Single-source shortest paths

• Two classic algorithms to solve single-source

shortest path problem
– Bellman-Ford algorithm
• A dynamic programming algorithm
• Works when some weights are negative
– Dijkstra’s algorithm
• A greedy algorithm
• Faster than Bellman-Ford
• Works when weights are all non-negative
 Bellman-Ford algorithm
Observation:
• If there is a negative cycle, there is no solution
– Add this cycle again can always produces a less weight path
• If there is no negative cycle, a shortest path has at most |V|-1 edges

Idea:
• Solve it using dynamic programming
• For all the paths have at most 0 edge, find all the shortest paths
• For all the paths have at most 1 edge, find all the shortest paths
• …
• For all the paths have at most |V|-1 edge, find all the shortest paths
 Bellman-Ford algorithm
Bellman-Ford(G, s)
for each v in G.V{ //Initialize 0-edge shortest paths
if(v==s) =0; else=//set the 0-edge shortest distance
from s to v
; //set the predecessor of v on the shortest path
} //bottom-up construct 0-to-(|V|-1)-edges shortest paths
Repeat |G.V|-1 times {
for each edge (u, v) in G.E{
if(+ ){
;
;
} //test negative cycle
}
for each edge (u, v) in G.E{
If (+ ) return false; // there is no solution
}
T(n)=O(VE)=O()
return true;
 Bellman-Ford algorithm
e.g.
20
0 ∞ ∞
10 1
1 2 3

What is the 0-edge shortest path from 1 to 1?

<> with path weight 0
What is the 0-edge shortest path from 1 to 2?
<>with path weight

What is the 0-edge shortest path from 1 to 3?

<>with path weight
 Bellman-Ford algorithm
e.g. ∞ > 0+20
20 𝑑1 , 3=20
0 ∞¿10 ∞¿ 20
1
10
2
1
3
∞ > 0+10
𝑑1 , 2=10
What is the at most 1-edge shortest path from 1 to 1? +1

<> with path weight 0 𝑑1 , 3 𝑢𝑛𝑐h𝑎𝑛𝑔𝑒𝑑

What is the at most 1-edge shortest path from 1 to 2?
<1, 2> with path weight 10

What is the at most 1-edge shortest path from 1 to 3?

<1, 3> with path weight 20

In Bellman-Ford, they are calculated by scan all edges once

 Bellman-Ford algorithm
e.g. 0 +20=20
20 𝑑1 , 3 𝑢𝑛𝑐h𝑎𝑛𝑔𝑒𝑑
0 ¿10
10 ¿ 11
20

1
10
2
1
3
10=0+ 10
𝑑1 , 2 𝑢𝑛𝑐h𝑎𝑛𝑔𝑒𝑑
What is the at most 2-edges shortest path from 1 to 1? +1

<> with path weight 0 𝑑1 , 3=11

What is the at most 2-edges shortest path from 1 to 2?
<1, 2> with path weight 10

What is the at most 2-edges shortest path from 1 to 3?

<1, 2, 3> with path weight 11

In Bellman-Ford, they are calculated by scan all edges once

 Bellman-Ford algorithm
All 0 edge shortest paths

5
∞ ∞
-2
2 3
6
-3
0
8
1 7
-4
7 2
4 5
∞ ∞
9
 Bellman-Ford algorithm
Calculate all at most 1 edge
shortest paths 5
∞¿ 6 ∞/
-2
2 3
6
-3
0 /0
8
1 7
-4
7 2
4 5
∞// ∞/
9
 Bellman-Ford algorithm
Calculate all at most 2 edges
shortest paths 5
6 ¿6 ∞ /11
/4
-2
2 3
6
-3
0 /0
8
1 7
-4
7 2
4 5
7 /7 ∞/2
9
 Bellman-Ford algorithm
Calculate all at most 3 edges
shortest paths 5
6 ¿¿62 4 /4
-2
2 3
6
-3
0 /0
8
1 7
-4
7 2
4 5
7 /7 2 /2
9
 Bellman-Ford algorithm
Calculate all at most 4 edges
shortest paths 5
2 ¿2 4 /4
-2
2 3
6
-3
0 /0
8
1 7
-4
7 2
4 5
7 /7 2 /-2
9
 Bellman-Ford algorithm
Final result:
5
2 4
-2
2 3
6
-3
0
8
1 7
-4
7 2
4 5
7 -2
9
What is the shortest path 1, 4, 3, 2, 5
from 1 to 5?
What is the shortest path
What is weight of this path? -2 from 1 to 2, 3, and 4?
 Count to infinity problem
• The Bellman–Ford algorithm does not prevent routing loops from happening and
suffers from the count-to-infinity problem.
• The core of the count-to-infinity problem is that if A tells B that it has a path
somewhere, there is no way for B to know if the path has B as a part of it. To see
the problem clearly, imagine a subnet connected like A–B–C–D–E–F, and let the
metric between the routers be "number of jumps".
• Now suppose that A is taken offline. In the vector-update-process B notices that the
route to A, which was distance 1, is down – B does not receive the vector update
from A. The problem is, B also gets an update from C, and C is still not aware of the
fact that A is down – so it tells B that A is only two jumps from C (C to B to A), which
is false. Since B doesn't know that the path from C to A is through itself (B), it
updates its table with the new value "B to A = 2 + 1".
• Later on, B forwards the update to C and due to the fact that A is reachable through
B (From C's point of view), C decides to update its table to "C to A = 3 + 1".
• This slowly propagates through the network until it reaches infinity (in which case
the algorithm corrects itself, due to the relaxation property of Bellman–Ford).
 All-pairs shortest paths
• All-pairs shortest path problem: given a
weighted, directed graph G=(V, E), for every
pair of vertices, find a shortest path.
• If there are negative weights, run Bellman-
Ford algorithm |V| times
– T(n)=
• If there are no negative weights, run Dijkstra’s
algorithm |V| times
– T(n)=
 All-pairs shortest paths
• There are other algorithms can do it more
efficient, such like Floyd-Warshall algorithm
• Floyd-Warshallalgorithm
– Negative weights may present, but no negative
cycle
– T(n)=
– A dynamic programming algorithm
 All-pairs shortest paths
• Floyd-Warshall(G)
Construct the shortest path matrix when there is no
intermediate vertex, D(0);
for(i=1 to |G.V|){
//D(i) is the shortest path matrix when the
intermediate //vertices could be:
Compute D(i) from D(i-1);
}
 All-pairs shortest paths
What are the weights of shortest paths with
no intermediate vertices, D(0)?

