Shortest Path Algorithms

Single Source Shortest Path

Dijkstra’s Algorithm

Shortest path examples

„ Highway system
– Distance
– Travel time
– Number of stoplights
– Krispy Kreme locations
„ Network of airports
– Travel time
– Fares
– Actual distance

1
Weighted path length
„ Consider an edge-weighted graph
G = (V,E).
„ Let C(vi,vj) be the weight on the edge
connecting vi to vj.
„ A path in G is a non-empty sequence of
vertices P = {v1, v2, v3, …, vk}.
„ The weighted path length is given by
k-1

∑C(v ,v
i=1
i i+1)

The general problem

„ Given an edge-weighted graph

G = (V,E) and two vertices, vs ∈ V and
vd ∈ V, find the path that starts at vs
and ends at vd that has the smallest
weighted path length

Single-source shortest path

„ Given an edge-weighted graph

G = (V,E) and a vertex, vs ∈ V, find the
shortest path from vs to every other
vertex in V

„ To find the shortest path from vs to vd,

we must find the shortest path from vs
to every vertex in G

2
Shortest weighted path
from b to f:
a
{b, a, c, e, f}
3 1 5

b 5 c 2 d

4 5

e 1 f

Shortest weighted path

from b to f:
a
{b, a, c, e, f}
3 1 5

b 5 c 2 d

4 5
Shortest unweighted
path from b to f: 1
e f
{b, c, e, f}

Shortest path problem

undefined for graphs
a
with negative-cost
cycles
3 1 5

b 5 c -8 d

4 5
{d, a, c, e, f} cost: 4
{d, a, c, d, a, c, e, f} -6
e f
cost: 2
{d, a, c, d, a, c, d, a, c, e, f} cost: 0

3
Dijkstra’s Algorithm

„ Greedy algorithm for solving shortest

path problem
„ Assume non-negative weights
„ Find shortest path from vs to each other
vertex

Dijkstra’s Algorithm

„ For each vertex v, need to know:

– kv: Is the shortest path from vs to v known?
(initially false for all v ∈ V)
– dv: What is the length of the shortest path
from vs to v? (initially ∞ for all v ∈ V, except
vs = 0)
– pv: What vertex precedes (is parent of) v on
the shortest path from vs to v? (initially
unknown for all v ∈ V)

v kv dv pv a

a F ∞ 3 1 5

b F 0
b 5 c 2 d
c F ∞
4 5
d F ∞
e 1 f
e F ∞
f F ∞

4
v kv dv pv a

a F 3 b 5
3 1

b T 0 --
b 5 c 2 d
c F 5 b
4 5
d F ∞
e 1 f
e F ∞
f F ∞

v kv dv pv a

a T 3 b 5
3 1

b T 0 --
b 5 c 2 d
c F 4 a
4 5
d F 8 a
e 1 f
e F ∞
f F ∞

v kv dv pv a

a T 3 b 5
3 1

b T 0 --
b 5 c 2 d
c T 4 a
4 5
d F 6 c
e 1 f
e F 8 c

f F ∞

5
v kv dv pv a

a T 3 b 5
3 1

b T 0 --
b 5 c 2 d
c T 4 a
4 5
d T 6 c
e 1 f
e F 8 c

f F 11 d

v kv dv pv a

a T 3 b 5
3 1

b T 0 --
b 5 c 2 d
c T 4 a
4 5
d T 6 c
e 1 f
e T 8 c

f F 9 e

v kv dv pv a

a T 3 b 5
3 1

b T 0 --
b 5 c 2 d
c T 4 a
4 5
d T 6 c
e 1 f
e T 8 c

f T 9 e

6
 3
v kv dv pv a

a T 3 b

b T 0 -- 6
b 4 c d
c T 4 a

d T 6 c
9
e T 8 c 8 e f

f T 9 e

Dijsktra’s Algorithm
Dijsktra(G,s0)
{
//Initialize
n = |V|; O( )
createtable(n); //stores k,d,p O( )
createpq(|E|); //empty pq O( )
table[s0].d = 0; O( )
insertpq(0,s0); O( )

Dijsktra’s Algorithm
//con’t
while (!pq.isempty) O( )
{
v0 = getMin(); O( )
if (!table[v0].k) //not known O( )
{
table[v0].k = true; O( )
for each vi ∈ Adj[v0] O( )
{
newd=table[v0].d + weight(vi,v0); O( )

7
Dijsktra’s Algorithm
//con’t
// newd=table[v0].d + weight(vi,v0);
if (table[vi].d) > newd) O( )
{
table[vi].d = newd; O( )
table[vi].p = v0; O( )
insertpq(newd,vi); O( )
}
}
}
}

Dijsktra’s Algorithm
//con’t
for each v ∈ G(V,E) O( )
//build vertex set in T
v ∈ T(V,E’);

for each v ∈ G(V,E) O( )

//build edge set in T
(v,table[v].p) ∈ T(V,E’);
}

All-pairs shortest path problem

„ Given an edge-weighted graph

G = (V,E), for each pair of vertices in V
find the length of the shortest weighted
path between the two vertices

Solution: Run Dijkstra V times