forked from scikit-learn/scikit-learn
-
Notifications
You must be signed in to change notification settings - Fork 0
/
Copy pathbench_plot_neighbors.py
186 lines (150 loc) · 6.35 KB
/
bench_plot_neighbors.py
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
"""
Plot the scaling of the nearest neighbors algorithms with k, D, and N
"""
from time import time
import numpy as np
import pylab as pl
from matplotlib import ticker
from sklearn import neighbors, datasets
def get_data(N, D, dataset='dense'):
if dataset == 'dense':
np.random.seed(0)
return np.random.random((N, D))
elif dataset == 'digits':
X = datasets.load_digits().data
i = np.argsort(X[0])[::-1]
X = X[:, i]
return X[:N, :D]
else:
raise ValueError("invalid dataset: %s" % dataset)
def barplot_neighbors(Nrange=2 ** np.arange(1, 11),
Drange=2 ** np.arange(7),
krange=2 ** np.arange(10),
N=1000,
D=64,
k=5,
leaf_size=30,
dataset='digits'):
algorithms = ('kd_tree', 'brute', 'ball_tree')
fiducial_values = {'N': N,
'D': D,
'k': k}
#------------------------------------------------------------
# varying N
N_results_build = dict([(alg, np.zeros(len(Nrange)))
for alg in algorithms])
N_results_query = dict([(alg, np.zeros(len(Nrange)))
for alg in algorithms])
for i, NN in enumerate(Nrange):
print("N = %i (%i out of %i)" % (NN, i + 1, len(Nrange)))
X = get_data(NN, D, dataset)
for algorithm in algorithms:
nbrs = neighbors.NearestNeighbors(n_neighbors=min(NN, k),
algorithm=algorithm,
leaf_size=leaf_size)
t0 = time()
nbrs.fit(X)
t1 = time()
nbrs.kneighbors(X)
t2 = time()
N_results_build[algorithm][i] = (t1 - t0)
N_results_query[algorithm][i] = (t2 - t1)
#------------------------------------------------------------
# varying D
D_results_build = dict([(alg, np.zeros(len(Drange)))
for alg in algorithms])
D_results_query = dict([(alg, np.zeros(len(Drange)))
for alg in algorithms])
for i, DD in enumerate(Drange):
print("D = %i (%i out of %i)" % (DD, i + 1, len(Drange)))
X = get_data(N, DD, dataset)
for algorithm in algorithms:
nbrs = neighbors.NearestNeighbors(n_neighbors=k,
algorithm=algorithm,
leaf_size=leaf_size)
t0 = time()
nbrs.fit(X)
t1 = time()
nbrs.kneighbors(X)
t2 = time()
D_results_build[algorithm][i] = (t1 - t0)
D_results_query[algorithm][i] = (t2 - t1)
#------------------------------------------------------------
# varying k
k_results_build = dict([(alg, np.zeros(len(krange)))
for alg in algorithms])
k_results_query = dict([(alg, np.zeros(len(krange)))
for alg in algorithms])
X = get_data(N, DD, dataset)
for i, kk in enumerate(krange):
print("k = %i (%i out of %i)" % (kk, i + 1, len(krange)))
for algorithm in algorithms:
nbrs = neighbors.NearestNeighbors(n_neighbors=kk,
algorithm=algorithm,
leaf_size=leaf_size)
t0 = time()
nbrs.fit(X)
t1 = time()
nbrs.kneighbors(X)
t2 = time()
k_results_build[algorithm][i] = (t1 - t0)
k_results_query[algorithm][i] = (t2 - t1)
pl.figure('scikit-learn nearest neighbors benchmark results',
figsize=(8, 11))
for (sbplt, vals, quantity,
build_time, query_time) in [(311, Nrange, 'N',
N_results_build,
N_results_query),
(312, Drange, 'D',
D_results_build,
D_results_query),
(313, krange, 'k',
k_results_build,
k_results_query)]:
ax = pl.subplot(sbplt, yscale='log')
pl.grid(True)
tick_vals = []
tick_labels = []
bottom = 10 ** np.min([min(np.floor(np.log10(build_time[alg])))
for alg in algorithms])
for i, alg in enumerate(algorithms):
xvals = 0.1 + i * (1 + len(vals)) + np.arange(len(vals))
width = 0.8
c_bar = pl.bar(xvals, build_time[alg] - bottom,
width, bottom, color='r')
q_bar = pl.bar(xvals, query_time[alg],
width, build_time[alg], color='b')
tick_vals += list(xvals + 0.5 * width)
tick_labels += ['%i' % val for val in vals]
pl.text((i + 0.02) / len(algorithms), 0.98, alg,
transform=ax.transAxes,
ha='left',
va='top',
bbox=dict(facecolor='w', edgecolor='w', alpha=0.5))
pl.ylabel('Time (s)')
ax.xaxis.set_major_locator(ticker.FixedLocator(tick_vals))
ax.xaxis.set_major_formatter(ticker.FixedFormatter(tick_labels))
for label in ax.get_xticklabels():
label.set_rotation(-90)
label.set_fontsize(10)
title_string = 'Varying %s' % quantity
descr_string = ''
for s in 'NDk':
if s == quantity:
pass
else:
descr_string += '%s = %i, ' % (s, fiducial_values[s])
descr_string = descr_string[:-2]
pl.text(1.01, 0.5, title_string,
transform=ax.transAxes, rotation=-90,
ha='left', va='center', fontsize=20)
pl.text(0.99, 0.5, descr_string,
transform=ax.transAxes, rotation=-90,
ha='right', va='center')
pl.gcf().suptitle("%s data set" % dataset.capitalize(), fontsize=16)
pl.figlegend((c_bar, q_bar), ('construction', 'N-point query'),
'upper right')
if __name__ == '__main__':
barplot_neighbors(dataset='digits')
barplot_neighbors(dataset='dense')
pl.show()