1 2
1 2 3 4
9
3 1
6 4 0∞ ∞ ∞
2
∞ 0 4
3 3 9 0 2
3 4 4 ∞ 6 ∞ 0
2
 All-pairs shortest paths
D(0)
1 2 3 4
1 0∞ ∞∞
2
1 2 ∞0 4
3 3 2 0 9
2
3 4 ∞6 ∞0
6 4
What are the weights of shortest paths with
intermediate vertex 1?
3 D(1)
4
1 2 3 4
9 1 0 ∞ ∞ ∞
2
∞ 0 ∞ 4
3 3 2 0 9
4 ∞ 6 ∞ 0
 All-pairs shortest paths
D(1)
1 2 3 4
1 0 ∞ ∞ ∞
1 2 2
∞ 0 ∞ 4
3 3 2 0 9
2
3 4 ∞ 6 ∞ 0
6 4
What are the weights of shortest paths with
intermediate vertices 1 and 2?
3 D(2)
4
1 2 3 4
9 1
∞ ∞ ∞
0
2
∞ 0 ∞ 4
3 3 2 0 6
4 ∞ 6 ∞ 0
 All-pairs shortest paths
D(2)
1 2 3 4
1 0 ∞ ∞ ∞
1 2 2 ∞ 4
∞ 0
2
3 3 2 0 6
3
6 4
4 ∞ 6 ∞ 0
What are the weights of shortest paths with
intermediate vertices 1, 2 and 3?
3 4 D(3)
1 2 3 4
9 1 0∞ ∞ ∞
2
∞ 0 ∞ 4
3 3 2 0 6
4 ∞ 6 ∞ 0
 All-pairs shortest paths
D(3)
1 2 3 4
1 0 ∞ ∞ ∞
1 2 2
∞ 0 ∞ 4
3 3 2 0 6
2
3 4 ∞ 6 ∞ 0
6 4
What are the weights of shortest paths with
intermediate vertices 1, 2, 3 and 4?
D(4)
3 4 1 2 3 4
1
9 0∞ ∞ ∞
2
∞ 0 ∞ 4
3 3 2 0 6
4 ∞ 6 ∞ 0
 All-pairs shortest paths
Add predecessor information to reconstruct a shortest path
If updated the predecessor i-j in D(k) is the same as the predecessor k-j in D(k-1)
D(0) D(1)
1 2 3 4 1 2 3 4
1 0/n∞ /𝑛∞ /𝑛∞ /𝑛 1 0/n ∞ /𝑛∞ /𝑛∞ /𝑛
2 ∞ /𝑛 0/n ∞ /𝑛 4/2 2 ∞ /𝑛 0/n ∞ /𝑛 4/2
3 3/3 2/ 3 0/n 9/3 3 3/3 2/ 3 0/n 9/3
4 ∞ /𝑛 6/4 ∞ /𝑛 0/n 4 ∞ /𝑛 6/4 ∞ /𝑛 0/n
D(2) D(3)
1 2 3 4 1 2 3 4
1 ∞ /𝑛∞ /𝑛∞ /𝑛
0/n 1 ∞ /𝑛∞ /𝑛∞ /𝑛
0/n
2 ∞ /𝑛 0/n ∞ /𝑛 4/2 2 ∞ /𝑛 0/n ∞ /𝑛 4/2
3 3/3 2/ 3 0/n 6/2 3 3/3 2/ 3 0/n 6/2
4 ∞ /𝑛 6/4 ∞ /𝑛 0/n 4 ∞ /𝑛 6/4 ∞ /𝑛 0/n
 All-pairs shortest paths

D(4)
1 2
1 2 3 4
1 0/n∞ /𝑛∞ /𝑛∞ /𝑛 3
2
2 ∞ /𝑛 0/n ∞ /𝑛 4/2 6 4
3 3/3 2/ 3 0/n 6/2
4 ∞ /𝑛 6/4 ∞ /𝑛 0/n
3 4

9
What is the shortest path 3, 2, 4
from 3 to 4?
What is weight of this path? 6
 Representation of the Input
We assume that the input is represented by a
weight matrix W= (wij)i,j in E that is defined by

wij= 0 if i=j
wij= w(i,j) if ij and (i,j) in E
wij=  if ij and (i,j) not in E

24
 The Floyd-Warshall Algorithm
Floyd-Warshall(W)
n = # of rows of W;
D(0) = W;
for k = 1 to n do
for i = 1 to n do
for j = 1 to n do
dij(k) = min{dij(k-1) , dik(k-1) + dkj(k-1)};
ef;
ef;
ef;
return D(n);

25
 Time and Space Requirements
The running time is obviously O(n3). Since n=V,
hence O(V3)

However, in this version, the space requirements

are high. One can reduce the space from O(V3)
to O(V2) by using a single array d.

26