Chapter 3 Breadth First Search, Dijkstras Algorithm for Shortest Paths

CS 473: Fundamental Algorithms, Spring 2011 January 25, 2011

3.1
3.1.0.1

Breadth First Search

Overview 1. BFS is obtained from BasicSearch by processing edges using a data structure called a queue. 2. It processes the vertices in the graph in the order of their shortest distance from the vertex s (the start vertex). As such... (A) DFS good for exploring graph structure (B) BFS good for exploring distances 3.1.0.2 Queues A queue is a list of elements which supports the following operations (A) enqueue: Adds an element to the end of the list (B) dequeue: Removes an element from the front of the list Elements are extracted in rst-in rst-out (FIFO) order, i.e., elements are picked in the order in which they were inserted. 1 Queue Data Structure

1 2 4 5 6 3

1 2 4 5 6 8 3 7

3.1.0.3

BFS Algorithm

Given (undirected or directed) graph G = (V, E) and node s V

Proposition 3.1.1 BFS(s) runs in O(n + m) time.

3.1.0.4

BFS: An Example in Undirected Graphs 1. 2. 3. [1] [2,3] [3,4,5] 4. [4,5,7,8] 5. [5,7,8] 6. [7,8,6] 7. [8,6] 8. [6] 9. []

BFS tree is the set of black edges. 2

Basic Graph Theory Breadth First search Depth First Search Directed Graphs

Digraphs and Connectivity Digraph Representation Searching

Graphs

3.1.0.5
B

BFS: An Example in Directed Graphs

on 3.1.0.6 BFS with Distance ed graph (also called a digraph) is G = (V , E ), where

s a set of vertices or nodes

V V is set of ordered
Viswanathan

BFS(s) Mark all vertices as unvisited and for each v set dist(v) = Initialize search tree T to be empty pairs of vertices called edges set dist(s) = 0 Mark vertex s as visited and set Q to be the empty queue enq(s) while Q is nonempty do CS473ug u = deq(Q) for each vertex v Adj(u) do if v is not visited do add edge (u, v) to T Mark v as visited, enq(v) and set dist(v) = dist(u) + 1

3.1.0.7

Properties of BFS: Undirected Graphs

Proposition 3.1.2 The following properties hold upon termination of BFS(s) 1. The search tree contains exactly the set of vertices in the connected component of s. 2. If dist(u) < dist(v) then u is visited before v. 3. For every vertex u, dist(u) is indeed the length of shortest path from s to u. 4. If u, v are in connected component of s and e = {u, v} is an edge of G, then either e is an edge in the search tree, or |dist(u) dist(v)| 1. Proof : Exercise. 3.1.0.8 Properties of BFS: Directed Graphs

Proposition 3.1.3 The following properties hold upon termination of BFS(s): 1. The search tree contains exactly the set of vertices reachable from s 3

1 2 4 5 6 3

2. If dist(u) < dist(v) then u is visited before v 3. For every vertex u, dist(u) is indeed the length of shortest path from s to u 4. If u is reachable from s and e = (u, v) is an edge of G, then either e is an edge in the search tree, or dist(v) dist(u) 1. Not necessarily the case that dist(u) dist(v) 1. Proof : Exercise. 3.1.0.9 BFS with Layers
BFSLayers(s): Mark all vertices as unvisited and initialize T to be empty Mark s as visited and set L0 = {s} i=0 while Li is not empty do initialize Li+1 to be an empty list for each u in Li do for each edge (u, v) Adj(u) do if v is not visited mark v as visited add (u, v) to tree T add v to Li+1 i=i+1

Running time: O(n + m) 3.1.0.10 3.1.0.11 Example BFS with Layers: Properties

Proposition 3.1.4 The following properties hold on termination of BFSLayers(s). (A) BFSLayers(s) outputs a BFS tree (B) Li is the set of vertices at distance exactly i from s (C) If G is undirected, each edge e = {u, v} is one of three types: (A) tree edge between two consecutive layers (B) non-tree forward/backward edge between two consecutive layers (C) non-tree cross-edge with both u, v in same layer (D) = Every edge in the graph is either between two vertices that are either (i) in the same layer, or (ii) in two consecutive layers. 4

3.1.1
3.1.1.1

BFS with Layers: Properties

Proposition 3.1.5 The following properties hold on termination of BFSLayers(s), if G is directed. For each edge e = (u, v) is one of four types: (A) a tree edge between consecutive layers, u Li , v Li+1 for some i 0 (B) a non-tree forward edge between consecutive layers (C) a non-tree backward edge (D) a cross-edge with both u, v in same layer

3.2
3.2.0.2

Bipartite Graphs and an application of BFS

Denition 3.2.1 (Bipartite Graph) Undirected graph G = (V, E) is a bipartite graph if V can be partitioned into X and Y s.t. all edges in E are between X and Y .

