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
|
/*
* 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._
import org.apache.spark.ml.param._
import org.apache.spark.ml.param.shared._
import org.apache.spark.ml.util._
import org.apache.spark.mllib.feature
import org.apache.spark.mllib.linalg._
import org.apache.spark.sql._
import org.apache.spark.sql.functions._
import org.apache.spark.sql.types.{StructField, StructType}
/**
* Params for [[PCA]] and [[PCAModel]].
*/
private[feature] trait PCAParams extends Params with HasInputCol with HasOutputCol {
/**
* The number of principal components.
* @group param
*/
final val k: IntParam = new IntParam(this, "k", "the number of principal components")
/** @group getParam */
def getK: Int = $(k)
}
/**
* :: Experimental ::
* PCA trains a model to project vectors to a low-dimensional space using PCA.
*/
@Experimental
class PCA (override val uid: String) extends Estimator[PCAModel] with PCAParams
with DefaultParamsWritable {
def this() = this(Identifiable.randomUID("pca"))
/** @group setParam */
def setInputCol(value: String): this.type = set(inputCol, value)
/** @group setParam */
def setOutputCol(value: String): this.type = set(outputCol, value)
/** @group setParam */
def setK(value: Int): this.type = set(k, value)
/**
* Computes a [[PCAModel]] that contains the principal components of the input vectors.
*/
@Since("2.0.0")
override def fit(dataset: Dataset[_]): PCAModel = {
transformSchema(dataset.schema, logging = true)
val input = dataset.select($(inputCol)).rdd.map { case Row(v: Vector) => v}
val pca = new feature.PCA(k = $(k))
val pcaModel = pca.fit(input)
copyValues(new PCAModel(uid, pcaModel.pc, pcaModel.explainedVariance).setParent(this))
}
override def transformSchema(schema: StructType): StructType = {
val inputType = schema($(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($(outputCol), new VectorUDT, false)
StructType(outputFields)
}
override def copy(extra: ParamMap): PCA = defaultCopy(extra)
}
@Since("1.6.0")
object PCA extends DefaultParamsReadable[PCA] {
@Since("1.6.0")
override def load(path: String): PCA = super.load(path)
}
/**
* :: Experimental ::
* Model fitted by [[PCA]].
*
* @param pc A principal components Matrix. Each column is one principal component.
* @param explainedVariance A vector of proportions of variance explained by
* each principal component.
*/
@Experimental
class PCAModel private[ml] (
override val uid: String,
val pc: DenseMatrix,
val explainedVariance: DenseVector)
extends Model[PCAModel] with PCAParams with MLWritable {
import PCAModel._
/** @group setParam */
def setInputCol(value: String): this.type = set(inputCol, value)
/** @group setParam */
def setOutputCol(value: String): this.type = set(outputCol, value)
/**
* Transform a vector by computed Principal Components.
* NOTE: Vectors to be transformed must be the same length
* as the source vectors given to [[PCA.fit()]].
*/
@Since("2.0.0")
override def transform(dataset: Dataset[_]): DataFrame = {
transformSchema(dataset.schema, logging = true)
val pcaModel = new feature.PCAModel($(k), pc, explainedVariance)
val pcaOp = udf { pcaModel.transform _ }
dataset.withColumn($(outputCol), pcaOp(col($(inputCol))))
}
override def transformSchema(schema: StructType): StructType = {
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 copy(extra: ParamMap): PCAModel = {
val copied = new PCAModel(uid, pc, explainedVariance)
copyValues(copied, extra).setParent(parent)
}
@Since("1.6.0")
override def write: MLWriter = new PCAModelWriter(this)
}
@Since("1.6.0")
object PCAModel extends MLReadable[PCAModel] {
private[PCAModel] class PCAModelWriter(instance: PCAModel) extends MLWriter {
private case class Data(pc: DenseMatrix, explainedVariance: DenseVector)
override protected def saveImpl(path: String): Unit = {
DefaultParamsWriter.saveMetadata(instance, path, sc)
val data = Data(instance.pc, instance.explainedVariance)
val dataPath = new Path(path, "data").toString
sqlContext.createDataFrame(Seq(data)).repartition(1).write.parquet(dataPath)
}
}
private class PCAModelReader extends MLReader[PCAModel] {
private val className = classOf[PCAModel].getName
/**
* Loads a [[PCAModel]] from data located at the input path. Note that the model includes an
* `explainedVariance` member that is not recorded by Spark 1.6 and earlier. A model
* can be loaded from such older data but will have an empty vector for
* `explainedVariance`.
*
* @param path path to serialized model data
* @return a [[PCAModel]]
*/
override def load(path: String): PCAModel = {
val metadata = DefaultParamsReader.loadMetadata(path, sc, className)
// explainedVariance field is not present in Spark <= 1.6
val versionRegex = "([0-9]+)\\.([0-9]+).*".r
val hasExplainedVariance = metadata.sparkVersion match {
case versionRegex(major, minor) =>
(major.toInt >= 2 || (major.toInt == 1 && minor.toInt > 6))
case _ => false
}
val dataPath = new Path(path, "data").toString
val model = if (hasExplainedVariance) {
val Row(pc: DenseMatrix, explainedVariance: DenseVector) =
sqlContext.read.parquet(dataPath)
.select("pc", "explainedVariance")
.head()
new PCAModel(metadata.uid, pc, explainedVariance)
} else {
val Row(pc: DenseMatrix) = sqlContext.read.parquet(dataPath).select("pc").head()
new PCAModel(metadata.uid, pc, Vectors.dense(Array.empty[Double]).asInstanceOf[DenseVector])
}
DefaultParamsReader.getAndSetParams(model, metadata)
model
}
}
@Since("1.6.0")
override def read: MLReader[PCAModel] = new PCAModelReader
@Since("1.6.0")
override def load(path: String): PCAModel = super.load(path)
}
|
