CS341 Lec09 Annotated Feb27

CS 341: Algorithms
Lec 09: Dijkstra’s Algorithms
Armin Jamshidpey Yang Lu
David R. Cheriton School of Computer Science, University of Waterloo
Winter 2024
A. Jamshidpey Y. Lu (CS, UW) Lec 09: Dijkstra’s Algorithms Winter 2024 1 / 15

Preliminaries
vk
w(vk−1 , vk )
G = (V, E) a directed graph
with a weight function:
w(v2 , v3 )
w:E→R
v2
The weight of path w(v1 , v2 )
P = ⟨v0 ,P
. . . , vk ⟩ is:
w(P ) = ki=1 w(vi−1 , vi ) w(v0 , v1 ) v1
v0

Preliminaries
Padlet (True/False)
Shortest path exists in any directed weighted graph.
https://padlet.com/arminjamshidpey/CS341

Preliminaries
Padlet (True/False)
Shortest path exists in any directed weighted graph.
https://padlet.com/arminjamshidpey/CS341
Assumption: G has no negative-weight cycles
The shortest path weight from u to v:

(
P
min{w(P ) : u ⇝ v} if there exists a path from u to v
δ(u, v) =
∞ otherwise

Preliminaries
Single-Source Shortest Path Problem
Input: G = (V, E), w : E → R and a source s ∈ V
Output: A shortest path from s to each v ∈ V
True/False: If ⟨v0 , v1 , . . . , vk ⟩ is a shortest path from v0 to vk , then ⟨v0 , v1 , . . . , vi ⟩

is a shortest path from v0 to vi , for any 0 ≤ i ≤ k.
vi
vk
v0
Padlet: https://padlet.com/arminjamshidpey/CS341
Overview
Explanation of Dijkstra’s algorithm
Pseudocode of the algorithm
An example
E. Dijkstra (1930-2002)
Complexity analysis Turing Award (1972)
Proof of correctness “Computer Science is no more about

computers than astronomy is about
telescopes.”
-E. Dijkstra
Dijkstra’s algorithm: Explanation
Dijkstra’s algorithm is a greedy algorithm

v
Input: A weighted directed graph with
non-negative edge weights
For all vertices, maintain quantities

▶
s
d[v]: a shortest-path estimate from s to v
▶ π[v]: predecessor in the path (a vertex or NIL)

Dijkstra’s algorithm is a greedy algorithm π[v]
v
Input: A weighted directed graph with d[v]
non-negative edge weights
For all vertices, maintain quantities

▶ d[v]: a shortest-path estimate from s to v s
▶ π[v]: predecessor in the path (a vertex or NIL)

Initialize C = ∅, repeat the following until

C =V: d[v] v
▶ Add u ∈ V − C with smallest d-value to C w(u, v)
▶ Update d-values of vertices v with (u, v) ∈ E: u
d[v] ← min{d[v], d[u] + w(u, v)}
▶ Update π[v] if d[v] is changed s d[u]

Which Abstract Data Type (ADT) should we use for vertices?

12 15
23 21 17 16
53 41 22

5
Cost of operations of a
binary min-heap
12 15 (of size n):
Insert: O(log n)
23 21 17 16 Extract-Min: O(log n)
Update-Key: O(log n)
53 41 22

Dijkstra’s algorithm: Pseudocode
Dijkstra(G, w, s)
1 for each vertex v ∈ V [G]
2 d[v] ← ∞
3 π[v] ← NIL
4 d[s] ← 0

Dijkstra(G, w, s)
2 d[v] ← ∞
3 π[v] ← NIL
4 d[s] ← 0
5 C←∅
6 Q ← V [G]

Dijkstra(G, w, s)
2 d[v] ← ∞
3 π[v] ← NIL
4 d[s] ← 0
5 C←∅
6 Q ← V [G]
7 while Q ̸= ∅
8 u ← Extract-Min(Q)
9 C ← C ∪ {u}

