数据结构代写 k-d tree CMSC 420: Data Structures

Writeup for fifth project of CMSC 420: “Data Structures” Section 0201, Fall 2018

Theme: KD-Trees & PR-QuadTrees Handout date: Sunday, 11-18-2018

On-time deadline: 12-12-2018, 11:59pm Late deadline (30% penalty): 12-14-2018, 11:59pm

Contents
1 Overview 2

  1. 2  Getting started 2
  2. 3  What you need to implement 2

4 Codebase 2 4.1 Top level . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2 4.2 KD-TreeandBoundedPriorityQueue …………………… 4 4.3 PR-QuadTree . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5

1

1 Overview

In this project you will implement KD-Trees and P-R (Point-Region) QuadTrees. For the former, you will also implement spatial queries (range, nearest neighbor). You will be tested against unit tests hosted on the department’s submit server.

Half of this project is the study of spatial data structures and half the practice of Object-Oriented Programming primitives, in particular, Inheritance and Polymorphism. You will need to spend some time studying the provided documentation and code structure, in order to understand how the various components are pieced together.

2 Getting started

All you need to do to get started is run a git pull from your working Git directory. This will update your project files with a subpackage called projects.spatial, which contains all the code we provide you with as well as the classes that you need to implement.

Speaking of…

3 What you need to implement

Everything you need to get started is available in our common Git repository. You will need to fill in the implementation of the following 4 classes:

• knnutils.BoundedPriorityQueue • nodes.KDTreeNode
• nodes.PRQuadBlackNode
• nodes.PRQuadGrayNode

The classes come with sufficient documentation that you will be able to find under the directory doc, so that you can have a full view of the functionality exposed by the class’ public interface. You are given the skeletons of the above classes, as well as various other classes and interfaces, further described in Section 4, which follows.

4 Code base 4.1 Top level

This project can essentially be divided into two “mini-projects”. It doesn’t matter which you will implement first and which one you will implement second, so we will just arbitrarily decide to call the KD-Tree part of the project as the first part, and the PR-QuadTree part of the project as its second part.

In this project, we supply you with a lot of code to use to build your own. Figure 1 provides a bird’s eye-view of the project.

2

SpatialDictionary
SpatialQuerySolver

+insert() +delete() +search() +height() + isEmpty() + count()

KDTree

+range() +nearestNeighbor() +kNearestNeighbors

-dims
-root -count +INFTY +DEFAULT_DIMS +KDTree(int) +getRoot()

KDPoint

Figure 1: A UML diagram describing the behavior and basic dependencies of the classes KDTree and PRQuadTree. Simple lines reflect one-to-many (1 – ∗) “has – a” relations, while arrows show “is-a” relations (derived class, implemented interface, etc).

Both KD-Trees and PR-QuadTrees are multi-dimensional indices. Since they are multi- dimensional indices, the first thing they need to know is the nature of the keys that they will store. The type of key stored is defined in kdpoint.KDPoint. Instances of this class will appear in virtually all of the methods that you will have to implement! You should study KDPoint extensively to understand how it works. In particular, notice that, since the internal buffer of KDPoints is exposed to the outside world for convenience, KDPoints are mutable objects! This means that you should always make deep copies of KDPoint instances when you have to! No aliasing! A copy constructor for KDPoints is provided for you.

One additional point about the implementation of KDPoints: as you can see by browsing ei- ther the documentation or the source code of the class KDPoint, the data type used to hold the coordinates of a d-dimensional point is double. This can cause some problems with com- paring doubles up to an arbitrary degree of precision.1 To have an idea of how you can alle- viate this problem of comparing doubles in your unit tests, please feel free to consult the file kdpoint.KDPointTests.java, which is a jUnit4 test suite for KDPoints that we include just for your own benefit.

KDTree implements both SpatialDictionary and SpatialQuerySolver, while PRQuadTree

1To alleviate these problems, ava has introduced the type BigDecimal, which is an immutable floating-point number type that allows for arbitrary precision.

