-
Notifications
You must be signed in to change notification settings - Fork 5
/
functions.py
278 lines (235 loc) · 8.48 KB
/
functions.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
from febo.utils.config import ConfigField, assign_config
from .benchmarks import BenchmarkEnvironment, BenchmarkEnvironmentConfig
import numpy as np
from febo.environment.domain import DiscreteDomain, ContinuousDomain
class FiniteLinearBandit(BenchmarkEnvironment):
"""
Quick sketch of a finite linear bandit
"""
def __init__(self, path=None):
super().__init__(path=path)
self._theta = np.ones(self.config.dimension)
self._theta = self._theta / np.linalg.norm(self._theta)
np.random.seed(self.seed)
self._domain_points = np.random.multivariate_normal(np.zeros(self.config.dimension),
np.eye(self.config.dimension),
size=self.config.num_domain_points)
np.random.seed() # reset to random state
self._domain_points = self._domain_points / np.maximum(np.linalg.norm(self._domain_points, axis=-1),
np.ones(self.config.num_domain_points)).reshape(-1, 1)
self._domain = DiscreteDomain(self._domain_points)
self._max_value = self._get_max_value()
@property
def name(self):
return "Finite Linear Bandit"
@property
def _requires_random_seed(self):
return True
def f(self, x):
return np.dot(x, self._theta)
def _get_max_value(self):
return np.max(self.f(self._domain_points))
class Camelback1D(BenchmarkEnvironment):
"""
1d Test Function
"""
def __init__(self, path=None):
super().__init__(path)
self._x = np.array([1.])
self._max_value = 1.0026469
self._domain = ContinuousDomain(np.array([-1]), np.array([2]))
def f(self, x):
return np.exp(-np.square(x-1.5)/0.05) + 1.98/(1+np.square(x-0.5)) - 1
class Camelback(BenchmarkEnvironment):
"""
Camelback benchmark function.
"""
def __init__(self, path=None):
super().__init__(path)
self._x0 = np.array([0.5, 0.2])
self._x0 = np.array([-0.12977758051079197, 0.2632096107305229])
self._max_value = 1.03162842
self._domain = ContinuousDomain(np.array([-2,-1]), np.array([2,1]))
def f(self, x):
x = np.atleast_2d(x)
xx = x[:,0]
yy = x[:,1]
y = (4. - 2.1*xx**2 + (xx**4)/3.)*(xx**2) + xx*yy + (-4. + 4*(yy**2))*(yy**2)
return np.maximum(-y, -2.5)
class Camelmod(BenchmarkEnvironment):
"""
Camelmod benchmark function with local minima
"""
def __init__(self, path=None):
super().__init__(path)
self._x0 = np.array([0., 0.])
self._max_value = 1.
self._domain = ContinuousDomain(np.array([-0.5,-0.5]), np.array([0.5,0.5]))
def f(self, X):
X = np.atleast_2d(X)
X, Y = X[:, 0], X[:, 1]
return ((8 * X) ** 4 - 16. * (8 * X) ** 2 + 5 * (8 * X) + (8 * Y) ** 4 - 16. * (8 * Y) ** 2 + 5 * (
8 * Y)) / -156.66466273
class GaussianConfig(BenchmarkEnvironmentConfig):
initial_value = ConfigField(0.1)
_section = 'environment.benchmark.gaussian'
@assign_config(GaussianConfig)
class Gaussian(BenchmarkEnvironment):
"""
Camelback benchmark function.
"""
def __init__(self, path=None):
super().__init__(path)
ones = np.ones(self.config.dimension)
self._dist_initial = np.sqrt(np.log(1 / self.config.initial_value) / 4)
self._x0 = self._dist_initial * ones / np.sqrt(self.config.dimension)
self._max_value = 1.0
self._domain = ContinuousDomain(-ones, ones)
def _get_random_initial_point(self):
dir = np.random.normal(size=self.config.dimension)
return self._dist_initial * dir / np.linalg.norm(dir)
def f(self, X):
X = np.atleast_2d(X)
Y = np.exp(-4*np.sum(np.square(X), axis=1))
return Y
class Quadratic(BenchmarkEnvironment):
"""
Camelback benchmark function.
"""
def __init__(self, path=None):
super().__init__(path)
ones = np.ones(self.config.dimension)
self._x0 = 0.5*ones/np.sqrt(self.config.dimension)
self._max_value = 1.0
self._domain = ContinuousDomain(-ones, ones)
def f(self, X):
X = np.atleast_2d(X)
Y = 2*np.sum(np.square(X), axis=1)
return 1 - Y
class CamelbackEmbedded(BenchmarkEnvironment):
"""
Camelback benchmark function.