Dijkstra(G, w, s)
2 d[v] ← ∞
3 π[v] ← NIL
4 d[s] ← 0
5 C←∅
6 Q ← V [G]
7 while Q ̸= ∅
9 C ← C ∪ {u}
10 for each vertex v ∈ Adj[u]
11 if d[v] > d[u] + w(u, v)
12 d[v] ← d[u] + w(u, v)
13 π[v] ← u

Dijkstra’s algorithm: Example
Dijkstra(G, w, s) a
1 for each vertex v ∈ V [G] 10 1
2 d[v] ← ∞
3 π[v] ← NIL 2
s 3 b
4 d[s] ← 0
5 C←∅ 9
6 Q ← V [G] 5 7 4 6
7 while Q ̸= ∅
8 u ← Extract-Min(Q) c
d
9 C ← C ∪ {u} 2
11 if d[v] > d[u] + w(u, v)
12 d[v] ← d[u] + w(u, v) Figure: An example from CLRS
13 π[v] ← u
Dijkstra(G, w, s) a
∞ NIL
1 for each vertex v ∈ V [G] 10 1
2 d[v] ← ∞
3 π[v] ← NIL 2
s 3 b
4 d[s] ← 0 0 NIL ∞ NIL
5 C←∅ 9
6 Q ← V [G] 5 7 4 6
7 while Q ̸= ∅
8 u ← Extract-Min(Q) c
d
9 C ← C ∪ {u} ∞ NIL 2 ∞ NIL
11 if d[v] > d[u] + w(u, v)
13 π[v] ← u
Dijkstra(G, w, s) a C
∞ NIL Q
10 1
2 d[v] ← ∞
3 π[v] ← NIL 2
s 3 b
4 d[s] ← 0 0 NIL ∞ NIL
5 C←∅
9
6 Q ← V [G] 5 4 6
7
7 while Q ̸= ∅
d c
9 C ← C ∪ {u} ∞ NIL ∞ NIL
2
11 if d[v] > d[u] + w(u, v)
13 π[v] ← u
∞ NIL Q
10 1
2 d[v] ← ∞
3 π[v] ← NIL 2
s 3 b
4 d[s] ← 0 0 NIL ∞ NIL
5 C←∅
9
6 Q ← V [G] 5 4 6
7
7 while Q ̸= ∅
d c
9 C ← C ∪ {u} ∞ NIL ∞ NIL
2
11 if d[v] > d[u] + w(u, v)
13 π[v] ← u
∞ NIL Q
10 1
2 d[v] ← ∞
3 π[v] ← NIL 2
s 3 b
4 d[s] ← 0 0 NIL ∞ NIL
5 C←∅
9
6 Q ← V [G] 5 4 6
7
7 while Q ̸= ∅
d c
9 C ← C ∪ {u} ∞ NIL ∞ NIL
2
11 if d[v] > d[u] + w(u, v)
13 π[v] ← u
10s Q
10 1
2 d[v] ← ∞
3 π[v] ← NIL 2
s 3 b
4 d[s] ← 0 0 NIL ∞ NIL
5 C←∅
9
6 Q ← V [G] 5 4 6
7
7 while Q ̸= ∅
d c
9 C ← C ∪ {u} 5 s ∞ NIL
2
11 if d[v] > d[u] + w(u, v)
13 π[v] ← u
10s Q
10 1
2 d[v] ← ∞
3 π[v] ← NIL 2
s 3 b
4 d[s] ← 0 0 NIL ∞ NIL
5 C←∅
9
6 Q ← V [G] 5 4 6
7
7 while Q ̸= ∅
d c
9 C ← C ∪ {u} 5 s ∞ NIL
2
11 if d[v] > d[u] + w(u, v)
13 π[v] ← u
10s Q
10 1
2 d[v] ← ∞
3 π[v] ← NIL 2
s 3 b
4 d[s] ← 0 0 NIL ∞ NIL
5 C←∅
9
6 Q ← V [G] 5 4 6
7
7 while Q ̸= ∅
d c
9 C ← C ∪ {u} 5 s ∞ NIL
2
11 if d[v] > d[u] + w(u, v)
13 π[v] ← u
10 S Q
10 1
