-
Notifications
You must be signed in to change notification settings - Fork 0
/
searches.py
205 lines (167 loc) · 7.02 KB
/
searches.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
import numpy as np
from random import randint
from random import random
import copy
# We will use Manhattan distance
def l1(p1, p2):
return np.sum(np.abs(np.array(p1) - np.array(p2)))
def l2(p1, p2):
return np.linalg.norm(np.array(p1) - np.array(p2))
def total_dist(data, dist_method=l1):
answer = 0
for i in range(len(data) - 1):
answer += dist_method(data[i], data[i + 1])
return answer
# =================MONTE CARLO=================#
def monte_carlo(cur_order, iters=100000, printDist=False, dist_method=l1):
best_cost = total_dist(cur_order, dist_method)
best_order = copy.deepcopy(cur_order)
for _ in range(iters):
cur_order = cur_order[np.random.permutation(cur_order.shape[0]), :]
cur_cost = total_dist(cur_order, dist_method)
if cur_cost < best_cost:
if printDist:
print(cur_cost, end=' ')
best_order = copy.deepcopy(cur_order)
best_cost = cur_cost
return best_order, best_cost
# =================RANDOM WALK=================#
def random_walk(cur_order, iters=100000, printDist=False, dist_method=l1):
def neighbour_dist(i, dist_method=l1):
answer = 0
if i != 0:
answer += dist_method(cur_order[i], cur_order[i - 1])
if i != len(cur_order) - 1:
answer += dist_method(cur_order[i], cur_order[i + 1])
return answer
cur_order = cur_order[np.random.permutation(cur_order.shape[0]), :]
best_order = copy.deepcopy(cur_order)
cur_cost = total_dist(cur_order, dist_method)
best_cost = cur_cost
n = len(cur_order)
for _ in range(iters):
i1 = randint(0, n - 1)
i2 = randint(0, n - 1)
# to be sure that i1 != i2
while i1 == i2:
i2 = randint(0, n - 1)
cur_cost -= neighbour_dist(i1, dist_method) + neighbour_dist(i2, dist_method)
cur_order[[i1, i2]] = cur_order[[i2, i1]]
cur_cost += neighbour_dist(i1, dist_method) + neighbour_dist(i2, dist_method)
if cur_cost < best_cost:
if printDist:
print(cur_cost, end=' ')
best_order = copy.deepcopy(cur_order)
best_cost = cur_cost
return best_order, best_cost
# =================HILL CLIMB=================#
# this is separate function for checking all neighbours
def try_all_swaps(cur_order, cur_cost, n, printDist=False, dist_method=l1):
def neighbour_dist(i, dist_method=l1):
answer = 0
if i != 0:
answer += dist_method(cur_order[i], cur_order[i - 1])
if i != len(cur_order) - 1:
answer += dist_method(cur_order[i], cur_order[i + 1])
return answer
best_order = copy.deepcopy(cur_order)
best_cost = cur_cost
for i1 in range(n):
for i2 in range(n):
if i1 == i2:
continue
this_cost = cur_cost
cur_cost -= neighbour_dist(i1, dist_method) + neighbour_dist(i2, dist_method)
cur_order[[i1, i2]] = cur_order[[i2, i1]]
cur_cost += neighbour_dist(i1, dist_method) + neighbour_dist(i2, dist_method)
if cur_cost < best_cost:
if printDist:
print(cur_cost, end=' ')
best_order = copy.deepcopy(cur_order)
best_cost = cur_cost
cur_order[[i1, i2]] = cur_order[[i2, i1]]
cur_cost = this_cost
return best_order, best_cost
def hill_climb(cur_order, iters=100000, printDist=False, dist_method=l1):
cur_order = cur_order[np.random.permutation(cur_order.shape[0]), :]
cur_cost = total_dist(cur_order, dist_method)
best_cost = cur_cost
best_order = copy.deepcopy(cur_order)
n = len(cur_order)
for _ in range(iters):
best_order, best_cost = try_all_swaps(best_order, best_cost, n, printDist=printDist, dist_method=dist_method)
return best_order, best_cost
# =================SIMULATED ANNEALING=================#
def annealing(cur_order, iters=100, delta_t=0.0001, t_max=100, printDist=False, dist_method=l1):
t_max = max(t_max, 2) # for protection
cur_order = cur_order[np.random.permutation(cur_order.shape[0]), :]
cur_order = cur_order.tolist()
n = len(cur_order)
cur_cost = total_dist(cur_order, dist_method)
best_order = copy.deepcopy(cur_order)
best_cost = cur_cost
for _ in range(iters):
for T in np.arange(t_max, 1, -1 * delta_t):
i1 = randint(0, n - 1)
i2 = randint(i1 + 1, n)
cur_order[i1: i2] = reversed(cur_order[i1: i2])
new_cost = total_dist(cur_order, dist_method)
dE = cur_cost - new_cost
if dE > 0 and np.exp(-dE / T) > random():
cur_cost = new_cost
if cur_cost < best_cost:
if printDist:
print(cur_cost, end=' ')
best_cost = cur_cost
best_order = copy.deepcopy(cur_order)
else:
cur_order[i1: i2] = reversed(cur_order[i1: i2])
return np.array(best_order), best_cost
# =================GENETIC ALGORITHM=================#
def generate_pair(n):
i1 = randint(0, n - 1)
return i1, randint(i1 + 1, n)
def crossover(parent1, parent2):
i1, i2 = generate_pair(len(parent1))
child = [[-1, -1] for _ in range(len(parent1))]
child[i1:i2] = parent1[i1:i2]
pivot = 0
for el in parent2:
if pivot == i1:
pivot = i2
if el not in parent1[i1:i2]:
child[pivot] = el
pivot += 1
return child
def genetic_algo(cur_order, iters, size=500, sel_rate=0.1, mut_rate=0.05, printDist=False, dist_method=l1):
mut_rate = min(1.0, mut_rate) # for protection
best_order = copy.deepcopy(cur_order)
best_cost = total_dist(best_order, dist_method)
world = [cur_order[np.random.permutation(cur_order.shape[0]), :].tolist() for _ in range(size)]
n = len(cur_order)
for i in range(iters):
# Selection: take best 10%
cur_relatives = sorted(world, key=total_dist)[0:int(size * sel_rate)]
world = []
# Making offsprings using crossover
for _ in range(size):
p1, p2 = generate_pair(int(size * sel_rate) - 1) # choose 2 random persons
new_order = crossover(cur_relatives[p1], cur_relatives[p2])
new_cost = total_dist(new_order, dist_method)
world.append(new_order)
if new_cost < best_cost:
if printDist:
print(new_cost, end=' ')
best_cost = new_cost
best_order = copy.deepcopy(new_order)
# Big mutations (reverse subarrays)
for _ in range(int(size * mut_rate)):
j = randint(0, size - 1)
i1, i2 = generate_pair(n)
world[j][i1: i2] = reversed(world[j][i1: i2])
# Make small mutations (swaps)
for _ in range(int(size * mut_rate)):
j = randint(0, size - 1)
i1, i2 = generate_pair(n - 1)
world[j][i1], world[j][i2] = world[j][i2], world[j][i1]
return np.array(best_order), best_cost