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
|
/*
* 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.spark.ml.feature
import org.apache.hadoop.fs.Path
import org.apache.spark.annotation.{Experimental, Since}
import org.apache.spark.ml.{Estimator, Model}
import org.apache.spark.ml.param.{DoubleParam, ParamMap, Params}
import org.apache.spark.ml.param.shared.{HasInputCol, HasOutputCol}
import org.apache.spark.ml.util._
import org.apache.spark.mllib.linalg.{Vector, Vectors, VectorUDT}
import org.apache.spark.mllib.stat.Statistics
import org.apache.spark.sql._
import org.apache.spark.sql.functions._
import org.apache.spark.sql.types.{StructField, StructType}
/**
* Params for [[MinMaxScaler]] and [[MinMaxScalerModel]].
*/
private[feature] trait MinMaxScalerParams extends Params with HasInputCol with HasOutputCol {
/**
* lower bound after transformation, shared by all features
* Default: 0.0
* @group param
*/
val min: DoubleParam = new DoubleParam(this, "min",
"lower bound of the output feature range")
/** @group getParam */
def getMin: Double = $(min)
/**
* upper bound after transformation, shared by all features
* Default: 1.0
* @group param
*/
val max: DoubleParam = new DoubleParam(this, "max",
"upper bound of the output feature range")
/** @group getParam */
def getMax: Double = $(max)
/** Validates and transforms the input schema. */
protected def validateAndTransformSchema(schema: StructType): StructType = {
validateParams()
val inputType = schema($(inputCol)).dataType
require(inputType.isInstanceOf[VectorUDT],
s"Input column ${$(inputCol)} must be a vector column")
require(!schema.fieldNames.contains($(outputCol)),
s"Output column ${$(outputCol)} already exists.")
val outputFields = schema.fields :+ StructField($(outputCol), new VectorUDT, false)
StructType(outputFields)
}
override def validateParams(): Unit = {
require($(min) < $(max), s"The specified min(${$(min)}) is larger or equal to max(${$(max)})")
}
}
/**
* :: Experimental ::
* Rescale each feature individually to a common range [min, max] linearly using column summary
* statistics, which is also known as min-max normalization or Rescaling. The rescaled value for
* feature E is calculated as,
*
* Rescaled(e_i) = \frac{e_i - E_{min}}{E_{max} - E_{min}} * (max - min) + min
*
* For the case E_{max} == E_{min}, Rescaled(e_i) = 0.5 * (max + min)
* Note that since zero values will probably be transformed to non-zero values, output of the
* transformer will be DenseVector even for sparse input.
*/
@Experimental
class MinMaxScaler(override val uid: String)
extends Estimator[MinMaxScalerModel] with MinMaxScalerParams with DefaultParamsWritable {
def this() = this(Identifiable.randomUID("minMaxScal"))
setDefault(min -> 0.0, max -> 1.0)
/** @group setParam */
def setInputCol(value: String): this.type = set(inputCol, value)
/** @group setParam */
def setOutputCol(value: String): this.type = set(outputCol, value)
/** @group setParam */
def setMin(value: Double): this.type = set(min, value)
/** @group setParam */
def setMax(value: Double): this.type = set(max, value)
override def fit(dataset: DataFrame): MinMaxScalerModel = {
transformSchema(dataset.schema, logging = true)
val input = dataset.select($(inputCol)).rdd.map { case Row(v: Vector) => v }
val summary = Statistics.colStats(input)
copyValues(new MinMaxScalerModel(uid, summary.min, summary.max).setParent(this))
}
override def transformSchema(schema: StructType): StructType = {
validateAndTransformSchema(schema)
}
override def copy(extra: ParamMap): MinMaxScaler = defaultCopy(extra)
}
@Since("1.6.0")
object MinMaxScaler extends DefaultParamsReadable[MinMaxScaler] {
@Since("1.6.0")
override def load(path: String): MinMaxScaler = super.load(path)
}
/**
* :: Experimental ::
* Model fitted by [[MinMaxScaler]].
*
* @param originalMin min value for each original column during fitting
* @param originalMax max value for each original column during fitting
*
* TODO: The transformer does not yet set the metadata in the output column (SPARK-8529).
*/
@Experimental
class MinMaxScalerModel private[ml] (
override val uid: String,
val originalMin: Vector,
val originalMax: Vector)
extends Model[MinMaxScalerModel] with MinMaxScalerParams with MLWritable {
import MinMaxScalerModel._
/** @group setParam */
def setInputCol(value: String): this.type = set(inputCol, value)
/** @group setParam */
def setOutputCol(value: String): this.type = set(outputCol, value)
/** @group setParam */
def setMin(value: Double): this.type = set(min, value)
/** @group setParam */
def setMax(value: Double): this.type = set(max, value)
override def transform(dataset: DataFrame): DataFrame = {
val originalRange = (originalMax.toBreeze - originalMin.toBreeze).toArray
val minArray = originalMin.toArray
val reScale = udf { (vector: Vector) =>
val scale = $(max) - $(min)
// 0 in sparse vector will probably be rescaled to non-zero
val values = vector.toArray
val size = values.size
var i = 0
while (i < size) {
val raw = if (originalRange(i) != 0) (values(i) - minArray(i)) / originalRange(i) else 0.5
values(i) = raw * scale + $(min)
i += 1
}
Vectors.dense(values)
}
dataset.withColumn($(outputCol), reScale(col($(inputCol))))
}
override def transformSchema(schema: StructType): StructType = {
validateAndTransformSchema(schema)
}
override def copy(extra: ParamMap): MinMaxScalerModel = {
val copied = new MinMaxScalerModel(uid, originalMin, originalMax)
copyValues(copied, extra).setParent(parent)
}
@Since("1.6.0")
override def write: MLWriter = new MinMaxScalerModelWriter(this)
}
@Since("1.6.0")
object MinMaxScalerModel extends MLReadable[MinMaxScalerModel] {
private[MinMaxScalerModel]
class MinMaxScalerModelWriter(instance: MinMaxScalerModel) extends MLWriter {
private case class Data(originalMin: Vector, originalMax: Vector)
override protected def saveImpl(path: String): Unit = {
DefaultParamsWriter.saveMetadata(instance, path, sc)
val data = new Data(instance.originalMin, instance.originalMax)
val dataPath = new Path(path, "data").toString
sqlContext.createDataFrame(Seq(data)).repartition(1).write.parquet(dataPath)
}
}
private class MinMaxScalerModelReader extends MLReader[MinMaxScalerModel] {
private val className = classOf[MinMaxScalerModel].getName
override def load(path: String): MinMaxScalerModel = {
val metadata = DefaultParamsReader.loadMetadata(path, sc, className)
val dataPath = new Path(path, "data").toString
val Row(originalMin: Vector, originalMax: Vector) = sqlContext.read.parquet(dataPath)
.select("originalMin", "originalMax")
.head()
val model = new MinMaxScalerModel(metadata.uid, originalMin, originalMax)
DefaultParamsReader.getAndSetParams(model, metadata)
model
}
}
@Since("1.6.0")
override def read: MLReader[MinMaxScalerModel] = new MinMaxScalerModelReader
@Since("1.6.0")
override def load(path: String): MinMaxScalerModel = super.load(path)
}
|