^', markersize=8, linewidth=2,
|
||||
label='PEG ring Z=32 (n=256, offset MS)',
|
||||
markeredgecolor='white', markeredgewidth=0.8,
|
||||
linestyle='--', alpha=0.7)
|
||||
|
||||
# Plot Z=128 curves (dashed lines, larger markers)
|
||||
ax.semilogy(lam_s_z128, [max(f, floor) for f in fer_z128_opt],
|
||||
color=z128_color, marker='D', markersize=9, linewidth=2.5,
|
||||
label=f'Optimized Z=128 (n=1024, norm MS, a={data_z128["alpha"]})',
|
||||
markeredgecolor='white', markeredgewidth=0.8)
|
||||
ax.semilogy(lam_s_z128, [max(f, floor) for f in fer_z128_peg],
|
||||
color=z128_peg_color, marker='v', markersize=9, linewidth=2.5,
|
||||
label=f'PEG ring Z=128 (n=1024, norm MS, a={data_z128["alpha"]})',
|
||||
markeredgecolor='white', markeredgewidth=0.8,
|
||||
linestyle='--', alpha=0.7)
|
||||
|
||||
ax.set_xlabel(r'Signal photons/slot ($\lambda_s$)')
|
||||
ax.set_ylabel('Frame Error Rate (FER)')
|
||||
ax.set_title('FER Comparison: Z=32 vs Z=128\nDE-optimized [7,4,4,4,4,3,3,3]')
|
||||
ax.legend(loc='upper right', framealpha=0.9, fontsize=10)
|
||||
ax.set_xlim(1.0, 11.0)
|
||||
ax.set_ylim(floor / 2, 1.1)
|
||||
ax.grid(True, alpha=0.3, which='both')
|
||||
ax.set_xticks([1.5, 2, 3, 4, 5, 7, 10])
|
||||
ax.set_xticklabels(['1.5', '2', '3', '4', '5', '7', '10'])
|
||||
|
||||
# Add Z=128 matrix info annotation
|
||||
girth_128 = data_z128.get('girth', '?')
|
||||
ax.text(0.02, 0.02,
|
||||
f'Z=128: girth={girth_128}, n=1024, k=128\n'
|
||||
f'Z=32: girth={data_z32["matrix_checks"]["girth"]}, n=256, k=32',
|
||||
transform=ax.transAxes, fontsize=9, verticalalignment='bottom',
|
||||
bbox=dict(boxstyle='round,pad=0.3', facecolor='wheat', alpha=0.5))
|
||||
|
||||
fig.savefig(OUT_DIR / 'z128_fer_comparison.png')
|
||||
plt.close(fig)
|
||||
print(f' Saved z128_fer_comparison.png')
|
||||
|
||||
|
||||
def main():
|
||||
OUT_DIR.mkdir(parents=True, exist_ok=True)
|
||||
data = load_results()
|
||||
|
||||
print('Generating plots...')
|
||||
plot_fer_comparison(data)
|
||||
plot_threshold_bars(data)
|
||||
plot_degree_distribution(data)
|
||||
plot_base_matrix_heatmap(data)
|
||||
|
||||
# Z=128 comparison plot
|
||||
data_z128 = load_z128_results()
|
||||
if data_z128 is not None:
|
||||
plot_z128_comparison(data, data_z128)
|
||||
else:
|
||||
print(' Skipping Z=128 comparison (data/de_results_z128.json not found)')
|
||||
|
||||
print(f'\nAll plots saved to {OUT_DIR}/')
|
||||
|
||||
|
||||
if __name__ == '__main__':
|
||||
main()
|
||||
Reference in New Issue
Block a user