-
Notifications
You must be signed in to change notification settings - Fork 0
/
my_sa_box.py
223 lines (184 loc) · 6.16 KB
/
my_sa_box.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
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
# Chris Riederer
# 2017-05-27
"""Use simulated annealing to move a point cloud to an X"""
import math
import random
import numpy as np
from numpy.random import normal
import pandas as pd
import pylab as pl
import simanneal
from shapely.geometry import LinearRing, Polygon, Point, MultiLineString
# Generate data set
np.random.seed(1337)
points = np.random.random((182, 2)) * 100
# points = [Point(p) for p in points]
# Create target Polygon
# poly = Polygon([(5, 20), (5, 80), (95, 80), (95, 20)]) # Rectangle
coords = [((15, 15), (15, 15)),
((15, 50), (15, 50)),
((15, 85), (15, 85)),
((50, 15), (50, 15)),
((50, 50), (50, 50)),
((50, 85), (50, 85)),
((85, 15), (85, 15)),
((85, 50), (85, 50)),
((85, 85), (85, 85)),
]
poly_ext = MultiLineString(coords)
poly = poly_ext
# poly_ext = LinearRing(poly.exterior.coords)
# poly_ext.xy = []
# px, py = poly_ext.xy
# Turn interactive mode on for plotting
pl.ion()
def plot_points(points, poly_ext, orig_points=False):
if orig_points:
orig_x = [p.x for p in orig_points]
orig_y = [p.y for p in orig_points]
pl.scatter(orig_x, orig_y, color="red", s=50)
x = [p.x for p in points]
y = [p.y for p in points]
pl.scatter(x, y)
pl.plot(px, py)
pl.xlim(0, 100)
pl.ylim(0, 100)
pl.pause(0.0001)
pl.show()
def plot_pointsNP(points, poly_ext, orig_points=None):
if orig_points is not None:
pl.scatter(orig_points[:, 0], orig_points[:, 1], color="blue", s=50, alpha=0.5)
x = points[:, 0]
y = points[:, 1]
# px, py = poly_ext.xy
# pl.plot(px, py, alpha=0.5)
pl.scatter(x, y, color="red")
pl.xlim(0, 100)
pl.ylim(0, 100)
pl.pause(0.0001)
pl.show()
def points_from_numpy(arr):
return []
# Utility functions
def get_x(points):
return points[:, 0] # [p.x for p in points]
def get_y(points):
return points[:, 1] #[p.y for p in points]
def get_point_mean_x(points):
return sum(get_x(points)) / len(points)
def get_point_mean_y(points):
return sum(get_y(points)) / len(points)
def get_point_sd_x(points):
return np.array(get_x(points)).std()
def get_point_sd_y(points):
return np.array(get_y(points)).std()
def get_stats(points):
return [
get_point_mean_x(points),
get_point_mean_y(points),
get_point_sd_x(points),
get_point_sd_y(points),
]
def print_stats(points):
print("%.6f\t%.6f\t%.6f\t%.6f" % tuple(get_stats(points)))
def write_points(points, fname):
df = pd.DataFrame()
df["x"] = get_x(points)
df["y"] = get_y(points)
df.to_csv(fname, index=False)
def TestAvgDifference(points, scale, num_loops=20):
orig_stats = np.array(get_stats(points))
new_stats = np.array([get_stats(MoveRandomPoints(points, scale))
for _ in range(num_loops)])
return np.abs(new_stats - orig_stats).mean(axis=0)
def TestErrorPassRate(points, scale, num_loops=100):
orig_stats = np.array(get_stats(points))
passes = np.array([IsErrorOk(points, MoveRandomPoints(points, scale))
for _ in range(num_loops)])
return passes.mean()
def Fit(points):
return -sum([poly_ext.distance(p) for p in points])
def FitOverall(points):
return -sum([poly_ext.distance(Point(x, y)) for x, y in points])
def Fit1(point):
return -poly_ext.distance(Point(point[0], point[1]))
def PerturbNP(ds, temp, scale=0.1):
# loop: ?
for i in range(1000):
# Move a random point
rand_idx = np.random.randint(len(ds))
old_point = ds[rand_idx, :]
new_point = old_point + normal(scale=scale, size= 2)
if temp > np.random.random() or Fit1(new_point) > Fit1(old_point):
out = ds.copy()
out[rand_idx, :] = new_point
return out
print("failed to pass")
return ds
def Perturb(ds, temp):
# loop: ?
for i in range(1000):
test = MoveRandomPoints(ds)
if temp > np.random.random() or Fit(test) > Fit(ds):
return test
print("failed to pass")
return ds
def constraint(old, new):
"""Constraint: first two decimal places remain the same"""
return math.floor(old*100) == math.floor(new*100)
def IsErrorOkNP(test_ds, initial_ds):
x_new = np.array(get_x(test_ds))
y_new = np.array(get_y(test_ds))
x_old = np.array(get_x(initial_ds))
y_old = np.array(get_y(initial_ds))
return (
constraint(x_new.mean(), x_old.mean())
and constraint(y_new.mean(), y_old.mean())
and constraint(x_new.std(), x_old.std())
and constraint(y_new.std(), y_old.std())
)
def IsErrorOkNP2(test_ds, initial_ds):
return (
constraint(test_ds[:, 0].mean(), initial_ds[:, 0].mean())
and constraint(test_ds[:, 1].mean(), initial_ds[:, 1].mean())
and constraint(test_ds[:, 0].std(), initial_ds[:, 0].std())
and constraint(test_ds[:, 1].std(), initial_ds[:, 1].std())
)
def IsErrorOkOrig(test_ds, initial_ds):
return (
constraint(get_point_mean_x(test_ds), get_point_mean_x(initial_ds))
and constraint(get_point_mean_y(test_ds), get_point_mean_y(initial_ds))
and constraint(get_point_sd_x(test_ds), get_point_sd_x(initial_ds))
and constraint(get_point_sd_y(test_ds), get_point_sd_y(initial_ds))
)
def MoveRandomPoints(ds, scale=0.019):
# Move between 1 and 10 points
num_points = np.random.randint(1, 30)
random.shuffle(ds)
update_points = ds[:num_points]
update_points = [Point(p.x + normal(scale=scale), p.y + normal(scale=scale))
for p in update_points]
new_points = update_points + ds[num_points:]
return new_points
NUM_LOOPS = 200000
initial_ds = points.copy()
current_ds = points
temperatures = np.logspace(np.log10(0.4), np.log10(0.1), num=NUM_LOOPS)
pl.figure()
pl.show()
for i, temp in enumerate(temperatures):
# test_ds = Perturb(current_ds, temp)
test_ds = PerturbNP(current_ds, temp)
if IsErrorOkNP2(test_ds, initial_ds):
current_ds = test_ds
if i % 100 == 0:
# print(i, "%.4f" % Fit(current_ds), end="\t")
print(i, "%.4f" % FitOverall(current_ds), sep="\t", end="\t")
print_stats(current_ds)
# print_stats(initial_ds)
pl.clf()
plot_pointsNP(current_ds, poly_ext, orig_points=initial_ds)
# plot_points(current_ds, poly_ext, orig_points=initial_ds)
write_points(current_ds, "grid_points.csv")
pl.ioff()
plot_pointsNP(current_ds, poly_ext, orig_points=initial_ds)