-
Notifications
You must be signed in to change notification settings - Fork 1
/
simplex.py
284 lines (246 loc) · 11.2 KB
/
simplex.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
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
#!/usr/bin/env python
#
# -*- Mode: python -*-
#
# $Id: Simplex.py,v 1.2 2004/05/31 14:01:06 vivake Exp $
#
# Copyright (c) 2002-2004 Vivake Gupta (vivakeATlab49.com). All rights reserved.
#
# This program is free software; you can redistribute it and/or
# modify it under the terms of the GNU General Public License as
# published by the Free Software Foundation; either version 2 of the
# License, or (at your option) any later version.
#
# This program is distributed in the hope that it will be useful,
# but WITHOUT ANY WARRANTY; without even the implied warranty of
# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
# GNU General Public License for more details.
#
# You should have received a copy of the GNU General Public License
# along with this program; if not, write to the Free Software
# Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307
# USA
#
# This software is maintained by Vivake (vivakeATlab49.com) and is available at:
# http://shell.lab49.com/~vivake/python/Simplex.py
#
# 1.2 ( 5/2004) - Fixed a bug found by Noboru Yamamoto <noboru.yamamotoATkek.jp>
# which caused minimize() not to converge, and reach the maxiter
# limit under some conditions.
# ( 1/2011) - Added **kwargs where appropriate to enable passing additional
# static parameters to the objective function (Filip Dominec)
""" Simplex - a regression method for arbitrary nonlinear function minimization
Simplex minimizes an arbitrary nonlinear function of N variables by the
Nedler-Mead Simplex method as described in:
Nedler, J.A. and Mead, R. "A Simplex Method for Function Minimization."
Computer Journal 7 (1965): 308-313.
It makes no assumptions about the smoothness of the function being minimized.
It converges to a local minimum which may or may not be the global minimum
depending on the initial guess used as a starting point.
"""
import math
import copy
import csv
import time
#from numpy import ndarray as nd
class Simplex:
def __init__(self, testfunc, guess, increments, kR = -1, kE = 2, kC = 0.5):
"""Initializes the simplex.
INPUTS
------
testfunc the function to minimize
guess[] an list containing initial guesses
increments[] an list containing increments, perturbation size
kR reflection constant
kE expansion constant
kC contraction constant
"""
self.testfunc = testfunc
self.guess = guess
self.increments = increments
self.kR = kR
self.kE = kE
self.kC = kC
self.numvars = len(self.guess)
load_time = time.strftime('%Y.%m.%d.%H.%M')
self.filename = "simplex_output_{}.log".format(load_time)
def minimize(self, epsilon = 0.0001, maxiters = 250, monitor = 1, **kwargs):
"""Walks to the simplex down to a local minima.
INPUTS
------
epsilon convergence requirement
maxiters maximum number of iterations
monitor if non-zero, progress info is output to stdout
OUTPUTS
-------
an array containing the final values
lowest value of the error function
number of iterations taken to get here
"""
self.simplex = []
self.lowest = -1
self.highest = -1
self.secondhighest = -1
self.errors = []
self.currenterror = 0
# Initialize vertices
for vertex in range(0, self.numvars + 3): # Two extras to store centroid and reflected point
self.simplex.append(copy.copy(self.guess))
# Use initial increments
for vertex in range(0, self.numvars + 1):
for x in range(0, self.numvars):
if x == (vertex - 1):
self.simplex[vertex][x] = self.guess[x] + self.increments[x]
self.errors.append(0)
self.calculate_errors_at_vertices(**kwargs)
if monitor:
with open(self.filename, 'wt') as save_file:
writer = csv.writer(save_file, dialect="excel-tab")
row = ['Step','Error']
for i in range(self.numvars):
row.append('Var {} '.format(i))
writer.writerow(row)
iter = 0
for iter in range(0, maxiters):
# Identify highest, second highest, and lowest vertices
self.highest = 0
self.lowest = 0
for vertex in range(0, self.numvars + 1):
if self.errors[vertex] > self.errors[self.highest]:
self.highest = vertex
if self.errors[vertex] < self.errors[self.lowest]:
self.lowest = vertex
self.secondhighest = 0
for vertex in range(0, self.numvars + 1):
if vertex == self.highest:
continue
if self.errors[vertex] > self.errors[self.secondhighest]:
self.secondhighest = vertex
# Test for convergence
S = 0.0
S1 = 0.0
for vertex in range(0, self.numvars + 1):
S = S + self.errors[vertex]
F2 = S / (self.numvars + 1)
for vertex in range(0, self.numvars + 1):
S1 = S1 + (self.errors[vertex] - F2)**2
T = math.sqrt(S1 / self.numvars)
# Optionally, print progress information
if monitor:
with open(self.filename, 'a') as save_file:
writer = csv.writer(save_file, dialect="excel-tab")
row = ["{} ".format(iter), "{:.4f}".format(self.errors[self.highest])]
for var in self.simplex[self.highest]:
row.append("{:.4e}".format(var))
writer.writerow(row)
if T <= epsilon: # We converged! Break out of loop!
break;
else: # Didn't converge. Keep crunching.
