CPSC 411, Fall 2008: Set 7 1 CPSC 411 Design and Analysis of Algorithms Set 7: Disjoint Sets Prof. Jennifer Welch Fall 2008.

Slides:



Advertisements
Similar presentations
Lecture 15. Graph Algorithms
Advertisements

1 Union-find. 2 Maintain a collection of disjoint sets under the following two operations S 3 = Union(S 1,S 2 ) Find(x) : returns the set containing x.
MA/CSSE 473 Day 38 Finish Disjoint Sets Dijkstra.
1 Disjoint Sets Set = a collection of (distinguishable) elements Two sets are disjoint if they have no common elements Disjoint-set data structure: –maintains.
1 Introduction to Algorithms 6.046J/18.401J/SMA5503 Lecture 20 Prof. Erik Demaine.
© 2004 Goodrich, Tamassia Union-Find1 Union-Find Partition Structures.
Andreas Klappenecker [Based on slides by Prof. Welch]
Disjoint-Set Operation
Discussion #36 Spanning Trees
CSE 780 Algorithms Advanced Algorithms Minimum spanning tree Generic algorithm Kruskal’s algorithm Prim’s algorithm.
Minimum Spanning Trees Definition Algorithms –Prim –Kruskal Proofs of correctness.
CPSC 311, Fall CPSC 311 Analysis of Algorithms Disjoint Sets Prof. Jennifer Welch Fall 2009.
© 2004 Goodrich, Tamassia Union-Find1 Union-Find Partition Structures.
Dynamic Sets and Data Structures Over the course of an algorithm’s execution, an algorithm may maintain a dynamic set of objects The algorithm will perform.
Data Structures, Spring 2004 © L. Joskowicz 1 Data Structures – LECTURE 17 Union-Find on disjoint sets Motivation Linked list representation Tree representation.
Lecture 16: Union and Find for Disjoint Data Sets Shang-Hua Teng.
CSE 373, Copyright S. Tanimoto, 2002 Up-trees - 1 Up-Trees Review of the UNION-FIND ADT Straight implementation with Up-Trees Path compression Worst-case.
© 2004 Goodrich, Tamassia Union-Find1 Union-Find Partition Structures.
CS2420: Lecture 42 Vladimir Kulyukin Computer Science Department Utah State University.
Kruskal’s algorithm for MST and Special Data Structures: Disjoint Sets
MA/CSSE 473 Day 36 Kruskal proof recap Prim Data Structures and detailed algorithm.
Ajinkya Nene, Lynbrook CS Club. 04/21/2014. Definition Allows you to figure the worst-case bound for the performance of an algorithm (most useful for.
2IL05 Data Structures Fall 2007 Lecture 13: Minimum Spanning Trees.
Spring 2015 Lecture 11: Minimum Spanning Trees
D ESIGN & A NALYSIS OF A LGORITHM 06 – D ISJOINT S ETS Informatics Department Parahyangan Catholic University.
Computer Algorithms Submitted by: Rishi Jethwa Suvarna Angal.
CS 473Lecture X1 CS473-Algorithms I Lecture X1 Properties of Ranks.
Homework remarking requests BEFORE submitting a remarking request: a)read and understand our solution set (which is posted on the course web site) b)read.
Disjoint Sets Data Structure (Chap. 21) A disjoint-set is a collection  ={S 1, S 2,…, S k } of distinct dynamic sets. Each set is identified by a member.
Lecture X Disjoint Set Operations
Disjoint Sets Data Structure. Disjoint Sets Some applications require maintaining a collection of disjoint sets. A Disjoint set S is a collection of sets.
Union-find Algorithm Presented by Michael Cassarino.
Union Find ADT Data type for disjoint sets: makeSet(x): Given an element x create a singleton set that contains only this element. Return a locator/handle.
David Luebke 1 12/12/2015 CS 332: Algorithms Amortized Analysis.
CSCE 411H Design and Analysis of Algorithms Set 7: Disjoint Sets Prof. Evdokia Nikolova* Spring 2013 CSCE 411H, Spring 2013: Set 7 1 * Slides adapted from.
Disjoint-Set Operation. p2. Disjoint Set Operations : MAKE-SET(x) : Create new set {x} with representative x. UNION(x,y) : x and y are elements of two.
Union-Find A data structure for maintaining a collection of disjoint sets Course: Data Structures Lecturers: Haim Kaplan and Uri Zwick January 2014.
Union-Find  Application in Kruskal’s Algorithm  Optimizing Union and Find Methods.
MA/CSSE 473 Days Answers to student questions Prim's Algorithm details and data structures Kruskal details.
Chapter 21 Data Structures for Disjoint Sets Lee, Hsiu-Hui Ack: This presentation is based on the lecture slides from Prof. Tsai, Shi-Chun as well as various.
WEEK 5 The Disjoint Set Class Ch CE222 Dr. Senem Kumova Metin
CSE Advanced Algorithms Instructor : Gautam Das Submitted by Raja Rajeshwari Anugula & Srujana Tiruveedhi.
David Luebke 1 3/1/2016 CS 332: Algorithms Dijkstra’s Algorithm Disjoint-Set Union.
21. Data Structures for Disjoint Sets Heejin Park College of Information and Communications Hanyang University.
Lecture 12 Algorithm Analysis Arne Kutzner Hanyang University / Seoul Korea.
MA/CSSE 473 Day 37 Student Questions Kruskal Data Structures and detailed algorithm Disjoint Set ADT 6,8:15.
CSE 373, Copyright S. Tanimoto, 2001 Up-trees - 1 Up-Trees Review of the UNION-FIND ADT Straight implementation with Up-Trees Path compression Worst-case.
CSE 373: Data Structures and Algorithms
Disjoint Sets Data Structure
CSE 373, Copyright S. Tanimoto, 2001 Up-trees -
MA/CSSE 473 Days Answers to student questions
CMSC 341 Disjoint Sets Based on slides from previous iterations of this course.
CSE373: Data Structures & Algorithms Lecture 11: Implementing Union-Find Linda Shapiro Spring 2016.
Data Structures and Algorithms
Disjoint Sets with Arrays
Disjoint Set Neil Tang 02/23/2010
Disjoint Set Neil Tang 02/26/2008
CSE 373 Data Structures and Algorithms
CSCE 411 Design and Analysis of Algorithms
CSE 373: Data Structures and Algorithms
CSE 373: Data Structures and Algorithms
Union-Find Partition Structures
Minimum Spanning Tree.
Union-Find Partition Structures
Lecture 20: Disjoint Sets
Kruskal’s algorithm for MST and Special Data Structures: Disjoint Sets
Disjoint Sets Data Structure (Chap. 21)
Lecture 21 Amortized Analysis
Review of MST Algorithms Disjoint-Set Union Amortized Analysis
CSE 373: Data Structures and Algorithms
Presentation transcript:

CPSC 411, Fall 2008: Set 7 1 CPSC 411 Design and Analysis of Algorithms Set 7: Disjoint Sets Prof. Jennifer Welch Fall 2008

CPSC 411, Fall 2008: Set 72 Disjoint Sets Abstract Data Type State: collection of disjoint dynamic sets number of sets and composition can change, but must always be disjoint each set has representative element, serves as name of the set Operations: Make-Set(x): creates {x} and adds to collection of sets Union(x,y): replaces x's set S x and y's set S y with S x U S y Find-Set(x): returns (pointer to) the representative of the set containing x

CPSC 411, Fall 2008: Set 73 Disjoint Sets Example Make-Set(a) Make-Set(b) Make-Set(c ) Make-Set(d) Union(a,b) Union(c,d) Find-Set(b) Find-Set(d) Union(b,d) b a c d returns a returns c

CPSC 411, Fall 2008: Set 74 Sample Use of Disjoint Sets Kruskal's MST algorithm Use Disjoint Sets data structure to test whether adding an edge to a set of edges would cause a cycle Each set represents nodes that are connected using edges chosen so far Initially put each node in a set by itself using Make-Set When testing edge (u,v), use Find-Set to check if u and v are in the same set - if so, then cycle When edge (u,v) is chosen, use Union(u,v)

CPSC 411, Fall 2008: Set 75 Linked List Representation Store set elements in a linked list each list node has a pointer to the next list node First list node is set representative (rep) Each list node also has a pointer to the first list node (rep) Keep external pointers to first list node (rep) and last list node (tail)

CPSC 411, Fall 2008: Set 76 Linked List Representation edafbc rep tailreptailreptail

CPSC 411, Fall 2008: Set 77 Linked List Representation Make-Set(x): make a new linked list containing just a node for x O(1) time Find-Set(x): given (pointer to) linked list node containing x, follow rep pointer to head of list O(1) time Union(x,y): append x's list to end of y's list and update all rep pointers in x's old list to point to head of y's list O(size of x's old list) time

