Github user rawkintrevo commented on a diff in the pull request:
https://github.com/apache/flink/pull/1898#discussion_r61582494
--- Diff:
flink-libraries/flink-ml/src/main/scala/org/apache/flink/ml/preprocessing/Splitter.scala
---
@@ -0,0 +1,215 @@
+/*
+ * 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.flink.ml.preprocessing
+
+import org.apache.flink.api.common.typeinfo.{TypeInformation,
BasicTypeInfo}
+import org.apache.flink.api.java.Utils
+import org.apache.flink.api.scala. DataSet
+import org.apache.flink.api.scala.utils._
+
+import org.apache.flink.ml.common.{FlinkMLTools, ParameterMap,
WithParameters}
+import _root_.scala.reflect.ClassTag
+
+object Splitter {
+
+ case class TrainTestDataSet[T: TypeInformation : ClassTag](training:
DataSet[T],
+ testing:
DataSet[T])
+
+ case class TrainTestHoldoutDataSet[T: TypeInformation :
ClassTag](training: DataSet[T],
+
testing: DataSet[T],
+
holdout: DataSet[T])
+ //
--------------------------------------------------------------------------------------------
+ // randomSplit
+ //
--------------------------------------------------------------------------------------------
+ /**
+ * Split a DataSet by the probability fraction of each element.
+ *
+ * @param input DataSet to be split
+ * @param fraction Probability that each element is chosen,
should be [0,1] without
+ * replacement, and [0, â) with replacement.
While fraction is larger
+ * than 1, the elements are expected to be
selected multi times into
+ * sample on average. This fraction refers to the
first element in the
+ * resulting array.
+ * @param precise Sampling by default is random and can result
in slightly lop-sided
+ * sample sets. When precise is true, equal
sample set size are forced,
+ * however this is somewhat less efficient.
+ * @param seed Random number generator seed.
+ * @return An array of two datasets
+ */
+
+ def randomSplit[T: TypeInformation : ClassTag]( input: DataSet[T],
+ fraction: Double,
+ precise: Boolean = false,
+ seed: Long =
Utils.RNG.nextLong())
+ : Array[DataSet[T]] = {
+ import org.apache.flink.api.scala._
+
+ val indexedInput: DataSet[(Long, T)] = input.zipWithIndex
+
+ val leftSplit: DataSet[(Long, T)] = precise match {
+ case false => indexedInput.sample(false, fraction, seed)
--- End diff --
I think boostrapping would be a cool feature- but would require a different
approach than the joins on the leftSplit/rightSplit.
If you over sample the leftSplit, there's not going to be anything left to
put in the right split (the whole points was to keep the training and testing
cases seperate).
I'm going to to add a boostrap method that will allow for oversampling in
the testing and training cases. Re: the next comment, I will test is
separately.
---
If your project is set up for it, you can reply to this email and have your
reply appear on GitHub as well. If your project does not have this feature
enabled and wishes so, or if the feature is enabled but not working, please
contact infrastructure at [email protected] or file a JIRA ticket
with INFRA.
---
