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
|
"""
===================================================
Comparing initial sampling methods on integer space
===================================================
Holger Nahrstaedt 2020 Sigurd Carlsen October 2019
.. currentmodule:: skopt
When doing baysian optimization we often want to reserve some of the
early part of the optimization to pure exploration. By default the
optimizer suggests purely random samples for the first n_initial_points
(10 by default). The downside to this is that there is no guarantee that
these samples are spread out evenly across all the dimensions.
Sampling methods as Latin hypercube, Sobol', Halton and Hammersly
take advantage of the fact that we know beforehand how many random
points we want to sample. Then these points can be "spread out" in
such a way that each dimension is explored.
See also the example on a real space
:ref:`sphx_glr_auto_examples_initial_sampling_method.py`
"""
print(__doc__)
import numpy as np
np.random.seed(1234)
import matplotlib.pyplot as plt
from scipy.spatial.distance import pdist
from skopt.sampler import Grid, Halton, Hammersly, Lhs, Sobol
from skopt.space import Space
#############################################################################
def plot_searchspace(x, title):
fig, ax = plt.subplots()
plt.plot(np.array(x)[:, 0], np.array(x)[:, 1], 'bo', label='samples')
plt.plot(np.array(x)[:, 0], np.array(x)[:, 1], 'bs', markersize=40, alpha=0.5)
# ax.legend(loc="best", numpoints=1)
ax.set_xlabel("X1")
ax.set_xlim([0, 5])
ax.set_ylabel("X2")
ax.set_ylim([0, 5])
plt.title(title)
ax.grid(True)
n_samples = 10
space = Space([(0, 5), (0, 5)])
#############################################################################
# Random sampling
# ---------------
x = space.rvs(n_samples)
plot_searchspace(x, "Random samples")
pdist_data = []
x_label = []
print("empty fields: %d" % (36 - np.size(np.unique(x, axis=0), 0)))
pdist_data.append(pdist(x).flatten())
x_label.append("random")
#############################################################################
# Sobol'
# ------
sobol = Sobol()
x = sobol.generate(space.dimensions, n_samples)
plot_searchspace(x, "Sobol'")
print("empty fields: %d" % (36 - np.size(np.unique(x, axis=0), 0)))
pdist_data.append(pdist(x).flatten())
x_label.append("sobol'")
#############################################################################
# Classic latin hypercube sampling
# --------------------------------
lhs = Lhs(lhs_type="classic", criterion=None)
x = lhs.generate(space.dimensions, n_samples)
plot_searchspace(x, 'classic LHS')
print("empty fields: %d" % (36 - np.size(np.unique(x, axis=0), 0)))
pdist_data.append(pdist(x).flatten())
x_label.append("lhs")
#############################################################################
# Centered latin hypercube sampling
# ---------------------------------
lhs = Lhs(lhs_type="centered", criterion=None)
x = lhs.generate(space.dimensions, n_samples)
plot_searchspace(x, 'centered LHS')
print("empty fields: %d" % (36 - np.size(np.unique(x, axis=0), 0)))
pdist_data.append(pdist(x).flatten())
x_label.append("center")
#############################################################################
# Maximin optimized hypercube sampling
# ------------------------------------
lhs = Lhs(criterion="maximin", iterations=10000)
x = lhs.generate(space.dimensions, n_samples)
plot_searchspace(x, 'maximin LHS')
print("empty fields: %d" % (36 - np.size(np.unique(x, axis=0), 0)))
pdist_data.append(pdist(x).flatten())
x_label.append("maximin")
#############################################################################
# Correlation optimized hypercube sampling
# ----------------------------------------
lhs = Lhs(criterion="correlation", iterations=10000)
x = lhs.generate(space.dimensions, n_samples)
plot_searchspace(x, 'correlation LHS')
print("empty fields: %d" % (36 - np.size(np.unique(x, axis=0), 0)))
pdist_data.append(pdist(x).flatten())
x_label.append("corr")
#############################################################################
# Ratio optimized hypercube sampling
# ----------------------------------
lhs = Lhs(criterion="ratio", iterations=10000)
x = lhs.generate(space.dimensions, n_samples)
plot_searchspace(x, 'ratio LHS')
print("empty fields: %d" % (36 - np.size(np.unique(x, axis=0), 0)))
pdist_data.append(pdist(x).flatten())
x_label.append("ratio")
#############################################################################
# Halton sampling
# ---------------
halton = Halton()
x = halton.generate(space.dimensions, n_samples)
plot_searchspace(x, 'Halton')
print("empty fields: %d" % (36 - np.size(np.unique(x, axis=0), 0)))
pdist_data.append(pdist(x).flatten())
x_label.append("halton")
#############################################################################
# Hammersly sampling
# ------------------
hammersly = Hammersly()
x = hammersly.generate(space.dimensions, n_samples)
plot_searchspace(x, 'Hammersly')
print("empty fields: %d" % (36 - np.size(np.unique(x, axis=0), 0)))
pdist_data.append(pdist(x).flatten())
x_label.append("hammersly")
#############################################################################
# Grid sampling
# -------------
grid = Grid(border="include", use_full_layout=False)
x = grid.generate(space.dimensions, n_samples)
plot_searchspace(x, 'Grid')
print("empty fields: %d" % (36 - np.size(np.unique(x, axis=0), 0)))
pdist_data.append(pdist(x).flatten())
x_label.append("grid")
#############################################################################
# Pdist boxplot of all methods
# ----------------------------
#
# This boxplot shows the distance between all generated points using
# Euclidian distance. The higher the value, the better the sampling method.
# It can be seen that random has the worst performance
fig, ax = plt.subplots()
ax.boxplot(pdist_data)
plt.grid(True)
plt.ylabel("pdist")
_ = ax.set_ylim(0, 6)
_ = ax.set_xticklabels(x_label, rotation=45, fontsize=8)
|