CPSC 411, Fall 2008: Set 78 Time Analysis What is worst-case time for any sequence of Disjoint Set operations, using the linked list representation? Let m be number of ops in the sequence Let n be number of Make-Set ops (i.e., number of elements)

CPSC 411, Fall 2008: Set 79 Expensive Case MS(x 1 ), MS(x 2 ), …, MS(x n ), U(x 1,x 2 ), U(x 2,x 3 ), …, U(x n-1,x n ) Total time is O(n 2 ), which is O(m 2 ) since m = 2n - 1 So amortized time per operation is O(m 2 )/m = O(m)

CPSC 411, Fall 2008: Set 710 Linked List with Weighted Union Always append smaller list to larger list Need to keep count of number of elements in list (weight) in rep node Calculate worst-case time for a sequence of m operations. Make-Set and Find-Set operations contribute O(m) total

CPSC 411, Fall 2008: Set 711 Analyzing Time for All Unions How many times can the rep pointer for an arbitrary node x be updated? First time x's rep pointer is updated, the new set has at least 2 elements Second time x's rep pointer is updated, the new set has at least 4 elements x's set has at least 2 and the other set is at least as large as x's set

CPSC 411, Fall 2008: Set 712 Analyzing Time for All Unions The maximum size of a set is n (the number of Make-Set ops in the sequence) So x's rep pointer can be updated at most log n times. Thus total time for all unions is O(n log n). Note style of counting - focus on one element and how it fares over all the Unions

CPSC 411, Fall 2008: Set 713 Amortized Time Grand total for sequence is O(m+n log n) Amortized cost per Make-Set and Find-Set is O(1) Amortized cost per Union is O(log n) since there can be at most n - 1 Union ops.

CPSC 411, Fall 2008: Set 714 Tree Representation Can we improve on the linked list with weighted union representation? Use a collection of trees, one per set The rep is the root Each node has a pointer to its parent in the tree

CPSC 411, Fall 2008: Set 715 Tree Representation a de b f c

CPSC 411, Fall 2008: Set 716 Analysis of Tree Implementation Make-Set: make a tree with one node O(1) time Find-Set: follow parent pointers to root O(h) time where h is height of tree Union(x,y): make the root of x's tree a child of the root of y's tree O(1) time So far, no better than original linked list implementation

CPSC 411, Fall 2008: Set 717 Improved Tree Implementation Use a weighted union, so that smaller tree becomes child of larger tree prevents long chains from developing can show this gives O(m log n) time for a sequence of m ops with n Make-Sets Also do path compression during Find-Set flattens out trees even more can show this gives O(m log*n) time!

CPSC 411, Fall 2008: Set 718 What is log*n ? The number of times you can successively take the log, starting with n, before reaching a number that is at most 1 More formally: log*n = min{i ≥ 0 : log (i) n ≤ 1} where log (i) n = n, if i = 0, and otherwise log (i) n = log(log (i-1) n)

CPSC 411, Fall 2008: Set 719 Examples of log*n nlog*nwhy 000 < ≤ 1 21log 2 = 1 32log 3 > 1, log(log 3) < 1 42log(log 4) = log 2 = 1 53log(log 5) > 1, log(log(log 5)) < 1 … 163log(log(log 16) = log(log 4) = log 2 = 1 174log(log(log 17) > 1, log(log(log(log 17))) < 1 … 65,5364log(log(log(log 65,536) = log(log(log 16) = log(log 4) = log 2 = 1 65,5375 … 2 65,537 5

CPSC 411, Fall 2008: Set 720 log*n Grows Slowly For all practical values of n, log*n is never more than 5.

CPSC 411, Fall 2008: Set 721 Make-Set Make-Set(x): parent(x) := x rank(x) := 0 // used for weighted union

CPSC 411, Fall 2008: Set 722 Union Union(x,y): r := Find-Set(x); s := Find-Set(y) if rank(r) > rank(s) then parent(s) := r else parent(r) := s if rank(r ) = rank(s) then rank(s)++

CPSC 411, Fall 2008: Set 723 Rank gives upper bound on height of tree is approximately the log of the number of nodes in the tree Example: MS(a), MS(b), MS(c), MS(d), MS(e), MS(f), U(a,b), U(c,d), U(e,f), U(a,c), U(a,e)

