-
-
Notifications
You must be signed in to change notification settings - Fork 182
/
MinMaxNormalizer.php
247 lines (205 loc) · 5.8 KB
/
MinMaxNormalizer.php
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
<?php
namespace Rubix\ML\Transformers;
use Rubix\ML\DataType;
use Rubix\ML\Persistable;
use Rubix\ML\Datasets\Dataset;
use Rubix\ML\Traits\AutotrackRevisions;
use Rubix\ML\Specifications\SamplesAreCompatibleWithTransformer;
use Rubix\ML\Exceptions\InvalidArgumentException;
use Rubix\ML\Exceptions\RuntimeException;
use function min;
use function max;
use const Rubix\ML\EPSILON;
/**
* Min Max Normalizer
*
* The *Min Max* Normalizer scales the input features to a value between
* a user-specified range (default 0 to 1).
*
* @category Machine Learning
* @package Rubix/ML
* @author Andrew DalPino
*/
class MinMaxNormalizer implements Transformer, Stateful, Elastic, Reversible, Persistable
{
use AutotrackRevisions;
/**
* The minimum value of the transformed features.
*
* @var float
*/
protected float $min;
/**
* The maximum value of the transformed features.
*
* @var float
*/
protected float $max;
/**
* The computed minimums of the fitted data.
*
* @var (int|float)[]|null
*/
protected ?array $minimums = null;
/**
* The computed maximums of the fitted data.
*
* @var (int|float)[]|null
*/
protected ?array $maximums = null;
/**
* The scale coefficients of each feature.
*
* @var float[]|null
*/
protected ?array $scales = null;
/**
* @param float $min
* @param float $max
* @throws InvalidArgumentException
*/
public function __construct(float $min = 0.0, float $max = 1.0)
{
if ($min > $max) {
throw new InvalidArgumentException('Minimum cannot be greater'
. ' than maximum.');
}
$this->min = $min;
$this->max = $max;
}
/**
* Return the data types that this transformer is compatible with.
*
* @internal
*
* @return list<\Rubix\ML\DataType>
*/
public function compatibility() : array
{
return DataType::all();
}
/**
* Is the transformer fitted?
*
* @return bool
*/
public function fitted() : bool
{
return $this->minimums and $this->maximums;
}
/**
* Return the minmums of each feature column.
*
* @return (int|float)[]|null
*/
public function minimums() : ?array
{
return $this->minimums;
}
/**
* Return the maximums of each feature column.
*
* @return (int|float)[]|null
*/
public function maximums() : ?array
{
return $this->maximums;
}
/**
* Fit the transformer to a dataset.
*
* @param Dataset $dataset
*/
public function fit(Dataset $dataset) : void
{
SamplesAreCompatibleWithTransformer::with($dataset, $this)->check();
$this->minimums = $this->maximums = $this->scales = [];
foreach ($dataset->featureTypes() as $column => $type) {
if ($type->isContinuous()) {
$values = $dataset->feature($column);
/** @var int|float $min */
$min = min($values);
/** @var int|float $max */
$max = max($values);
$scale = ($this->max - $this->min) / (($max - $min) ?: EPSILON);
$this->minimums[$column] = $min;
$this->maximums[$column] = $max;
$this->scales[$column] = $scale;
}
}
}
/**
* Update the fitting of the transformer.
*
* @param Dataset $dataset
*/
public function update(Dataset $dataset) : void
{
if (!isset($this->minimums, $this->maximums, $this->scales)) {
$this->fit($dataset);
return;
}
SamplesAreCompatibleWithTransformer::with($dataset, $this)->check();
foreach ($this->scales as $column => &$scale) {
$values = $dataset->feature($column);
/** @var int|float $min */
$min = min($this->minimums[$column], ...$values);
/** @var int|float $max */
$max = max($this->maximums[$column], ...$values);
$scale = ($this->max - $this->min) / (($max - $min) ?: EPSILON);
$this->minimums[$column] = $min;
$this->maximums[$column] = $max;
}
}
/**
* Transform the dataset in place.
*
* @param list<list<mixed>> $samples
* @throws RuntimeException
*/
public function transform(array &$samples) : void
{
if (!isset($this->minimums, $this->scales)) {
throw new RuntimeException('Transformer has not been fitted.');
}
foreach ($samples as &$sample) {
foreach ($this->scales as $column => $scale) {
$value = &$sample[$column];
$min = $this->minimums[$column];
$value *= $scale;
$value += $this->min - $min * $scale;
}
}
}
/**
* Perform the reverse transformation to the samples.
*
* @param list<list<mixed>> $samples
* @throws RuntimeException
*/
public function reverseTransform(array &$samples) : void
{
if (!isset($this->minimums, $this->scales)) {
throw new RuntimeException('Transformer has not been fitted.');
}
foreach ($samples as &$sample) {
foreach ($this->scales as $column => $scale) {
$value = &$sample[$column];
$min = $this->minimums[$column];
$value -= $this->min - $min * $scale;
$value /= $scale;
}
}
}
/**
* Return the string representation of the object.
*
* @internal
*
* @return string
*/
public function __toString() : string
{
return "Min Max Normalizer (min: {$this->min}, max: {$this->max})";
}
}