3.2.0.3

Bipartite Graph Characterization

Question When is a graph bipartite? Proposition 3.2.2 Every tree is a bipartite graph. Proof : Root tree T at some node r. Let Li be all nodes at level i, that is, Li is all nodes at distance i from root r. Now dene X to be all nodes at even levels and Y to be all nodes at odd level. Only edges in T are between levels. Proposition 3.2.3 An odd length cycle is not bipartite. 5

3.2.0.4

Odd Cycles are not Bipartite

Proposition 3.2.4 An odd length cycle is not bipartite. Proof : Let C = u1 , u2 , . . . , u2k+1 , u1 be an odd cycle. Suppose C is a bipartite graph and let X, Y be the bipartition. Without loss of generality u1 X. Implies u2 Y . Implies u3 X. Inductively, ui X if i is odd ui Y if i is even. But {u1 , u2k+1 } is an edge and both belong to X! 3.2.0.5 Subgraphs

Denition 3.2.5 Given a graph G = (V, E) a subgraph of G is another graph H = (V , E ) where V V and E E. Proposition 3.2.6 If G is bipartite then any subgraph H of G is also bipartite. Proposition 3.2.7 A graph G is not bipartite if G has an odd cycle C as a subgraph. Proof : If G is bipartite then since C is a subgraph, C is also bipartite (by above proposition). However, C is not bipartite! 3.2.0.6 Bipartite Graph Characterization

Theorem 3.2.8 A graph G is bipartite if and only if it has no odd length cycle as subgraph. Proof : Only If: G has an odd cycle implies G is not bipartite. If: G has no odd length cycle. Assume without loss of generality that G is connected. (A) Pick u arbitrarily and do BFS(u) (B) X = i is even Li and Y = i is odd Li (C) Claim: X and Y is a valid bipartition if G has no odd length cycle.

3.2.0.7

Proof of Claim

Claim 3.2.9 In BFS(u) if a, b Li and (a, b) is an edge then there is an odd length cycle containing (a, b). Proof : Let v be least common ancestor of a, b in BFS tree T . v is in some level j < i (could be u itself). Path from v a in T is of length j i. Path from v b in T is of length j i. These two paths plus (a, b) forms an odd cycle of length 2(j i) + 1. Corollary 3.2.10 There is an O(n+m) time algorithm to check if G is bipartite and output an odd cycle if it is not. 6

3.3
3.3.0.8

Shortest Paths and Dijkstras Algorithm

Shortest Path Problems Input A (undirected or directed) graph G = (V, E) with edge lengths (or costs). For edge e = (u, v), (e) = (u, v) is its length. (A) Given nodes s, t nd shortest path from s to t. (B) Given node s nd shortest path from s to all other nodes. (C) Find shortest paths for all pairs of nodes. Many applications! 3.3.0.9 Single-Source Shortest Paths: Non-Negative Edge Lengths

Single-Source Shortest Path Problems Input A (undirected or directed) graph G = (V, E) with non-negative edge lengths. For edge e = (u, v), (e) = (u, v) is its length. (A) Given nodes s, t nd shortest path from s to t. (B) Given node s nd shortest path from s to all other nodes. (A) Restrict attention to directed graphs (B) Undirected graph problem can be reduced to directed graph problem - how? (A) Given undirected graph G, create a new directed graph G by replacing each edge {u, v} in G by (u, v) and (v, u) in G . (B) set (u, v) = (v, u) = ({u, v}) (C) Exercise: show reduction works 3.3.0.10 Single-Source Shortest Paths via BFS

