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
|
/*
* Licensed to the Apache Software Foundation (ASF) under one or more
* contributor license agreements. See the NOTICE file distributed with
* this work for additional information regarding copyright ownership.
* The ASF licenses this file to You under the Apache License, Version 2.0
* (the "License"); you may not use this file except in compliance with
* the License. You may obtain a copy of the License at
*
* http://www.apache.org/licenses/LICENSE-2.0
*
* Unless required by applicable law or agreed to in writing, software
* distributed under the License is distributed on an "AS IS" BASIS,
* WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
* See the License for the specific language governing permissions and
* limitations under the License.
*/
package org.apache.commons.math3.analysis.function;
import java.util.Arrays;
import org.apache.commons.math3.analysis.FunctionUtils;
import org.apache.commons.math3.analysis.UnivariateFunction;
import org.apache.commons.math3.analysis.DifferentiableUnivariateFunction;
import org.apache.commons.math3.analysis.ParametricUnivariateFunction;
import org.apache.commons.math3.analysis.differentiation.DerivativeStructure;
import org.apache.commons.math3.analysis.differentiation.UnivariateDifferentiableFunction;
import org.apache.commons.math3.exception.NullArgumentException;
import org.apache.commons.math3.exception.DimensionMismatchException;
import org.apache.commons.math3.util.FastMath;
/**
* <a href="http://en.wikipedia.org/wiki/Sigmoid_function">
* Sigmoid</a> function.
* It is the inverse of the {@link Logit logit} function.
* A more flexible version, the generalised logistic, is implemented
* by the {@link Logistic} class.
*
* @since 3.0
*/
public class Sigmoid implements UnivariateDifferentiableFunction, DifferentiableUnivariateFunction {
/** Lower asymptote. */
private final double lo;
/** Higher asymptote. */
private final double hi;
/**
* Usual sigmoid function, where the lower asymptote is 0 and the higher
* asymptote is 1.
*/
public Sigmoid() {
this(0, 1);
}
/**
* Sigmoid function.
*
* @param lo Lower asymptote.
* @param hi Higher asymptote.
*/
public Sigmoid(double lo,
double hi) {
this.lo = lo;
this.hi = hi;
}
/** {@inheritDoc}
* @deprecated as of 3.1, replaced by {@link #value(DerivativeStructure)}
*/
@Deprecated
public UnivariateFunction derivative() {
return FunctionUtils.toDifferentiableUnivariateFunction(this).derivative();
}
/** {@inheritDoc} */
public double value(double x) {
return value(x, lo, hi);
}
/**
* Parametric function where the input array contains the parameters of
* the {@link Sigmoid#Sigmoid(double,double) sigmoid function}, ordered
* as follows:
* <ul>
* <li>Lower asymptote</li>
* <li>Higher asymptote</li>
* </ul>
*/
public static class Parametric implements ParametricUnivariateFunction {
/**
* Computes the value of the sigmoid at {@code x}.
*
* @param x Value for which the function must be computed.
* @param param Values of lower asymptote and higher asymptote.
* @return the value of the function.
* @throws NullArgumentException if {@code param} is {@code null}.
* @throws DimensionMismatchException if the size of {@code param} is
* not 2.
*/
public double value(double x, double ... param)
throws NullArgumentException,
DimensionMismatchException {
validateParameters(param);
return Sigmoid.value(x, param[0], param[1]);
}
/**
* Computes the value of the gradient at {@code x}.
* The components of the gradient vector are the partial
* derivatives of the function with respect to each of the
* <em>parameters</em> (lower asymptote and higher asymptote).
*
* @param x Value at which the gradient must be computed.
* @param param Values for lower asymptote and higher asymptote.
* @return the gradient vector at {@code x}.
* @throws NullArgumentException if {@code param} is {@code null}.
* @throws DimensionMismatchException if the size of {@code param} is
* not 2.
*/
public double[] gradient(double x, double ... param)
throws NullArgumentException,
DimensionMismatchException {
validateParameters(param);
final double invExp1 = 1 / (1 + FastMath.exp(-x));
return new double[] { 1 - invExp1, invExp1 };
}
/**
* Validates parameters to ensure they are appropriate for the evaluation of
* the {@link #value(double,double[])} and {@link #gradient(double,double[])}
* methods.
*
* @param param Values for lower and higher asymptotes.
* @throws NullArgumentException if {@code param} is {@code null}.
* @throws DimensionMismatchException if the size of {@code param} is
* not 2.
*/
private void validateParameters(double[] param)
throws NullArgumentException,
DimensionMismatchException {
if (param == null) {
throw new NullArgumentException();
}
if (param.length != 2) {
throw new DimensionMismatchException(param.length, 2);
}
}
}
/**
* @param x Value at which to compute the sigmoid.
* @param lo Lower asymptote.
* @param hi Higher asymptote.
* @return the value of the sigmoid function at {@code x}.
*/
private static double value(double x,
double lo,
double hi) {
return lo + (hi - lo) / (1 + FastMath.exp(-x));
}
/** {@inheritDoc}
* @since 3.1
*/
public DerivativeStructure value(final DerivativeStructure t)
throws DimensionMismatchException {
double[] f = new double[t.getOrder() + 1];
final double exp = FastMath.exp(-t.getValue());
if (Double.isInfinite(exp)) {
// special handling near lower boundary, to avoid NaN
f[0] = lo;
Arrays.fill(f, 1, f.length, 0.0);
} else {
// the nth order derivative of sigmoid has the form:
// dn(sigmoid(x)/dxn = P_n(exp(-x)) / (1+exp(-x))^(n+1)
// where P_n(t) is a degree n polynomial with normalized higher term
// P_0(t) = 1, P_1(t) = t, P_2(t) = t^2 - t, P_3(t) = t^3 - 4 t^2 + t...
// the general recurrence relation for P_n is:
// P_n(x) = n t P_(n-1)(t) - t (1 + t) P_(n-1)'(t)
final double[] p = new double[f.length];
final double inv = 1 / (1 + exp);
double coeff = hi - lo;
for (int n = 0; n < f.length; ++n) {
// update and evaluate polynomial P_n(t)
double v = 0;
p[n] = 1;
for (int k = n; k >= 0; --k) {
v = v * exp + p[k];
if (k > 1) {
p[k - 1] = (n - k + 2) * p[k - 2] - (k - 1) * p[k - 1];
} else {
p[0] = 0;
}
}
coeff *= inv;
f[n] = coeff * v;
}
// fix function value
f[0] += lo;
}
return t.compose(f);
}
}
|