"""
def __init__(self, path=None):
super().__init__(path)
self._max_value = 1.03162842
d = self.config.dimension
if d <= 2:
raise Exception("Need dimension at least 3 to create embedded version of Camelback")
self._x0 = np.array([0.5, 0.2] + [0.]*(d-2))
self._domain = ContinuousDomain(np.array([-2,-1] + [-1]*(d-2)), np.array([2,1]+ [1]*(d-2)))
def f(self, x):
xx = x[0]
yy = x[1]
y = (4. - 2.1*xx**2 + (xx**4)/3.)*(xx**2) + xx*yy + (-4. + 4*(yy**2))*(yy**2)
return -y
class LinSin1D(BenchmarkEnvironment):
"""
d=1 benchmark function
"""
def __init__(self, path=None):
super().__init__(path)
self._x = np.array([15])
self._max_value = 1.25375424 # determined using scipy.minimze
self._domain = ContinuousDomain(np.array([-20]), np.array([20]))
def f(self, X):
return 10. + 0.05*X + np.sin(X-5)/(X-5) - 10
class CosUnique1D(BenchmarkEnvironment):
"""
d=1 benchmark function
"""
def __init__(self, path=None):
super().__init__(path)
self._x = np.array([0.1])
self._max_value = 1.1 # at 0.5
self._domain = ContinuousDomain(np.array([0.]), np.array([1]))
def f(self, X):
return -np.cos(10*np.pi*X) + 0.1 - 0.1*np.abs(X-0.5)
#
# class DSafetyConstraintsEnv(BenchmarkEnvironment):
# """ implements a _get_saftey_constraint methods, which returns self.domain_dimension safety constraints. """
# def _get_safety_constraints(self, x):
# constraints = []
# # if self.config.TEST_ENVS_USE_SAFETY_CONSTRAINTS:
# # for i in range(self.domain_dimension):
# # constraints.append(0.5 * math.cos(10 * x[i] - i % 2) - 1 + 2 * x[max(0,i-1)])
#
# return np.array(constraints)
#
#
# def initialize(self):
# if self.config.TEST_ENVS_USE_SAFETY_CONSTRAINTS:
# # self._num_safety_constraints = self.domain_dimension
# self._lower_bound_objective_value = 0.3
# else:
# self._num_safety_constraints = 0
# self._lower_bound_objective_value = None
# super(DSafetyConstraintsEnv, self).initialize()
#
# class Simple2D(DSafetyConstraintsEnv):
#
# @property
# def domain_dimension(self):
# return 2
#
# def _f(self, x):
# return math.sin(.3*x[0]) + .3*math.cos(1.3*x[1]*x[0])
# class Simple1D(DSafetyConstraintsEnv):
#
# @property
# def domain_dimension(self):
# return 1
#
# def f(self, x):
# return math.sin(5*x[0]) + 0.5 *math.cos(10*x[0])
# class Michalewicz(DSafetyConstraintsEnv):
#
# def __init__(self, *args, **kwargs):
# super(Michalewicz, self).__init__(*args, **kwargs)
# self.d = 2
#
# def initialize(self):
# super(Michalewicz, self).initialize()
#
# # overwrite initial point
# self._current_x = np.array([0.2]*self.d)
# self.set_parameters(self._current_x) # refresh objective value
# self._lower_bound_objective_value = -2.0 if self.config.TEST_ENVS_USE_SAFETY_CONSTRAINTS else None
#
# @property
# def domain_dimension(self):
# return self.d
#
# def f(self, x):
# x = x / np.pi
# (d) = x.shape
# ar = np.arange(1,self.d+1,1)
# sum_ = np.sin(x) * np.power((np.sin(ar * np.power(x, 2) / np.pi)), (2*d))
# sum_ = np.sum(sum_)
# return -0.5*sum_
#
# class SineMultiDim(DSafetyConstraintsEnv):
#
# def __init__(self, *args, **kwargs):
# super(SineMultiDim, self).__init__(*args, **kwargs)
# self._dim = 2
#
# def initialize(self):
# super(Camelback, self).initialize()
#
# # overwrite initial point
# self._current_x = np.array([0.2] * self.domain_dimension)
# self.set_parameters(self._current_x) # refresh objective value
# self._lower_bound_objective_value = None
#
# def set_dimension(self, dim):
# self._dim = dim
#
# @property
# def domain_dimension(self):
# return 2
#
# def f(self, x):
# y = -np.sin(np.sum(x**2)) + 1
# return y
#