2 d[v] ← ∞
3 π[v] ← NIL 2
s 3 b
4 d[s] ← 0 0 NIL ∞ NIL
5 C←∅
9
6 Q ← V [G] 5 4 6
7
7 while Q ̸= ∅
d c
9 C ← C ∪ {u} 5 s ∞ NIL
2
11 if d[v] > d[u] + w(u, v)
13 π[v] ← u
8 d Q
10 1
2 d[v] ← ∞
3 π[v] ← NIL 2
s 3 b
4 d[s] ← 0 0 NIL 14 d
5 C←∅
9
6 Q ← V [G] 5 4 6
7
7 while Q ̸= ∅
d c
9 C ← C ∪ {u} 5 s 7
2 d
11 if d[v] > d[u] + w(u, v)
13 π[v] ← u
8 d Q
10 1
2 d[v] ← ∞
3 π[v] ← NIL 2
s 3 b
4 d[s] ← 0 0 NIL 14 d
5 C←∅
9
6 Q ← V [G] 5 4 6
7
7 while Q ̸= ∅
d c
9 C ← C ∪ {u} 5 s 7
2 d
11 if d[v] > d[u] + w(u, v)
13 π[v] ← u
8 d Q
10 1
2 d[v] ← ∞
3 π[v] ← NIL 2
s 3 b
4 d[s] ← 0 0 NIL 14 d
5 C←∅
9
6 Q ← V [G] 5 4 6
7
7 while Q ̸= ∅
d c
9 C ← C ∪ {u} 5 s 7
2 d
11 if d[v] > d[u] + w(u, v)
13 π[v] ← u
8 d Q
10 1
2 d[v] ← ∞
3 π[v] ← NIL 2
s 3 b
4 d[s] ← 0 0 NIL 14 d
5 C←∅
9
6 Q ← V [G] 5 4 6
7
7 while Q ̸= ∅
d c
9 C ← C ∪ {u} 5 s 7
2 d
11 if d[v] > d[u] + w(u, v)
13 π[v] ← u
8 d Q
10 1
2 d[v] ← ∞
3 π[v] ← NIL 2
s 3 b
4 d[s] ← 0 0 NIL 13 c
5 C←∅
9
6 Q ← V [G] 5 4 6
7
7 while Q ̸= ∅
d c
9 C ← C ∪ {u} 5 s 7
2 d
11 if d[v] > d[u] + w(u, v)
13 π[v] ← u
8 d Q
10 1
2 d[v] ← ∞
3 π[v] ← NIL 2
s 3 b
4 d[s] ← 0 0 NIL 13 c
5 C←∅
9
6 Q ← V [G] 5 4 6
7
7 while Q ̸= ∅
d c
9 C ← C ∪ {u} 5 s 7
2 d
11 if d[v] > d[u] + w(u, v)
13 π[v] ← u
Dijkstra(G, w, s) C
a
1 for each vertex v ∈ V [G] 8 d Q
2 d[v] ← ∞ 10 1
3 π[v] ← NIL
2
4 d[s] ← 0 s 3 b
5 C←∅ 0 NIL 13 c
6 Q ← V [G] 9
5 4 6
7 while Q ̸= ∅ 7
9 C ← C ∪ {u} d c
10 for each vertex v ∈ Adj[u] 5 s 2 7 d
11 if d[v] > d[u] + w(u, v)
13 π[v] ← u
Dijkstra(G, w, s) C
a
2 d[v] ← ∞ 10 1
3 π[v] ← NIL
2
4 d[s] ← 0 s 3 b
5 C←∅ 0 NIL 13 c
6 Q ← V [G] 9
5 4 6
7 while Q ̸= ∅ 7
9 C ← C ∪ {u} d c
11 if d[v] > d[u] + w(u, v)
13 π[v] ← u
Dijkstra(G, w, s) C
a
2 d[v] ← ∞ 10 1
3 π[v] ← NIL
2
4 d[s] ← 0 s 3 b
5 C←∅ 0 NIL 9 a
6 Q ← V [G] 9
5 4 6
7 while Q ̸= ∅ 7
9 C ← C ∪ {u} d c
11 if d[v] > d[u] + w(u, v)
13 π[v] ← u
Dijkstra(G, w, s) C
a