PRQuadTree

-maxPoints -root
-n
-count +PRQuadTree(int, int)

-coords +KDPoint(int) +KDPoint(double…) +KDPoint(KDPoint) +distance

3

only implements SpatialDictionary. This has been deliberate, so that we can make the project easier for you. However, it also shows you one of the benefits of Object-Oriented Design. If you are ever interested in extending the functionality of PRQuadTree to allow for spatial queries (range, 1 or m nearest neighbors), all you need to do is extend that interface (in addition to SpatialDictonary). Your IDE will then prompt you for an implementation of those methods, with their exact signatures, and even copy over the JavaDocs so that you don’t need to re- write them! This is an example of how you should be thinking later on in your careers, if you end up having to do a lot of object-oriented software development: if you see that you have many classes that share a good portion of a common public interface, then maybe you should package all of their declarations into an actual Java interface, instead of copying – and – pasting declarations of methods! Those methods might be in the hundreds! Not to mention that if the actual implementation is entirely identical, then the interface can be made into an abstract class, and the same implementation can be re-used! For example, methods such as height(), count(), isEmpty() tend to be implemented in exactly the same way line-by- line in many tree-based data structures. Perhaps if we had a common interface for all those structures called Tree, we would be able to include all the common information there and not waste time on re-writing stub methods!

The less code you have to write for any given task, the better. For one, you reduce the likelihood of inserting bugs in the code base. You also avoid going through the trouble of having to generate new documentation, which you can also introduce bugs in! Finally, you also make compilation faster: you avoid compiling and linking new sources with your existing ones.

4.2 KD-Tree and Bounded Priority Queue

Figure 2 shows the structure of the first part of the project:

1 …

*

java.lang.Iterable<T>

+forEach

+iterator() +spliterator()

+insert() +delete()
+…. +DEFAULT_DIMS… +INFTY
-root
-dims -count

BoundedPriorityQueue<T>

+BoundedPriorityQueue(int) +enqueue()

+dequeue() +first() +last() +size() +isEmpty() -UNIMPL_METHOD

KDTreeNode

+insert() +delete()

+….
+range +kNearestNeighbors +… +getPoint() -UNIMPL_METHOD

1…1

1…1

KDTree

NNData<T>

+bestGuess

+bestDist +NNData(T, double)

Figure 2: Structure of the first part of the project, dedicated to the implementation of KD-Trees and “plug-ins” that allow for spatial queries.

Interestingly, the implementation of the class KDTree is provided for you! However, this is not much respite: If you were to browse the class’ source code, you would note that all the work

4

you have to do is now part of the class KDTreeNode, located inside the package nodes. Section 4.3 will shed some light on why we structure the code you have to submit in this manner.

KDTreeNode uses BoundedPriorityQueue only for the implementation of m-nearest neighbor queries, with m ≥ 2. For 1-nearest neighbor, it uses the simple struct – like class NNData. These types are declared as parameters of the relevant methods of the class KDTree and KDTreeNode, so your project won’t compile against our tests if you don’t have them exactly where they are in the code tree!

For your implementation of BoundedPriorityQueue, you are given complete implementa- tional freedom. That is to say, if you wanted to use your Priority Queue from project 1 and adjust it to the semantics of a priority queue bounded above, you can do this. If you want to extend Java’s built-in Priority Queue to do what you want it to, that is also fine with us. You do not even have to adhere to a particular implementation of a Priority Queue: you can use a min-heap, an array of lists made up with elements of the same priority, sorted by insertion order, anything really, as long as the implementation is correct and provides elementary efficiency for first(), dequeue(), enqueue(), last(). By “elementary efficiency”, we essentially mean: “Don’t make it so bad that the submit server will hang.” Doing this would be a major achievement in its own right, and we trust that we don’t need to define it any further. However, the semantics of BoundedPriorityQueue are different from that of a classic Priority Queue in that it limits the number of “best neighbors” that a given KDPoint can have. Consult the lecture slides and recordings for more information on what we mean by that.

