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
|
#!/usr/bin/env python
#
# Author: Mike McKerns (mmckerns @uqfoundation)
# Copyright (c) 2024 The Uncertainty Quantification Foundation.
# License: 3-clause BSD. The full license text is available at:
# - https://github.com/uqfoundation/mystic/blob/master/LICENSE
'''
noisy emulator:
* input parameters (a_alpha, c_alpha, a_steel)
with solved values at:
[2.9306538, 4.6817646, 3.6026807]
measured error of:
[1e-4, 1e-4, 1e-4]
bounds at:
[(0.95 * i, 1.05 * i) for i in solved]
yielding a minimum at:
a_beta = 3.233392
with a measured error of:
a_beta_error = 1e-4
and gaussian noise on the input parameters of:
sigma = 0.001
'''
from itertools import repeat
from mystic.monitors import LoggingMonitor
from emulators import cost3, x3, bounds3
from noisy import noisy
try: # parallel maps
from pathos.maps import Map
from pathos.pools import ProcessPool, ThreadPool, SerialPool
cmap = Map(SerialPool) #ProcessPool
except ImportError:
cmap = map
# use a logging monitor to track cost function evaluations
monitor3 = LoggingMonitor(1, filename='truth.txt', label='truth')
# generate an objective with gaussian noise on the inputs
noisycost3 = lambda x: cost3(noisy(x, sigma=.001))
# generate a tracked cost function
def trackcost3(x):
fx = noisycost3(x)
monitor3(x, fx)
return fx
# average the cost over the number of cost evaluations per objective call
avecost3 = lambda x, n=1: sum(cmap(trackcost3, repeat(x, n)))/n
# set the number of cost evaluations per objective call
ave10cost3 = lambda x: avecost3(x, n=10)
|
