-
Notifications
You must be signed in to change notification settings - Fork 1
/
Copy pathfina.py
266 lines (231 loc) · 7.58 KB
/
fina.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
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
# 引入库
import numpy as cp
import random as rd
# from scipy.optimize import minimize
import pickle
import json
N = 5
M = 3
L = 3
K = 1
mN = 3
mM = 2
mL = 2
mK = 1
mN -= 1
mM -= 1
mL -= 1
mK -= 1
class model:
def __init__(self):
self.c1 = cp.ones((N, M)) # 本组数据
self.c2 = cp.ones((L, K)) # 外组数据\
self.lamba = [cp.zeros((N, M)), cp.zeros((L, K))]
self.maxt = 2 # 最多遍历次数
# self.mine = 1000 # 最小误差
self.step = 10 # 步长
# self.mint = 1 # 最少遍历次数
self.istraining = 0
def save(self, files):
f = open(files, "wb")
s = [self.c1, self.c2]
pickle.dump(s, f)
f.close()
def read(self, files):
f = open(files, "rb")
s = pickle.load(f)
self.c1 = s[0]
self.c2 = s[1]
f.close()
def b(self, a=cp.array([])): # 标准化
# print(a)
mx = a.max()
for i in range(0, len(a)):
a[i] /= mx
return a
def change(self): # 处理变动函数
self.c1 += self.lamba[0] * self.getStep()
self.c2 += self.lamba[1] * self.getStep()
def getDistance(self, vector1=cp.array([]), vector2=cp.array([])):
# print((vector1-vector2),427)
return cp.sum((vector1 - vector2) ** 2)
def compute(self, tree=cp.array([[[]]]), l=cp.array([[]]), goal=cp.array([[]])):
self.lamba = [cp.zeros((N, M)), cp.zeros((L, K))]
for i in range(0, len(tree)):
w = self.dcf(goal[i], tree[i], l[i])
self.lamba[0] += w[0]
self.lamba[1] += w[1]
# print(r)
def fmax(self, func, x0, args=(), **kwargs):
return minimize(
lambda x, tree, l: -func(x, tree, l),
x0,
args=args,
# options={"maxiter": 100},
**kwargs
)
def realf(self, x=cp.array([]), tree=cp.array([[]]), l=cp.array([])):
return -self.allf(x, tree, l) ** 2
def allf(self, x=cp.array([]), tree=cp.array([[]]), l=cp.array([])):
rt = 0
# print(tree,l,12432)
treeDistance = cp.array([])
for j in range(0, len(tree)):
treeDistance = cp.append(treeDistance, self.getDistance(tree[j], x))
# print(treeDistance[j])
# print(treeDistance,234)
# if not self.istraining:
# print(treeDistance)
for i in range(0, len(treeDistance)):
rt += self.f(treeDistance, l, i)
return rt
def dcf(self, x=cp.array([]), tree=cp.array([[]]), l=cp.array([])):
# 对c的梯度
dc1 = cp.zeros((N, M)) # 本组数据
dc2 = cp.zeros((L, K)) # 外组数据
treeDistance = cp.array([])
for j in range(0, len(tree)):
treeDistance = cp.append(treeDistance, self.getDistance(tree[j], x))
for i in range(0, len(tree)):
w = self.f(treeDistance, l, i)
for i1 in range(0, N):
for i2 in range(0, M):
dc1 -= 2 * w * (treeDistance[i] ** (i1 - mN)) * (l[i] ** (i2 - mM))
for i1 in range(0, L):
for i2 in range(0, K):
dc2 -= 2 * w * (treeDistance[i] ** (i1 - mL)) * (l[i] ** (i2 - mK))
# print(dc1)
# print(dc2)
return [dc1, dc2]
def f(self, r=cp.array([]), l=cp.array([]), i=0):
j = 0
score = 0
# if not self.istraining:
# print(l)
# print(r,l)
while j < len(r):
if r[j] == 0:
r[j] += 0.0001
if j == i:
for i1 in range(0, N):
for i2 in range(0, M):
# if not self.istraining:
# print(r[j],i1,l[j],i2,j,self.c1[i1][i2],816)
score = (
score
+ (r[j] ** (i1 - mN))
* (l[j] ** (i2 - mM))
* self.c1[i1][i2]
)
else:
for i1 in range(0, L):
for i2 in range(0, K):
# print(r[j],i1,l[j],i2)
score = (
score
+ (r[j] ** (i1 - mL))
* (l[j] ** (i2 - mK))
* self.c2[i1][i2]
)
j = j + 1
return score
def getStep(self):
return cp.exp(-rd.random() * self.t) * self.step
def training(self, tree=cp.array([[[]]]), l=cp.array([[]]), goal=cp.array([[]])):
# 第一层是样本个数，第二层是对应基因树
self.istraining = 1
# self.residual = cp.zeros(
# (48, 3, len(tree))
# ) # 共4层，最外面一层是不同变量，第二层是移动量，第三层是不同样本
self.t = 0
tree = tree.astype(cp.float16)
for i in tree:
for j in tree:
self.b(j)
l = l.astype(cp.float16)
for i in l:
self.b(i)
goal = goal.astype(cp.float16)
for i in goal:
self.b(i)
print("begin")
while self.t < self.maxt:
self.compute(tree, l, goal)
self.change()
self.t += 1
print(self.t)
self.istraining = 0
def worked(self, tree=cp.array([[]]), l=cp.array([])):
tree = tree.astype(cp.float16)
mx = tree.max()
for i in tree:
self.b(i)
l = l.astype(cp.float16)
self.b(l)
return self.fmax(self.realf, cp.mean(tree, axis=0), args=(tree, l)).x * mx
def sdf(a=cp.array([[]])):
s = cp.array([])
for i in a:
for j in i:
s = cp.append(s, j)
return s
# ek = []
# eue = []
# euc = []
# for i in range(1,4):
# # i = 1
# k = []
# for j in range(5):
# print("READIN: %d %d" % (i,j))
# with open("output_%d\\output_%d%d_.json" % (i,i,j),"r") as f:
# data = json.load(f)
# # print(sdf(cp.array(data)).tolist())
# k.append(sdf(cp.array(data)).tolist())
# # print(k)
# # k = cp.array(k)
# with open("output_%d\\output_%dA_.json" % (i,i),"r") as f:
# ue = json.load(f)
# ue = sdf(cp.array(ue)).tolist()
# uc = [100 for i in range(100)]
# ek.append(k)
# eue.append(ue)
# euc.append(uc)
# a = model()
# a.training(
# cp.array(ek),
# cp.array(euc),
# cp.array(eue),
# )
# a.save("modelA.model")
def runmodel(modelpath, distances, lengths):
a = model()
a.read(modelpath)
return a.worked(cp.array(distances), cp.array(lengths)).tolist()
ek = []
eue = []
euc = []
for i in range(1,4):
# i = 1
k = []
for j in range(5):
print("READIN: %d %d" % (i,j))
with open("output_%d\\output_%d%d_.json" % (i,i,j),"r") as f:
data = json.load(f)
# print(sdf(cp.array(data)).tolist())
k.append(sdf(cp.array(data)).tolist())
# print(k)
# k = cp.array(k)
with open("output_%d\\output_%dA_.json" % (i,i),"r") as f:
ue = json.load(f)
ue = sdf(cp.array(ue)).tolist()
uc = [100 for i in range(100)]
ek.append(k)
eue.append(ue)
euc.append(uc)
a = model()
a.training(
cp.array(ek),
cp.array(euc),
cp.array(eue),
)
a.save("modelA.model")