Design and Analysis of Algorithms
Single Source Shortest Path Problem
(Bellman-Ford Algorithm)
Dr. D. P. Acharjya
Professor, SCOPE
SJT Annex – 201 E
3/9/2024 Dr. D. P. Acharjya 1
Bellman Ford Algorithm
Let G = (V, E) be a weighted directed graph with
source s and weight function w : E → R .
The Bellman-Ford algorithm solves the single-source
shortest-path problem in which edge weights may be
negative.
The algorithm returns a Boolean value indicating
whether or not there is a negative-weight cycle that is
reachable from the source.
The algorithm fails if there is a negative weight cycle.
Otherwise, the algorithm produces the shortest path.
3/9/2024 Dr. D. P. Acharjya 2
Bellman-Ford (G, w, s)
1. INITIALIZE – SINGLE SOURCE (G, s)
2. for i ← 1 to n – 1 = |V| - 1
3. do for each edge (u, v) E
4. do RELAX (u, v, w)
5. for each edge (u, v) E
6. do if d[v] > d[u] + w(u, v)
7. then return FALSE
8. Return TRUE
3/9/2024 Dr. D. P. Acharjya 3
Initialize-Single Source d (G, s)
1. Initialize – Single Source (G, s)
2. for each vertex v V[G]
3. do d[v] ← ∞
4. π[v] ← NIL
5. d[s] = 0
After initialization
π[v] = NIL for all v V (Predecessor)
d[s] = 0, and d[v] = ∞ for v V -{s}
(Distance of source is 0)
3/9/2024 Dr. D. P. Acharjya 4
Relax (u,v,w)
1. RELAX (u, v, w)
2. if d[v] > d[u] + w(u, v)
3. then d[v] ← d[u] + w(u, v)
4. π[v] ← u d[v] > d[u] + w(u, v)
After Relax
Relaxing test whether we can
improve the shortest path to v
going through u and so, update
d[v] and π[v].
d[v] < d[u] + w(u, v)
3/9/2024 Dr. D. P. Acharjya 5
Numerical Illustration
Consider the following graph
and compute the shortest
distance to all the vertices from
the source vertex.
After Initialization
d[s] = 0, and d[t] = ∞, d[x] = ∞,
d[y] = ∞, d[z] = ∞
π[s] = NIL, π[t] = NIL, π[x] =
NIL, π[y] = NIL, π[z] = NIL
3/9/2024 Dr. D. P. Acharjya 6
Continued …
For i = 1
Relax edges (s, t) and (s, y)
Relax (s, t, w = 6)
d[t] = ∞, d[s] + 6 = 6
d[t] > d[s] + 6 = 6
d[t] = 6
Relax (s, y, w = 7)
d[y] = ∞, d[s] + 7 = 7
d[y] > d[s] + 7 = 7
d[y] = 7
3/9/2024 Dr. D. P. Acharjya 7
Continued …
For i = 2
Relax edges (t, x), (t, z), (t, y),
(y, z), and (y, x)
Relax (t, x, w = 5)
d[x] = ∞, d[t] + 5 = 6+5=11
d[x] > d[t] + 5 = 11
d[x] = 11
Relax (t, z, w = – 4)
d[z] = ∞, d[t] + (– 4) = 6 – 4 = 2
d[z] > d[t] – 4 = 2
d[z] = 2
3/9/2024 Dr. D. P. Acharjya 88
Continued …
Relax (t, y, w = 8)
d[y] = 7, d[t] + 8 = 6+8=14
d[y] > d[t] + 8 (False)
d[y] = 7
Relax (y, z, w = 9)
d[z] = 2, d[y] + 9 = 7 + 9 = 16
d[z] > d[y] + 9 (False)
d[z] = 2
Relax (y, x, w = – 3)
d[x] = 11, d[y] + (–3) = 7–3 = 4
d[x] > d[y] –3
d[x] = 4
3/9/2024 Dr. D. P. Acharjya 99
Continued …
For i = 3
Relax edges (x, t) and (z, x)
Relax (x, t, w = – 2)
d[t] = 6, d[x] + (–2) = 4–2 = 2
d[t] > d[x] –2
d[t] = 2
Relax (z, x, w = 7)
d[x] = 4, d[z] + 7 = 2+7 = 9
d[x] > d[z] + 7 (False)
For i = 4
Relax edges (t, x), (t, z), (t, y)
3/9/2024 Dr. D. P. Acharjya 10
10
Continued …
Relax (t, x, w = 5)
d[x] = 4, d[t] + 5 = 2 + 5 = 7
d[x] > d[t] + 5 (False)
d[x] = 4
Relax (t, y, w = 8)
d[y] = 7, d[t] + 8 = 2+8 = 10
d[y] > d[t] + 8 (False)
d[y] = 7
Relax (t, z, w = – 4)
d[z] = 2, d[t] + (– 4) = 2 – 4 = – 2
d[z] > d[t] – 4
d[z] = – 2 (Process Terminates)
3/9/2024 Dr. D. P. Acharjya 11
11