# Calculate centroid of simplex, excluding highest vertex
for x in range(0, self.numvars):
S = 0.0
for vertex in range(0, self.numvars + 1):
if vertex == self.highest:
continue
S = S + self.simplex[vertex][x]
self.simplex[self.numvars + 1][x] = S / self.numvars
self.reflect_simplex()
self.currenterror = self.testfunc(self.guess, **kwargs)
if self.currenterror < self.errors[self.lowest]:
tmp = self.currenterror
self.expand_simplex()
self.currenterror = self.testfunc(self.guess, **kwargs)
if self.currenterror < tmp:
self.accept_expanded_point()
else:
self.currenterror = tmp
self.accept_reflected_point()
elif self.currenterror <= self.errors[self.secondhighest]:
self.accept_reflected_point()
elif self.currenterror <= self.errors[self.highest]:
self.accept_reflected_point()
self.contract_simplex()
self.currenterror = self.testfunc(self.guess, **kwargs)
if self.currenterror < self.errors[self.highest]:
self.accept_contracted_point()
else:
self.multiple_contract_simplex(**kwargs)
elif self.currenterror >= self.errors[self.highest]:
self.contract_simplex()
self.currenterror = self.testfunc(self.guess, **kwargs)
if self.currenterror < self.errors[self.highest]:
self.accept_contracted_point()
else:
self.multiple_contract_simplex(**kwargs)
# Either converged or reached the maximum number of iterations.
# Return the lowest vertex and the currenterror.
for x in range(0, self.numvars):
self.guess[x] = self.simplex[self.lowest][x]
self.currenterror = self.errors[self.lowest]
if monitor:
with open(self.filename, 'a') as save_file:
writer = csv.writer(save_file, dialect="excel-tab")
row = [iter + 1, self.currenterror]
for var in self.guess:
row.append(var)
writer.writerow(row)
with open('best.csv', 'wb') as save_file:
writer = csv.writer(save_file, dialect='excel-tab')
writer.writerow(self.guess)
return self.guess, self.currenterror, iter
def contract_simplex(self):
for x in range(0, self.numvars):
self.guess[x] = self.kC * self.simplex[self.highest][x] + (1 - self.kC) * self.simplex[self.numvars + 1][x]
return
def expand_simplex(self):
for x in range(0, self.numvars):
self.guess[x] = self.kE * self.guess[x] + (1 - self.kE) * self.simplex[self.numvars + 1][x]
return
def reflect_simplex(self):
for x in range(0, self.numvars):
self.guess[x] = self.kR * self.simplex[self.highest][x] + (1 - self.kR) * self.simplex[self.numvars + 1][x]
self.simplex[self.numvars + 2][x] = self.guess[x] # REMEMBER THE REFLECTED POINT
return
def multiple_contract_simplex(self, **kwargs):
for vertex in range(0, self.numvars + 1):
if vertex == self.lowest:
continue
for x in range(0, self.numvars):
self.simplex[vertex][x] = 0.5 * (self.simplex[vertex][x] + self.simplex[self.lowest][x])
self.calculate_errors_at_vertices(**kwargs)
return
def accept_contracted_point(self):
self.errors[self.highest] = self.currenterror
for x in range(0, self.numvars):
self.simplex[self.highest][x] = self.guess[x]
return
def accept_expanded_point(self):
self.errors[self.highest] = self.currenterror
for x in range(0, self.numvars):
self.simplex[self.highest][x] = self.guess[x]
return
def accept_reflected_point(self):
self.errors[self.highest] = self.currenterror
for x in range(0, self.numvars):
self.simplex[self.highest][x] = self.simplex[self.numvars + 2][x]
return
def calculate_errors_at_vertices(self,**kwargs):
for vertex in range(0, self.numvars + 1):
if vertex == self.lowest:
continue
for x in range(0, self.numvars):
self.guess[x] = self.simplex[vertex][x]
self.currenterror = self.testfunc(self.guess, **kwargs)
self.errors[vertex] = self.currenterror
return
def my_objective_function(args, static_par):
return abs(- (args[0]-static_par[0])**2 - (args[1]-static_par[1])**2)
def main():
s = Simplex(my_objective_function, [1, 1, 1], [.01, .01, .01])
values, err, iter = s.minimize(epsilon = 0.0000001, maxiters = 250, monitor = 1, static_par=[2,4])
del(s)
print 'args = ', values
print 'error = ', err
print 'iterations = ', iter
if __name__ == '__main__':
main()