1 / 36

David Kaplan davek@cs

CSE 326: Data Structures Lecture #19 Disjoint Sets Dynamic Equivalence Weighted Union & Path Compression. David Kaplan davek@cs. Today’s Outline. Making a “good” maze Disjoint Set Union/Find ADT Up-trees Maze revisited Weighted Union Path Compression An a maz ing complexity analysis.

jdemarest
Download Presentation

David Kaplan davek@cs

An Image/Link below is provided (as is) to download presentation Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author. Content is provided to you AS IS for your information and personal use only. Download presentation by click this link. While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server. During download, if you can't get a presentation, the file might be deleted by the publisher.

E N D

Presentation Transcript

  1. CSE 326: Data StructuresLecture #19Disjoint SetsDynamic EquivalenceWeighted Union & Path Compression David Kaplan davek@cs

  2. Today’s Outline • Making a “good” maze • Disjoint Set Union/Find ADT • Up-trees • Maze revisited • Weighted Union • Path Compression • An amazing complexity analysis

  3. The Maze Construction Problem • Represent maze environment as graph {V,E} • collection of rooms: V • connections between rooms (initially all closed): E • Construct a maze: • collection of rooms: V = V • designated rooms in, iV, and out, oV • collection of connections to knock down: E  E such that one unique path connects every two rooms

  4. The Middle of the Maze • So far, some walls have been knocked down while others remain. • Now, we consider the wall between A and B. • Should we knock it down? • no, if A and Bare otherwise connected • yes, if A and Bare not otherwise connected A B

  5. Maze Construction Algorithm While edges remain in E • Remove a random edge e = (u, v) from E • If u and v have not yet been connected • add e to E • mark u and v as connected

  6. Equivalence Relations An equivalence relation R has three properties: • reflexive: for any x, xRx is true • symmetric: for any x and y, xRy implies yRx • transitive: for any x, y,and z, xRy and yRz implies xRz Connection between rooms is an equivalence relation (call it C) For any rooms a, b, c aCa (A room connects to itself!) If aCb, then bCa If aCb and bCc, then aCc Examples of other equivalence relations?

  7. find(4) {1,4,8} {6} 8 {7} {2,3,6} {5,9,10} {2,3} union(2,6) Disjoint Set Union/Find ADT • Union/Find operations • create • destroy • union • find • Disjoint set equivalence property: every element of a DS U/F structure belongs to exactly one set • Dynamic equivalence property: the set of an element can change after execution of a union

  8. Disjoint Set Union/FindMore Formally Given a set U = {a1, a2, … , an} Maintain a partition of U, a set of subsets of U {S1, S2, … , Sk} such that: • each pair of subsets Si and Sj are disjoint: • together, the subsets cover U: • The Si are mutually exclusive and collectively exhaustive Union(a, b) creates a new subset which is the union of a’s subset and b’s subset Find(a) returns a unique name for a’s subset • NOTE: set names are arbitrary! We only care that: • Find(a) == Find(b)  a and b are in the same subset • NOTE: outside agent must decide when/what to union; ADT is just the bookkeeper.

  9. a b c 3 10 2 1 6 d e f 4 7 11 9 8 g h i 12 5 Example Construct the maze on the right Initial state (set names underlined): {a}{b}{c}{d}{e}{f}{g}{h}{i} Maze constructor (outside agent) traverses edges in numeric order and decides whether to union

  10. 3 10 2 1 6 4 7 11 9 8 12 5 Example, First Step {a}{b}{c}{d}{e}{f}{g}{h}{i} find(b) b find(e) e find(b)  find(e) so: add 1 to E union(b, e) {a}{b,e}{c}{d}{f}{g}{h}{i} a b c d e f g h i Order of edges in blue

  11. Up-Tree Intuition Finding the representative member of a set is somewhat like the opposite of finding whether a given item exists in a set. So, instead of using trees with pointers from each node to its children; let’s use trees with a pointer from each node to its parent.

  12. Up-Tree Union-Find Data Structure • Each subset is an up-tree with its root as its representative member • All members of a given set are nodes in that set’s up-tree • Hash table maps input data to the node associated with that data a c g h f i d b e Up-trees are not necessarily binary!

  13. Find find(f) find(e) a b c 10 a c g h d e f 7 11 9 8 f i d b g h i 12 e Just traverse to the root! runtime:

  14. Union union(a,c) a b c 10 a c g h d e f 11 9 8 f i d b g h i 12 e Just hang one root from the other! runtime:

  15. a b c 3 10 2 1 6 The Whole Example (1/11) d e f 4 7 11 9 8 union(b,e) g h i 12 5 a b c d e f g h i a b c d f g h i e

  16. a b c 3 10 2 6 The Whole Example (2/11) d e f 4 7 11 9 8 union(a,d) g h i 12 5 a b c d f g h i e a b c f g h i d e

  17. a b c 3 10 6 The Whole Example (3/11) d e f 4 7 11 9 8 union(a,b) g h i 12 5 a b c f g h i d e a c f g h i b d e

