File size: 8,796 Bytes
19591f3
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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
#!/usr/bin/env python3
"""Rebuild the manuscript's static scientific figures.

Requires matplotlib (optional; the selection and certificate code uses only the
Python standard library). All rational sample values are computed with Fraction.
figure_data.json stores rational values as strings and irrational lower bounds
as exact quadratic-surd records. Floating conversion is used only for plotting.
Curves join the displayed sample points; they do not assert interpolation laws.

Run: python make_figures.py
"""
from fractions import Fraction as F
from pathlib import Path
import json
import math

import matplotlib
matplotlib.use('Agg')
import matplotlib.pyplot as plt
from matplotlib.ticker import MultipleLocator

from selection import coefficient
from profile_minimax import solve_profile

HERE=Path(__file__).resolve().parent
FIGURES=HERE/'figures'
NAVY='#173654'
TEAL='#087E8B'
AMBER='#C88026'
SLATE='#647382'
GRID='#E4EAF0'


def surd(rational_part,radical_coefficient,radicand):
    return {'rational_part':str(F(rational_part)),
            'radical_coefficient':str(F(radical_coefficient)),
            'radicand':str(F(radicand)),
            'interpretation':'rational_part + radical_coefficient * sqrt(radicand)'}


def surd_float(record):
    return float(F(record['rational_part']))+float(F(record['radical_coefficient']))*math.sqrt(float(F(record['radicand'])))


def style_axes(ax):
    ax.set_facecolor('white')
    ax.grid(axis='y',color=GRID,linewidth=.65,zorder=0)
    ax.set_axisbelow(True)
    for spine in ('top','right'):
        ax.spines[spine].set_visible(False)
    for spine in ('left','bottom'):
        ax.spines[spine].set_color('#ABB7C3')
        ax.spines[spine].set_linewidth(.7)
    ax.tick_params(axis='both',colors=NAVY,length=3,width=.6,pad=3)


def save_figure(fig,stem):
    fig.savefig(FIGURES/(stem+'.pdf'),facecolor='white',bbox_inches='tight',pad_inches=.07)
    fig.savefig(FIGURES/(stem+'.png'),dpi=250,facecolor='white',bbox_inches='tight',pad_inches=.07)
    plt.close(fig)


def main():
    FIGURES.mkdir(exist_ok=True)
    plt.rcParams.update({
        'font.family':'DejaVu Sans','font.size':9,
        'axes.labelsize':9,'axes.titlesize':10,
        'xtick.labelsize':8,'ytick.labelsize':8,
        'text.color':NAVY,'axes.labelcolor':NAVY,
        'pdf.fonttype':42,'ps.fonttype':42,
        'mathtext.fontset':'dejavusans',
    })
    budget_data=[]
    for n in range(2,41):
        # n*B_(4,n) = (4n-1 - sqrt(4n^2-8n+1))/(4n).
        lower=surd(F(4*n-1,4*n),F(-1,4*n),4*n*n-8*n+1)
        upper=F(n)*coefficient(4,n)
        budget_data.append({'n':n,'scaled_lower':lower,
                            'scaled_upper':str(upper),'scaled_iid':str(F(1))})
    dimension_data=[]
    for k in range(4,33):
        gamma=surd(F(2*(k-2),k),F(-1,k),2*(k-2)*(k-3))
        dimension_data.append({'k':k,'gamma_lower':gamma,'planar_reference':str(F(1,2))})

    fig,axes=plt.subplots(1,2,figsize=(7,3.18),gridspec_kw={'wspace':.32})
    left,right=axes
    ns=[row['n'] for row in budget_data]
    lowers=[surd_float(row['scaled_lower']) for row in budget_data]
    uppers=[float(F(row['scaled_upper'])) for row in budget_data]
    style_axes(left)
    left.fill_between(ns,lowers,uppers,color=TEAL,alpha=.10,linewidth=0,zorder=1)
    left.axhline(1,color=SLATE,linewidth=1.1,linestyle=(0,(4,3)),zorder=2)
    left.plot(ns,uppers,color=NAVY,linewidth=1.8,zorder=3)
    left.plot(ns,lowers,color=TEAL,linewidth=1.9,zorder=4)
    left.scatter([2],[lowers[0]],color=TEAL,s=18,zorder=5)
    left.set_xlim(1.3,40.7)
    left.set_ylim(.475,1.065)
    left.set_xticks([2,10,20,30,40])
    left.set_yticks([.5,.6,.7,.8,.9,1.0])
    left.set_xlabel('Selection budget n',labelpad=6)
    left.set_ylabel('Scaled excess-loss coefficient',labelpad=5)
    left.set_title('(a) Dimension four: certified bounds',loc='left',pad=11,fontweight='bold')
    left.text(39.8,1.012,'Independent sampling: 1',ha='right',va='bottom',color=SLATE,fontsize=7.5)
    left.text(39.8,.780,r'Upper bound $nU_4(n)$',ha='right',va='bottom',color=NAVY,fontsize=8)
    left.text(39.8,.535,r'Lower bound $nB_{4,n}$',ha='right',va='bottom',color=TEAL,fontsize=8)
    left.text(22,.640,'Unresolved interval',ha='center',va='center',color=SLATE,fontsize=7.7)