CPSC 411, Fall 2008: Set 724 End Result of Rank Example d c b a f e

CPSC 411, Fall 2008: Set 725 Find-Set Find-Set(x): if x  parent(x) then parent(x) := Find-Set(parent(x)) return parent(x) Unroll recursion: first, follow parent pointers up the tree then go back down the path, making every node on the path a child of the root

CPSC 411, Fall 2008: Set 726 Find-Set(a) a b c d e e dcba

CPSC 411, Fall 2008: Set 727 Amortized Analysis Show any sequence of m Disjoint Set operations, n of which are Make-Sets, takes O(m log*n) time with the improved tree implementation. Use aggregate method. Assume Union always operates on roots otherwise analysis is only affected by a factor of 3

CPSC 411, Fall 2008: Set 728 Charging Scheme Charge 1 unit for each Make-Set Charge 1 unit for each Union Set 1 unit of charge to be large enough to cover the actual cost of these constant-time operations

CPSC 411, Fall 2008: Set 729 Charging Scheme Actual cost of Find-Set(x) is proportional to number of nodes on path from x to its root. Assess 1 unit of charge for each node in the path (make unit size big enough). Partition charges into 2 different piles: block charges and path charges

CPSC 411, Fall 2008: Set 730 Overview of Analysis For each Find-Set, partition charges into block charges and path charges To calculate all the block charges, bound the number of block charges incurred by each Find-Set To calculate all the path charges, bound the number of path charges incurred by each node (over all the Find-Sets that it participates in)

CPSC 411, Fall 2008: Set 731 Blocks Consider all possible ranks of nodes and group ranks into blocks Put rank r into block log*r ranks: … … … blocks:

CPSC 411, Fall 2008: Set 732 Charging Rule for Find-Set Fact: ranks of nodes along path from x to root are strictly increasing. Fact: block values along the path are non- decreasing Rule: root, child of root, and any node whose rank is in a different block than the rank of its parent is assessed a block charge each remaining node is assessed a path charge

CPSC 411, Fall 2008: Set 733 Find-Set Charging Rule Figure … … … block b'' > b' block b' > b block b 1 block charge (root) 1 block charge 1 path charge 1 block charge 1 path charge

CPSC 411, Fall 2008: Set 734 Total Block Charges Consider any Find-Set(x). Worst case is when every node on the path from x to the root is in a different block. Fact: There are at most log*n different blocks. So total cost per Find-Set is O(log*n) Total cost for all Find-Sets is O(m log*n)

CPSC 411, Fall 2008: Set 735 Total Path Charges Consider a node x that is assessed a path charge during a Find-Set. Just before the Find-Set executes: x is not a root x is not a child of a root x is in same block as its parent As a result of the Find-Set executing: x gets a new parent due to the path compression

CPSC 411, Fall 2008: Set 736 Total Path Charges x could be assessed another path charge in a subsequent Find-Set execution. However, x is only assessed a path charge if it's in same block as its parent Fact: A node's rank only increases while it is a root. Once it stops being a root, its rank, and thus its block, stay the same. Fact: Every time a node gets a new parent (because of path compression), new parent's rank is larger than old parent's rank

CPSC 411, Fall 2008: Set 737 Total Path Charges So x will contribute path charges in multiple Find- Sets as long as it can be moved to a new parent in the same block Worst case is when x has lowest rank in its block and is successively moved to a parent with every higher rank in the block Thus x contributes at most M(b) path charges, where b is x's block and M(b) is the maximum rank in block b

CPSC 411, Fall 2008: Set 738 Total Path Charges Fact: There are at most n/M(b) nodes in block b. Thus total path charges contributed by all nodes in block b is M(b)*n/M(b) = n. Since there are log*n different blocks, total path charges is O(n log*n), which is O(m log*n).

CPSC 411, Fall 2008: Set 739 Even Better Bound By working even harder, and using a potential function analysis, can show that the worst-case time is O(m  (n)), where  is a function that grows even more slowly than log*n: for all practical values of n,  (n) is never more than 4.

CPSC 411, Fall 2008: Set 740 Effect on Running Time of Kruskal's Algorithm |V| Make-Sets, one per node 2|E| Find-Sets, two per edge |V| - 1 Unions, since spanning tree has |V| - 1 edges in it So sequence of O(E) operations, |V| of which are Make-Sets Time for Disjoint Sets ops is O(E log*V) Dominated by time to sort the edges, which is O(E log E) = O(E log V).