  18. a c f g h i b d e While we’re finding e, could we do anything else? a b c 10 6 The Whole Example (4/11) d e f 4 7 11 9 8 find(d) = find(e) No union! g h i 12 5

  19. a b c 10 6 The Whole Example (5/11) d e f 7 11 9 8 union(h,i) g h i 12 5 a c f g h i a c f g h b d i b d e e

  20. a b c 10 6 The Whole Example (6/11) d e f 7 11 9 8 union(c,f) g h i 12 a c f g h a c g h b d b d f i i e e

  21. a c g h c g h b d f a f i i b d e Could we do a better job on this union? e a b c 10 The Whole Example (7/11) d e f 7 find(e) find(f) union(a,c) 11 9 8 g h i 12

  22. a b c 10 The Whole Example (8/11) d e f find(f) find(i) union(c,h) 11 9 8 g h i 12 c g h c g a f a f h i b d b d i e e

  23. a b c 10 The Whole Example (9/11) d e f find(e) = find(h) and find(b) = find(c) So, no unions for either of these. 11 9 g h i 12 c g a f h b d i e

  24. a b c The Whole Example (10/11) d e f find(d) find(g) union(c, g) 11 g h i 12 c g g c a f h a f h b d i b d i e e

  25. a b c The Whole Example (11/11) d e f find(g) = find(h) So, no union. And, we’re done! g h i 12 g a b c c d e f a f h g h i b d i Ooh… scary! Such a hard maze! e

  26. a c g h b d f i e 0 (a)1 (b)2 (c)3 (d)4 (e)5 (f)6 (g)7 (h)8 (i) -1 0 -1 0 1 2 -1 -1 7 DS/DE data structure A forest of up-trees can easily be stored in an array. Also, if the node names are integers or characters, we can use a very simple, perfect hash. Nifty storage trick! up-index:

  27. Implementation typedef ID int; ID find(Object x) { assert(hTable.contains(x)); ID xID = hTable[x]; while(up[xID] != -1) { xID = up[xID]; } return xID; } ID union(ID x, ID y) { assert(up[x] == -1); assert(up[y] == -1); up[y] = x; } runtime: O(depth) or … runtime: O(1)

  28. a c g h b d f i e Room for Improvement:Weighted Union • Always makes the root of the larger tree the new root • Often cuts down on height of the new up-tree c g h a g h a f b d i c i b d f e Could we do a better job on this union? e Weighted union!

  29. Weighted Union Code typedef ID int; ID union(ID x, ID y) { assert(up[x] == -1); assert(up[y] == -1); if (weight[x] > weight[y]) { up[y] = x; weight[x] += weight[y]; } else { up[x] = y; weight[y] += weight[x]; } } new runtime of union: new runtime of find:

  30. Weighted Union Find Analysis • Finds with weighted union are O(max up-tree height) • But, an up-tree of height h with weighted union must have at least 2h nodes •  2max height n and max height  log n • So, find takes O(log n) Base case: h = 0, tree has 20 = 1 node Induction hypothesis: assume true for h < h A merge can only increase tree height by one over the smaller tree. So, a tree of height h-1 was merged with a larger tree to form the new tree. Each tree then has  2h-1 nodes by the induction hypotheses for a total of at least 2h nodes. QED.

  31. a c f g h i b d e Room for Improvement: Path Compression • Points everything along the path of a find to the root • Reduces the height of the entire access path to 1 a c f g h i b d e While we’re finding e, could we do anything else? Path compression!

  32. Path Compression Example find(e) c c g g a f h a f h b e b d d i i e

  33. Path Compression Code typedef ID int; ID find(Object x) { assert(hTable.contains(x)); ID xID = hTable[x]; ID hold = xID; while(up[xID] != -1) { xID = up[xID]; } while(up[hold] != -1) { temp = up[hold]; up[hold] = xID; hold = temp; } return xID; } runtime:

  34. Complexity of Weighted Union + Path Compression Ackermann created a function A(x, y) which grows very fast! Inverse Ackermann function (x, y) grows very slooooowwwly … Single-variable inverse Ackermann function is called log* n How fast does log n grow?log n = 4 for n = 16 Let log(k) n = log (log (log … (log n))) Then, let log* n = minimum k such that log(k) n  1 How fast does log* n grow?log* n = 4 for n = 65536 log* n = 5 for n = 265536 (20,000 digit number!) How fast does (x, y) grow? Even sloooowwwer than log* n (x, y) = 4 for n far larger than the number of atoms in the universe (2300) k logs

  35. Complex Complexity of Weighted Union + Path Compression • Tarjan proved that m weighted union and find operations on a set of n elements have worst case complexity O(m(m, n)) • This is essentially amortized constant time • In some practical cases, weighted union or path compression or both are unnecessary because trees do not naturally get very deep.

  36. Disjoint Set Union/Find ADT Summary • Simple ADT, simple data structure, simple code • Complex complexity analysis, but extremely useful result: essentially, constant time! • Lots of potential applications • Object-property collections • Index partitions: e.g. parts inspection • To say nothing of maze construction • In some applications, it may make sense to have meaningful (non-arbitrary) set names

More Related