2 d[v] ← ∞ 10 1
3 π[v] ← NIL
2
4 d[s] ← 0 s 3 b
5 C←∅ 0 NIL 9 a
6 Q ← V [G] 9
5 4 6
7 while Q ̸= ∅ 7
9 C ← C ∪ {u} d c
11 if d[v] > d[u] + w(u, v)
13 π[v] ← u
Dijkstra(G, w, s) C
a
2 d[v] ← ∞ 10 1
3 π[v] ← NIL
2
4 d[s] ← 0 s 3 b
0 9 a
5 C←∅ NIL
6 Q ← V [G] 9
5 7 4 6
7 while Q ̸= ∅
9 C ← C ∪ {u} d c
11 if d[v] > d[u] + w(u, v)
13 π[v] ← u
Dijkstra(G, w, s) C
a
2 d[v] ← ∞ 10 1
3 π[v] ← NIL
2
4 d[s] ← 0 s 3 b
0 9 a
5 C←∅ NIL
6 Q ← V [G] 9
5 7 4 6
7 while Q ̸= ∅
9 C ← C ∪ {u} d c
11 if d[v] > d[u] + w(u, v)
13 π[v] ← u
Dijkstra(G, w, s) C
a
2 d[v] ← ∞ 10 1
3 π[v] ← NIL
2
4 d[s] ← 0 s 3 b
0 9 a
5 C←∅ NIL
6 Q ← V [G] 9
5 7 4 6
7 while Q ̸= ∅
9 C ← C ∪ {u} d c
11 if d[v] > d[u] + w(u, v)
13 π[v] ← u
Dijkstra(G, w, s) C
a
2 d[v] ← ∞ 10 1
3 π[v] ← NIL
2
4 d[s] ← 0 s 3 b
0 9 a
5 C←∅ NIL
6 Q ← V [G] 9
5 7 4 6
7 while Q ̸= ∅
9 C ← C ∪ {u} d c
11 if d[v] > d[u] + w(u, v)
13 π[v] ← u
Dijkstra’s algorithm: Complexity analysis
Dijkstra(G, w, s) line Array Imp. Heap Imp.
1
1 for each vertex v ∈ V [G] 2
2 d[v] ← ∞ 3 O(|V |)
3 π[v] ← NIL 4
4 d[s] ← 0 5
5 C←∅ 6 O(|V |)
6 Q ← V [G] 7 O(|V |) iterations
8 O(|V |2 )
7 while Q ̸= ∅
9
10 O(|E|) iterations
9 C ← C ∪ {u} 11
10 for each vertex v ∈ Adj[u] 12 O(|E|)
11 if d[v] > d[u] + w(u, v) 13
12 d[v] ← d[u] + w(u, v) Tot. O(|V |2 )
13 π[v] ← u Note: The effects of number of iterations on lines 7
and 10, are already considered in the cost of lines 8-9
A. Jamshidpey Y. Lu (CS, UW) Lec 09: Dijkstra’sand 11-13
Algorithms Winter 2024 11 / 15
Dijkstra’s algorithm: Complexity analysis
Dijkstra(G, w, s) line Array Imp. Heap Imp.
1
2
2 d[v] ← ∞ 3 O(|V |) O(|V |)
3 π[v] ← NIL 4
4 d[s] ← 0 5
5 C←∅ 6 O(|V |) O(|V |)
6 Q ← V [G] 7 O(|V |) iter. O(|V |) iter.
7 while Q ̸= ∅ 8 O(|V |2 ) O(|V | log |V |)
9
10 O(|E|) iter. O(|E|) iter.
9 C ← C ∪ {u}
11
10 for each vertex v ∈ Adj[u] 12 O(|E|) O(|E| log |V |)
11 if d[v] > d[u] + w(u, v) 13

