src/make_gauge_figures.py

from pathlib import Path
import json
import numpy as np
import matplotlib
matplotlib.use('Agg')
import matplotlib.pyplot as plt
from gauge_contracts import uniform_reserve_yield

ROOT=Path(__file__).resolve().parents[1]
DATA=ROOT/'results/gauge';FIG=ROOT/'figures'
plt.rcParams.update({'font.size':9,'axes.spines.top':False,
                     'axes.spines.right':False,'savefig.dpi':200})
phase=json.loads((DATA/'phase.json').read_text())
palette=['#17647f','#3c8d99','#768c45','#b07e3c','#9c4f6e']
fig,ax=plt.subplots(1,2,figsize=(10,3.65),layout='constrained')
for color,N in zip(palette,[4,16,64,256,1024]):
    cs=np.linspace(.4,.65,301)
    probability=[uniform_reserve_yield(N,.65,1.35,c,.04) for c in cs]
    ax[0].plot(cs,probability,color=color,label=f'{N} modules')
    rows=[r for r in phase['rows'] if r['modules']==N and r['capacity']<=.65]
    ax[0].scatter([r['capacity'] for r in rows],
                  [r['passes']/r['trials'] for r in rows],s=10,color=color)
ax[0].axvline(phase['critical_capacity'],color='#263542',ls='--',lw=1)
ax[0].set(xlabel='Reachable additive reserve per module',
          ylabel='Probability of G2 feasibility',ylim=(-.025,1.04),
          title='Exact finite-size reserve transition')
ax[0].legend(frameon=False,fontsize=7,loc='upper left')
for color,c in zip(palette,[.55,.58,.59,.60]):
    Ns=np.arange(2,1025,2)
    ys=np.maximum([uniform_reserve_yield(int(N),.65,1.35,c,.04) for N in Ns],1e-40)
    ax[1].semilogy(Ns,ys,color=color,label=f'Reserve = {c:.2f}')
ax[1].set(xlabel='Number of modules',ylabel='Exact feasibility probability',
          ylim=(1e-32,1.5),title='Subcritical yield vanishes exponentially')
ax[1].legend(frameon=False,fontsize=8)
fig.savefig(FIG/'gauge_reserve_phase.png');plt.close(fig)

cal=json.loads((DATA/'calibration.json').read_text())
fig,ax=plt.subplots(1,2,figsize=(10,3.4),layout='constrained')
for dim,color in zip([1,2,3],palette):
    rows=[r for r in cal['scaling'] if r['dimension']==dim]
    ax[0].loglog([r['nodes'] for r in rows],[r['rho_max'] for r in rows],
                 'o-',color=color,label=f'{dim}D')
    ax[1].loglog([r['nodes'] for r in rows],
                 [r['samples_for_radius_006'] for r in rows],
                 'o-',color=color,label=f'{dim}D')
ax[0].set(xlabel='Comparison nodes',ylabel='Maximum effective resistance',
          title='Comparison topology controls noise amplification')
ax[1].set(xlabel='Comparison nodes',ylabel='Samples per edge',
          title='Radius 0.006; 128 inspections; failure budget 0.005')
for a in ax:a.legend(frameon=False);a.grid(alpha=.18)
fig.savefig(FIG/'gauge_calibration.png');plt.close(fig)

summary=json.loads((DATA/'summary.json').read_text())
methods=['projective','common_detector_gain','common_material_scale',
         'conventional_joint','fixed_representative','insufficient_reserve',
         'differential_bias','early_reference_release']
labels=['Projective repair','Shared detector gain','Common material scale',
        'Matched conventional','Fixed representative','Insufficient reserve',
        'Differential bias','Early reference release']
fig,ax=plt.subplots(1,2,figsize=(10,4.4),layout='constrained',sharey=True)
y=np.arange(len(methods))
for d,shift,color in [(2,-.18,'#17647f'),(3,.18,'#b07e3c')]:
    rows={r['method']:r for r in summary if r['dimension']==d}
    ax[0].barh(y+shift,[rows[m]['functional']/32 for m in methods],
               height=.32,color=color,label=f'{d}D')
    ax[1].barh(y+shift,[rows[m]['false_certificates']/32 for m in methods],
               height=.32,color=color)
ax[0].set(yticks=y,yticklabels=labels,xlim=(0,1.03),
          xlabel='Completed functional objects / 32',
          title='Model results with explicit negative controls')
ax[0].invert_yaxis()
ax[1].set(xlim=(0,1.03),xlabel='False local certificates / 32',
          title='Differential bias remains a hard failure')
ax[0].legend(frameon=False,loc='lower right')
fig.savefig(FIG/'gauge_results.png');plt.close(fig)
print('Created three v3 scientific figures.')