Something important regarding the first part of your project is that we will be testing to make sure you don’t solve the nearest neighbor problem naively. That is to say, while you do have implementational freedom for BoundedPriorityQueue, the implementation of the nearest neighbor queries for KDTree cannot simply consist of a depth-first collection of all the nodes, their sorting based on the distance from the “anchor” point and the returning of the k closest neighbors! We will be manually inspecting your code to make sure that you are not implementing KNN in this naive manner! You must use an instance of BoundedPriorityQueue to implement the nearest neighbor methods!

4.3 PR-QuadTree

Figure 3 contains a UML diagram that shows the structure of the second part of the project, which concerns the data structure known as a “Point-Region” (P-R) QuadTree .

As with the class KD-Tree, the “central” class PRQuadTree has been given to you! However, all the work that you need to do is in the derived classes PRQuadGrayNode and PRQuadBlackNode. Since any node in the PR-QuadTree can dynamically change status between black, gray and white node, we need to deal with those nodes polymorphically: i.e, we want to call insertion and deletion routines that do the right thing given the runtime class of the PRQuadNode instance for which they are called! For this reason, we provide you with the abstract class PRQuadNode, which provides the common interface that we expect from any given node in a PR-QuadTree. In detail, every node should have some way to insert and delete a KDPoint, query the node about the height of the subtree rooted at the node, ask for the number of KDPoints stored either in the node itself (if it’s black) or anywhere in the (sub)tree rooted at the node (if it’s gray).

5

PRQuadTree

-root -maxPoints
-n
-count +PRQuadTree(int, int)

KDPoint

-coords +KDPoint(int) +KDPoint(double…) +KDPoint(KDPoint) +distance

1 …

PRQuadNode

*

#PRQuadNode(KDPoint, int, int) +insert()
+delete()
+….
#centroid
#k #bucketingParam

* …

1

PRQuadBlackNode

Figure 3: Structure of the second part of the project, dedicated to the polymorphic implementation of a PR-QuadTree. Note that PRQuadGrayNode instances are both PRQuadNode instances and contain up to 4 PRQuadNode instances!

Some points of interest:

  • You might wonder why there is no class PRQuadWhiteNode in the provided code base. This is because such a class would be fundamentally useless, since white nodes don’t really do or contain anything! This means that one can model them adequately (and cheaply!) using a null reference. As always, you yourselves absolutely can use a separate class called, e.g, PRQuadWhiteNode for your own purposes (debugging or otherwise), if you want to. Remember: you should not alter the existing code base in any way, but adding your own functionality is always fine. The submit server unit test suites only care about what they can call, and what they can call is your public methods!
  • The base class PRQuadNode is made into an abstract class instead of an interface be- cause it contains a protected data field of type KDPoint, which is called centroid, and another int field called k. Refer to their documentation2 to understand what they are useful for. In the Java programming language, one cannot have data fields in interfaces. Every one of our nodes, irrespective of color, has a geometrical interpretation, since it models a certain quadrant of our space. As you think about the project, you might wonder why we have centroid and k as data types in the base class, instead of just in PRQuadGrayNode. This you might think because, at least initially, it doesn’t seem as if a PRQuadBlackNode actually uses any geometrical information; it just stores KDPoints. We will not answer this question in the writeup; it is for you to think about. 2Yes, you can generate JavaDocs for fields too, and even private or protected fields or methods. All modern IDEs have ways to toggle whether documentation of such “access-restricted” elements of a given class will appear in the documentation. We provide you with JavaDocs for everything in this project.

PRQuadGrayNode

-UNIMPL_METHOD

+DEFAULT_BUCKETSIZE=1

+PRQuadBlackNode(KDPoint, int, int) +PRQuadBlackNode(KDPoint, int, int, KDPoint)

-UNIMPL_METHOD +PRQuadGrayNode(KDPoint, int, int)

6

4 …

1