Special case: All edge lengths are 1. (A) Run BFS(s) to get shortest path distances from s to all other nodes. (B) O(m + n) time algorithm. Special case: Suppose (e) is an integer for all e? Can we use BFS? Reduce to unit edge-length problem by placing (e) 1 dummy nodes on e

Let L = maxe (e). New graph has O(mL) edges and O(mL + n) nodes. BFS takes O(mL + n) time. Not ecient if L is large. 7

3.3.0.11

Towards an algorithm

Why does BFS work? BFS(s) explores nodes in increasing distance from s

Lemma 3.3.1 Let G be a directed graph with non-negative edge lengths. Let dist(s, v) denote the shortest path length from s to v. If s = v0 v1 v2 . . . vk is a shortest path from s to vk then for 1 i < k: (A) s = v0 v1 v2 . . . vi is a shortest path from s to vi (B) dist(s, vi ) dist(s, vk ).

Proof : Suppose not. Then for some i < k there is a path P from s to vi of length strictly less than that of s = v0 v1 . . . vi . Then P concatenated with vi vi+1 . . . vk contains a strictly shorter path to vk than s = v0 v1 . . . vk .

3.3.0.12

A proof by picture

Shorter path from v0 to v4 s = v0 v2 v4 v1 v3 Shortest path from v0 to v6 Shortest path from v0 to v6 v1 v3 v5 s = v0 v2 v4 v5

v6

v6

s=v

Shor from

3.3.0.13

A Basic Strategy

Explore vertices in increasing order of distance from s: (For simplicity assume that nodes are at dierent distances from s and that no edge has zero length) 8

Initialize for each node v, dist(s, v) = Initialize S = , for i = 1 to |V | do (* Invariant: S contains the i 1 closest nodes to s *) Among nodes in V \ S, find the node v that is the ith closest to s Update dist(s, v) S = S {v}

How can we implement the step in the for loop? 3.3.0.14 Finding the ith closest node

(A) S contains the i 1 closest nodes to s (B) Want to nd the ith closest node from V S. What do we know about the ith closest node?

Claim 3.3.2 Let P be a shortest path from s to v where v is the ith closest node. Then, all intermediate nodes in P belong to S. Proof : If P had an intermediate node u not in S then u will be closer to s than v. Implies v is not the ith closest node to s - recall that S already has the i 1 closest nodes.

3.3.1
3.3.1.1

Finding the ith closest node repeatedly

a
0

10 9 18 6 30 13 8 20 25 16 6 11 19 6

a
0
6

10

f
6 6 11 19 6

9 6

c
20

18 30

a
g 0

13 8

d
25

16

3.3.1.2

Finding the ith closest node

Corollary 3.3.3 The ith closest node is adjacent to S. 9

a
0

10 9 18 6 30 13 8 20 25 16 6 11 19 6 6

3.3.1.3

Finding the ith closest node

(A) S contains the i 1 closest nodes to s (B) Want to nd the ith closest node from V S. (A) For each u V S let P (s, u, S) be a shortest path from s to u using only nodes in S as intermediate vertices. (B) Let d (s, u) be the length of P (s, u, S) Observations: for each u V S, (A) dist(s, u) d (s, u) since we are constraining the paths (B) d (s, u) = minaS (dist(s, a) + (a, u)) - Why? Lemma 3.3.4 If v is the ith closest node to s, then d (s, v) = dist(s, v). 3.3.1.4 Finding the ith closest node

Lemma 3.3.5 If v is an ith closest node to s, then d (s, v) = dist(s, v). Proof : Let v be the ith closest node to s. Then there is a shortest path P from s to v that contains only nodes in S as intermediate nodes (see previous claim). Therefore d (s, v) = dist(s, v). 3.3.1.5 Finding the ith closest node

Lemma 3.3.6 If v is an ith closest node to s, then d (s, v) = dist(s, v). Corollary 3.3.7 The ith closest node to s is the node v V S such that d (s, v) = minuV S d (s, u). Proof : For every node u V S, dist(s, u) d (s, u) and for the ith closest node v, dist(s, v) = d (s, v). Moreover, dist(s, u) dist(s, v) for each u V S. 10

