Dijkstra’s Algorithm
Shortest Path Problem
Directed graph G = (V, E)
Source s
le = length of edge e
le ≥ 0 for all edges e
Shortest path problem: (Google Maps!) find shortest path
from s to all other nodes
1
a c
2
1 3 1 e 1 f
4
b d
1
Simplification
For now, let’s just find the lengths of the
shortest paths to every other node
We’ll show how to recover the paths later
Dijkstra’s Algorithm
Idea: explore outward by distance
Maintain set S of explored nodes: for u ∈ S, we know
the length d(u) of the shortest path from s to u.
Initialize S = { s }, d(s) = 0.
Repeatedly find shortest path to any node v ∉ S that
remains in S until the final edge e = (u, v)
S = S ∪ {v}, d(v) = d(u) + le
1
a c
2
3 1
S 1 1 e f
4
b d
1
Dijkstra’s Algorithm
Repeatedly find shortest path to any node v ∉ S that
remains in S until the final edge e = (u, v)
Minimize:
d’(v) = min { d(u) + le : u ∈ S, e = (u, v) ∈ E }
1
a c
2
3 1
S 1 1 e f
4
b d
1
Dijkstra’s Algorithm
Node d()
1
a c
2
1 3 1 e 1 f
4
b d
1
Dijkstra’s Algorithm
Node d()
a
1
c
2
a 0
1 3 1 e 1 f
4
b d
1
Dijkstra’s Algorithm
Node d()
a
1
c
2
a 0
1 3 1 e 1 f b 1
4
b d
1
Dijkstra’s Algorithm
Node d()
a
1
c
2
a 0
1 3 1 e 1 f b 1
4 c 1
b d
1
Dijkstra’s Algorithm
Node d()
a
1
c
2
a 0
1 3 1 e 1 f b 1
4 c 1
b d
1 d 2
Dijkstra’s Algorithm
Node d()
a
1
c
2
a 0
1 3 1 e 1 f b 1
4 c 1
b d
1 d 2
e 3
Dijkstra’s Algorithm
Node d()
a
1
c
2
a 0
1 3 1 e 1 f b 1
4 c 1
b d
1 d 2
e 3
f 4
Example
Second example on board
Dijkstra’s Algorithm: Implementation
Dijkstra’s Algorithm (G, s) {
" S = {s} // S is the set of explored nodes
" d(s) = 0" // d is the distance to the node from s
" while S ≠ V { // there are unexplored nodes
select a node v ∉ S with at least one edge from S to
minimize d’(v) = min { d(u) + le : u ∈ S, e = (u, v) ∈ E }
" " add v to S
" " d(v) = d’(v)
"}
How do we recover a path with length d(v)?
Dijkstra’s Algorithm: Implementation
Dijkstra’s Algorithm (G, s) {
" S = {s} // S is the set of explored nodes
" d(s) = 0" // d is the distance to the node from s
" while S ≠ V { // there are unexplored nodes
select a node v ∉ S with at least one edge from S to
minimize d’(v) = min { d(u) + le : u ∈ S, e = (u, v) ∈ E }
" " add v to S
" " d(v) = d’(v)
prev(v) = argmin { d(u) + le : u ∈ S, e = (u, v) ∈ E }
"}
Proof of correctness: start in
small groups, complete on board
Dijkstra’s Algorithm: Implementation
Dijkstra’s Algorithm (G, s) {
" S = {s} // S is the set of explored nodes
" d(s) = 0" // d is the distance to the node from s
" while S ≠ V { // there are unexplored nodes
select a node v ∉ S with at least one edge from S to
minimize d’(v) = min { d(u) + le : u ∈ S, e = (u, v) ∈ E }
" " add v to S
" " d(v) = d’(v)
prev(v) = argmin { d(u) + le : u ∈ S, e = (u, v) ∈ E }
"}
How do we implement this efficiently?
Dijkstra’s Algorithm (G, s) {
" S = {s} // S is the set of explored nodes
" d’(s) = 0" // explicitly maintain the d’ values
d’(v) = ∞ for all v ∉ S
while S != V { // there are unexplored nodes
Let v be the node that minimizes d’(v)
d(v) = d’(v)
for each edge (v, w) where v ∈ S, w ∉ S {
" " " d’(v) = min( d’(v), d(v) + l(v, w) )
"" }
"}
}
Data structure?
Running Time
With heap-based priority queue, running time
of Dijkstra’s algorithm is:
O(m) - traverse edges
O(n log n) - n extractMin operations
O(m log n) - m changeKey operations
Total: O(m log n)
