Github user staple commented on a diff in the pull request:
https://github.com/apache/spark/pull/4934#discussion_r25986928
--- Diff:
mllib/src/main/scala/org/apache/spark/mllib/optimization/AcceleratedGradientDescent.scala
---
@@ -0,0 +1,237 @@
+/*
+ * 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.mllib.optimization
+
+import scala.collection.mutable.ArrayBuffer
+
+import breeze.linalg.{DenseVector => BDV, norm}
+
+import org.apache.spark.Logging
+import org.apache.spark.annotation.DeveloperApi
+import org.apache.spark.mllib.linalg.{Vector, Vectors}
+import org.apache.spark.rdd.RDD
+
+/**
+ * :: DeveloperApi ::
+ * This class optimizes a vector of weights via accelerated (proximal)
gradient descent.
+ * The implementation is based on TFOCS [[http://cvxr.com/tfocs]],
described in Becker, Candes, and
+ * Grant 2010.
+ * @param gradient Delegate that computes the loss function value and
gradient for a vector of
+ * weights.
+ * @param updater Delegate that updates weights in the direction of a
gradient.
+ */
+@DeveloperApi
+class AcceleratedGradientDescent (private var gradient: Gradient, private
var updater: Updater)
+ extends Optimizer {
+
+ private var stepSize: Double = 1.0
+ private var convergenceTol: Double = 1e-4
+ private var numIterations: Int = 100
+ private var regParam: Double = 0.0
+
+ /**
+ * Set the initial step size, used for the first step. Default 1.0.
+ * On subsequent steps, the step size will be adjusted by the
acceleration algorithm.
+ */
+ def setStepSize(step: Double): this.type = {
--- End diff --
@mengxr Thanks for taking a look. I was advised by Reza Zadeh to implement
a version without line search, at least for the initial implementation.
Please see discussion here:
https://issues.apache.org/jira/browse/SPARK-1503?focusedCommentId=14225295, and
in the following comments. I also attached some optimization benchmarks to the
jira, which include performance of both backtracking line search and non line
search implementations. Per your suggestion that it's hard to choose a proper
stepSize I can attest that, anecdotally, acceleration seems somewhat more
sensitive to diverging with nominal stepSize than the existing gradient descent.
---
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.
---
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