12 d[v] ← d[u] + w(u, v) Tot. O(|V |2 ) O (|V | + |E|) log |V |
13 π[v] ← u Note: The of number of iterations on lines 7 and 10,
are already considered in the cost of lines 8-9 and
A. Jamshidpey Y. Lu (CS, UW) Lec 09: Dijkstra’s11-13
Algorithms Winter 2024 11 / 15
Dijkstra’s algorithm: Proof of Correctness
Claim: For each vertex v ∈ V , we have d[v] = δ(s, v) at the time when v is added
to set C. To prove the claim, we follow these steps:

1 Proof by contradiction: Assume the claim is not correct and u ∈ V is the

first vertex for which d[u] ̸= δ(s, u) when it is added to C.


2 Time t: Beginning of the iteration in which u is added to C.


2 Time t: Beginning of the iteration in which u is added to C.
3 Use a shortest path P , from s to u (needs justification)

C u
s
P
C covers a part of P at time t

C u
s
4 On P , find y (with predecessor x), the first vertex in P
V − C.
x y

C u
s
V − C.
x y
5 At time t, we have d[y] = δ(s, y). (needs Justification)

C u
s
V − C.
x y

6 Fact 1: d[y] = δ(s, y) ≤ δ(s, u) ≤ d[u]

C u
s
V − C.
x y

6 Fact 1: d[y] = δ(s, y) ≤ δ(s, u) ≤ d[u]
7 Fact 2: at time t the algorithm chose u. Hence
d[u] ≤ d[y]

C u
s
V − C.
x y

6 Fact 1: d[y] = δ(s, y) ≤ δ(s, u) ≤ d[u]
7 Fact 2: at time t the algorithm chose u. Hence
d[u] ≤ d[y]
8 From equations facts 1 and 2 we conclude: d[y] = δ(s, y) = δ(s, u) = d[u]

3 Use a shortest path P , from s to u
(needs justification)
Proof: C u
s
P
C covers a part of P at time t

5 At time t, we have d[y] = δ(s, y).
Proof:
C u
s
P
x y

CS341 Lec09 Annotated Feb27

Uploaded by

Document Information

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

CS341 Lec09 Annotated Feb27

Uploaded by

Copyright:

Available Formats

CS 341: Algorithms

Lec 09: Dijkstra’s Algorithms

Armin Jamshidpey Yang Lu

David R. Cheriton School of Computer Science, University of Waterloo

A. Jamshidpey Y. Lu (CS, UW) Lec 09: Dijkstra’s Algorithms Winter 2024 1 / 15

A. Jamshidpey Y. Lu (CS, UW) Lec 09: Dijkstra’s Algorithms Winter 2024 2 / 15

A. Jamshidpey Y. Lu (CS, UW) Lec 09: Dijkstra’s Algorithms Winter 2024 3 / 15

Assumption: G has no negative-weight cycles

The shortest path weight from u to v:

A. Jamshidpey Y. Lu (CS, UW) Lec 09: Dijkstra’s Algorithms Winter 2024 3 / 15

Output: A shortest path from s to each v ∈ V

True/False: If ⟨v0 , v1 , . . . , vk ⟩ is a shortest path from v0 to vk , then ⟨v0 , v1 , . . . , vi ⟩

