Github user jkbradley commented on a diff in the pull request:
https://github.com/apache/spark/pull/3637#discussion_r21559541
--- Diff: mllib/src/main/scala/org/apache/spark/ml/LabeledPoint.scala ---
@@ -0,0 +1,52 @@
+/*
+ * 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
+
+import scala.beans.BeanInfo
+
+import org.apache.spark.annotation.AlphaComponent
+import org.apache.spark.mllib.linalg.Vector
+
+/**
+ * :: AlphaComponent ::
+ * Class that represents an instance (data point) for prediction tasks.
+ *
+ * @param label Label to predict
+ * @param features List of features describing this instance
+ * @param weight Instance weight
+ */
+@AlphaComponent
+@BeanInfo
+case class LabeledPoint(label: Double, features: Vector, weight: Double) {
--- End diff --
The main pros & cons I see for having continuous & categorical types for
labels & features are:
Pros:
* Type safety.
Cons:
* Algorithms may need to make copies of the data, depending on how much we
expose internals of a features type.
* Users may have to worry more about types.
* For labels, if we load data from a file without metadata (like libsvm),
we may need to assume that everything is continuous. Users will have to
explicitly cast labels to categorical for classification.
* For features, strong typing implies a stronger contract, where the
assumption is that users specify the correct types. I've been wondering about
having more "best effort" APIs, where we take suggestions from users (like
DecisionTree's categoricalFeaturesInfo) but otherwise try to infer the best
types to use under the hood.
These lists started out much more balanced, but I guess I'm voting for the
old system where everything is a Double.
---
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 infrastruct...@apache.org or file a JIRA ticket
with INFRA.
---
---------------------------------------------------------------------
To unsubscribe, e-mail: reviews-unsubscr...@spark.apache.org
For additional commands, e-mail: reviews-h...@spark.apache.org