    ks=[row['k'] for row in dimension_data]
    gammas=[surd_float(row['gamma_lower']) for row in dimension_data]
    style_axes(right)
    right.axhline(.5,color=AMBER,linewidth=1.25,linestyle=(0,(4,3)),zorder=2)
    right.plot(ks,gammas,color=TEAL,linewidth=1.9,zorder=3)
    right.scatter([4,8,16,32],[gammas[k-4] for k in [4,8,16,32]],color=TEAL,s=15,zorder=4)
    right.set_xlim(3.3,32.7)
    right.set_ylim(.491,.59)
    right.set_xticks([4,8,16,24,32])
    right.set_yticks([.50,.52,.54,.56,.58])
    right.set_xlabel('Ambient dimension k',labelpad=6)
    right.set_ylabel(r'Asymptotic lower coefficient $\gamma_k$',labelpad=5)
    right.set_title('(b) Fixed-dimension asymptotics',loc='left',pad=11,fontweight='bold')
    right.text(31.8,.505,'Planar coefficient: 1/2',ha='right',va='bottom',color=AMBER,fontsize=8)
    right.text(15,.582,r'$\liminf_{n\to\infty}n(F(k,n)-1)\geq\gamma_k$',
               ha='center',va='center',fontsize=8.1,color=NAVY)
    fig.subplots_adjust(left=.085,right=.99,bottom=.19,top=.87)
    save_figure(fig,'selection_bounds')

    profile_data=[]
    for j in range(1,100):
        t=F(j,100)
        solution=solve_profile(5,4,t)
        c=F(solution['excess_minimax'])
        profile_data.append({'t':str(t),'excess_minimax':str(c),
                             'axial_fraction':solution['primal']['axial_fraction'],
                             'dual_count_types':solution['dual']['types']})
    ts=[float(F(row['t'])) for row in profile_data]
    cs=[float(F(row['excess_minimax'])) for row in profile_data]
    fig,ax=plt.subplots(figsize=(7,3.8))
    style_axes(ax)
    ax.axhline(float(F(1,7)),color=AMBER,linewidth=1.3,linestyle=(0,(5,3)),zorder=2)
    ax.plot(ts,cs,color=TEAL,linewidth=2,zorder=3)
    ax.scatter([.5],[float(F(3,20))],color=NAVY,s=42,zorder=5,
               edgecolors='white',linewidths=.9)
    ax.plot([.5,.5],[0,float(F(3,20))],color=NAVY,linewidth=.7,
            linestyle=(0,(2,4)),alpha=.45,zorder=1)
    ax.annotate(r'$\mathfrak{C}_{5,4}(1/2)=3/20$',xy=(.5,.15),xytext=(.65,.166),
                ha='center',va='center',color=NAVY,fontsize=10,
                arrowprops={'arrowstyle':'-','color':NAVY,'linewidth':.8,
                            'connectionstyle':'angle3,angleA=0,angleB=60'})
    ax.text(.985,float(F(1,7))+.005,'Planar excess reference: 1/7',
            ha='right',va='bottom',color=AMBER,fontsize=8.6)
    ax.set_xlim(0,1)
    ax.set_ylim(0,.18)
    ax.set_xticks([0,.1,.2,.3,.4,.5,.6,.7,.8,.9,1])
    ax.yaxis.set_major_locator(MultipleLocator(.03))
    ax.set_xlabel(r'Total rare-point probability $t$',labelpad=7)
    ax.set_ylabel('Exact fixed-profile minimax excess',labelpad=7)
    ax.set_title(r'Symmetric profile: $k=5$, $n=4$, $p=(1-t,t/5,\ldots,t/5)$',
                 loc='left',pad=12,fontweight='bold',fontsize=10.5)
    ax.text(.02,.169,'Worst case over all support geometries',
            color=SLATE,ha='left',va='center',fontsize=8.5)
    fig.subplots_adjust(left=.105,right=.985,bottom=.18,top=.86)
    save_figure(fig,'profile_minimax')

    data={
        'schema':'neiru-siha-figure-data-v1',
        'exact_arithmetic':'All rational values are Fraction strings. Surds are stored as exact rational parts, radical coefficients, and radicands.',
        'plotting_note':'Floating conversion is used only to draw figures. Lines join sampled values; no interpolation theorem is asserted.',
        'selection_bounds':{
            'dimension_four_budget_samples':budget_data,
            'asymptotic_dimension_samples':dimension_data,
            'claim_scope':'These are lower and upper bounds, not the exact unrestricted profile.'},
        'profile_minimax':{
            'k':5,'n':4,'sampling_grid':'t=j/100, j=1,...,99',
            'planar_reference':str(F(1,7)),
            'marked_exact_point':{'t':str(F(1,2)),'excess_minimax':str(F(3,20))},
            'samples':profile_data,
            'claim_scope':'Exact minimax over every support geometry for the specified symmetric mass vector; not the unrestricted F(d,n).'}
    }
    (HERE/'figure_data.json').write_text(json.dumps(data,indent=2)+'\n',encoding='utf-8')
    print(json.dumps({'status':'PASS','files':['figures/selection_bounds.pdf','figures/selection_bounds.png',
          'figures/profile_minimax.pdf','figures/profile_minimax.png','figure_data.json'],
          'budget_samples':len(budget_data),'dimension_samples':len(dimension_data),
          'exact_profile_samples':len(profile_data)},indent=2))


if __name__=='__main__':
    main()