Explanation of Dijkstra’s algorithm

Pseudocode of the algorithm

Proof of correctness “Computer Science is no more about

Dijkstra’s algorithm is a greedy algorithm

For all vertices, maintain quantities

A. Jamshidpey Y. Lu (CS, UW) Lec 09: Dijkstra’s Algorithms Winter 2024 6 / 15

Dijkstra’s algorithm is a greedy algorithm π[v]

For all vertices, maintain quantities

A. Jamshidpey Y. Lu (CS, UW) Lec 09: Dijkstra’s Algorithms Winter 2024 6 / 15

Initialize C = ∅, repeat the following until

A. Jamshidpey Y. Lu (CS, UW) Lec 09: Dijkstra’s Algorithms Winter 2024 7 / 15

Which Abstract Data Type (ADT) should we use for vertices?

A. Jamshidpey Y. Lu (CS, UW) Lec 09: Dijkstra’s Algorithms Winter 2024 8 / 15

Which Abstract Data Type (ADT) should we use for vertices?

A. Jamshidpey Y. Lu (CS, UW) Lec 09: Dijkstra’s Algorithms Winter 2024 8 / 15

Which Abstract Data Type (ADT) should we use for vertices?

A. Jamshidpey Y. Lu (CS, UW) Lec 09: Dijkstra’s Algorithms Winter 2024 8 / 15

A. Jamshidpey Y. Lu (CS, UW) Lec 09: Dijkstra’s Algorithms Winter 2024 9 / 15

A. Jamshidpey Y. Lu (CS, UW) Lec 09: Dijkstra’s Algorithms Winter 2024 9 / 15

A. Jamshidpey Y. Lu (CS, UW) Lec 09: Dijkstra’s Algorithms Winter 2024 9 / 15

A. Jamshidpey Y. Lu (CS, UW) Lec 09: Dijkstra’s Algorithms Winter 2024 9 / 15

A. Jamshidpey Y. Lu (CS, UW) Lec 09: Dijkstra’s Algorithms Winter 2024 12 / 15

1 Proof by contradiction: Assume the claim is not correct and u ∈ V is the

A. Jamshidpey Y. Lu (CS, UW) Lec 09: Dijkstra’s Algorithms Winter 2024 12 / 15

1 Proof by contradiction: Assume the claim is not correct and u ∈ V is the

A. Jamshidpey Y. Lu (CS, UW) Lec 09: Dijkstra’s Algorithms Winter 2024 12 / 15

1 Proof by contradiction: Assume the claim is not correct and u ∈ V is the

3 Use a shortest path P , from s to u (needs justification)

C covers a part of P at time t

A. Jamshidpey Y. Lu (CS, UW) Lec 09: Dijkstra’s Algorithms Winter 2024 13 / 15

5 At time t, we have d[y] = δ(s, y). (needs Justification)

A. Jamshidpey Y. Lu (CS, UW) Lec 09: Dijkstra’s Algorithms Winter 2024 13 / 15

5 At time t, we have d[y] = δ(s, y). (needs Justification)

A. Jamshidpey Y. Lu (CS, UW) Lec 09: Dijkstra’s Algorithms Winter 2024 13 / 15

5 At time t, we have d[y] = δ(s, y). (needs Justification)

A. Jamshidpey Y. Lu (CS, UW) Lec 09: Dijkstra’s Algorithms Winter 2024 13 / 15

5 At time t, we have d[y] = δ(s, y). (needs Justification)

8 From equations facts 1 and 2 we conclude: d[y] = δ(s, y) = δ(s, u) = d[u]

A. Jamshidpey Y. Lu (CS, UW) Lec 09: Dijkstra’s Algorithms Winter 2024 13 / 15

C covers a part of P at time t

A. Jamshidpey Y. Lu (CS, UW) Lec 09: Dijkstra’s Algorithms Winter 2024 14 / 15

A. Jamshidpey Y. Lu (CS, UW) Lec 09: Dijkstra’s Algorithms Winter 2024 15 / 15

You might also like