flink-issues mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From chiwanpark <...@git.apache.org>
Subject [GitHub] flink pull request: [FLINK-1745] Add exact k-nearest-neighbours al...
Date Thu, 08 Oct 2015 15:18:34 GMT
Github user chiwanpark commented on a diff in the pull request:

    https://github.com/apache/flink/pull/1220#discussion_r41526734
  
    --- Diff: flink-staging/flink-ml/src/main/scala/org/apache/flink/ml/nn/QuadTree.scala
---
    @@ -0,0 +1,305 @@
    +
    +/*
    + * 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.nn.util
    +
    +import org.apache.flink.ml.math.Vector
    +import org.apache.flink.ml.metrics.distances.DistanceMetric
    +
    +import scala.collection.mutable.ListBuffer
    +import scala.collection.mutable.PriorityQueue
    +
    +/**
    + * n-dimensional QuadTree data structure; partitions
    + * spatial data for faster queries (e.g. KNN query)
    + * The skeleton of the data structure was initially
    + * based off of the 2D Quadtree found here:
    + * http://www.cs.trinity.edu/~mlewis/CSCI1321-F11/Code/src/util/Quadtree.scala
    + *
    + * Many additional methods were added to the class both for
    + * efficient KNN queries and generalizing to n-dim.
    + *
    + * @param minVec
    + * @param maxVec
    + */
    +class QuadTree(minVec:ListBuffer[Double], maxVec:ListBuffer[Double],distMetric:DistanceMetric){
    +  var maxPerBox = 20
    +
    +  class Node(c:ListBuffer[Double],L:ListBuffer[Double], var children:ListBuffer[Node])
{
    +
    +    var objects = new ListBuffer[Vector]
    +
    +    /** for testing purposes only; used in QuadTreeSuite.scala
    +      *
    +      * @return
    +      */
    +    def getCenterLength(): (ListBuffer[Double], ListBuffer[Double]) = {
    +      (c, L)
    +    }
    +
    +    def contains(obj: Vector): Boolean = {
    +      overlap(obj, 0.0)
    +    }
    +
    +    /** Tests if obj is within a radius of the node
    +      *
    +      * @param obj
    +      * @param radius
    +      * @return
    +      */
    +    def overlap(obj: Vector, radius: Double): Boolean = {
    +      var count = 0
    +      for (i <- 0 to obj.size - 1) {
    +        if (obj(i) - radius < c(i) + L(i) / 2 && obj(i) + radius > c(i)
- L(i) / 2) {
    +          count += 1
    +        }
    +      }
    +
    +      if (count == obj.size) {
    +        return true
    +      } else {
    +        return false
    +      }
    +    }
    +
    +    /** Tests if obj is near a node:  minDist is defined so that every point in the box
    +      * has distance to obj greater than minDist
    +      * (minDist adopted from "Nearest Neighbors Queries" by N. Roussopoulos et al.)
    +      *
    +      * @param obj
    +      * @param radius
    +      * @return
    +      */
    +    def isNear(obj: Vector, radius: Double): Boolean = {
    +      if (minDist(obj) < radius) {
    +        true
    +      } else {
    +        false
    +      }
    +    }
    +
    +    def minDist(obj: Vector): Double = {
    +      var minDist = 0.0
    +      for (i <- 0 to obj.size - 1) {
    +        if (obj(i) < c(i) - L(i) / 2) {
    +          minDist += math.pow(obj(i) - c(i) + L(i) / 2, 2)
    +        } else if (obj(i) > c(i) + L(i) / 2) {
    +          minDist += math.pow(obj(i) - c(i) - L(i) / 2, 2)
    +        }
    +      }
    +      return minDist
    +    }
    +
    +    def whichChild(obj:Vector):Int = {
    +
    +      var count = 0
    +      for (i <- 0 to obj.size - 1){
    +        if (obj(i) > c(i)) {
    +          count += Math.pow(2,i).toInt
    +        }
    +      }
    +      count
    +    }
    +
    +    def makeChildren() {
    +      var cBuff = new ListBuffer[ListBuffer[Double]]
    +      cBuff += c
    +      var Childrennodes = new ListBuffer[Node]
    +      val cPart = partitionBox(cBuff,L,L.length)
    +      for (i <- cPart.indices){
    +        Childrennodes = Childrennodes :+ new Node(cPart(i), L.map(x => x/2.0), null)
    +
    +      }
    --- End diff --
    
    Maybe we can add a method such as `multiply(value: Double)` in `Vector` trait. But currently,
there is no easy method for this situation. I think that following is best method currently:
    
    ```scala
    val mappedWidth = width match {
      case SparseVector(size, indices, data) =>
        val newData = data.map(_ / 2.0)
        SparseVector(size, indices, data)
      case DenseVector(data) =>
        val newData = data.map(_ / 2.0)
        DenseVector(data)
    }


---
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 infrastructure@apache.org or file a JIRA ticket
with INFRA.
---

Mime
View raw message