10 9 s 6 6 6 13 8 7 20 30 5 16 25 t 11 4 19 6 2 18 6 3

10 9 0 6 6 6 13 25 11 36 9 19

25 13 38

3.3.1.6

Algorithm
Initialize for each node v: dist(s, v) = Initialize S = , d (s, s) = 0 for i = 1 to |V | do (* Invariant: S contains the i-1 closest nodes to s *) (* Invariant: d(s,u) is shortest path distance from u to s using only S as intermediate nodes*) Let v be such that d(s,v) = minuV S d(s,u) dist(s, v) = d (s, v) S = S {v} for each node u in V \ S compute d(s,u) = minaS (dist(s, a) + (a, u))

Correctness: By induction on i using previous lemmas. Running time: O(n (n + m)) time. (A) n outer iterations. In each iteration, d (s, u) for each u by scanning all edges out of nodes in S; O(m + n) time/iteration.

3.3.1.7 3.3.1.8

Example Improved Algorithm

(A) Main work is to compute the d (s, u) values in each iteration (B) d (s, u) changes from iteration i to i + 1 only because of the node v that is added to S in iteration i. 11

Initialize for each node v, dist(s, v) = d (s, v) = Initialize S = , d(s,s) = 0 for i = 1 to |V | do // S contains the i 1 closest nodes to s, // and the values of d (s, u) are current Let v be such that d(s,v) = minuV S d(s,u) dist(s, v) = d (s, v) S = S {v} Update d(s,u) for each u in V-S as follows: d (s, u) = min(d (s, u), dist(s, v) + (v, u))

Running time: O(m + n2 ) time. (A) n outer iterations and in each iteration following steps (B) updating d (s, u) after v added takes O(deg(v)) time so total work is O(m) since a node enters S only once (C) Finding v from d (s, u) values is O(n) time 3.3.1.9 Dijkstras Algorithm

(A) eliminate d (s, u) and let dist(s, u) maintain it (B) update dist values after adding v by scanning edges out of v
Initialize for each node v, dist(s, v) = Initialize S = {s}, dist(s, s) = 0 for i = 1 to |V | do Let v be such that dist(s, v) = minuV S dist(s, u) S = S {v} for each u in Adj(v) do dist(s, u) = min(dist(s, u), dist(s, v) + (v, u))

Priority Queues to maintain dist values for faster running time (A) Using heaps and standard priority queues: O((m + n) log n) (B) Using Fibonacci heaps: O(m + n log n).

3.3.2
3.3.2.1

Priority Queues
Priority Queues

Data structure to store a set S of n elements where each element v S has an associated real/integer key k(v) such that the following operations (A) makeQ: create an empty queue (B) findMin: nd the minimum key in S (C) extractMin: Remove v S with smallest key and return it (D) add(v, k(v)): Add new element v with key k(v) to S (E) delete(v): Remove element v from S 12

(F) decreaseKey(v, k(v)): decrease key of v from k(v) (current key) to k (v) (new key). Assumption: k (v) k(v) (G) meld: merge two separate priority queues into one can be performed in O(log n) time each. decreaseKey via delete and add 3.3.2.2 Dijkstras Algorithm using Priority Queues
Q = makePQ() insert(Q, (s, 0)) for each node u = s do insert(Q, (u,)) S= for i = 1 to |V | do (v, dist(s, v)) = extractM in(Q) S = S {v} For each u in Adj(v) do decreaseKey(Q, (u, min(dist(s, u), dist(s, v) + (v, u))))

Priority Queue operations: (A) O(n) insert operations (B) O(n) extractMin operations (C) O(m) decreaseKey operations 3.3.2.3 Implementing Priority Queues via Heaps

Using Heaps Store elements in a heap based on the key value (A) All operations can be done in O(log n) time

Dijkstras algorithm can be implemented in O((n + m) log n) time. 3.3.2.4 Priority Queues: Fibonacci Heaps/Relaxed Heaps

Fibonacci Heaps (A) extractMin, add, delete, meld in O(log n) time (B) decreaseKey in O(1) amortized time: decreaseKey operations for n take together O( ) time (C) Relaxed Heaps: decreaseKey in O(1) worst case time but at the expense of meld (not necessary for Dijkstras algorithm) Dijkstras algorithm can be implemented in O(n log n + m) time. If m = (n log n), running time is linear in input size. 13