2

Consider an adjacency matrix $A$ with elements $[A]_{ij}=1$ if one can reach state $i$ from state $j$ in one timestep, and $0$ otherwise. The matrix $[A^k]_{ij}$ represents the number of paths that lead from state $j$ to $i$ in $k$ timesteps. Derive an algorithm that will find the minimum number of step to get from state $j$ to state $i$.

Here is my attempt:

The algorithm takes as input:

  • $A$ - Adjacency matrix
  • $j$ and $i$ - the vertices for which we want to find the minimum path inbetween

Then it calculates in a while loop for which value of $k$, $[A^k]_{ij}\neq 0$ and returns that value of $k$. This way we get the minimum value of $k$ for which there exists a path.

Here is the pseudocode:

(Is this the correct way to do it and would there be a faster way?)

shortestPath(A,j,i):
      if [A]_{ij}=1:
           return 1
      else:
           f=0; k=1
           while f=0:
               f = [A^k]_{ij}
               k = k + 1
           return k                ​

​

Slim Shady
  • 155
  • 4

2 Answers2

2

This is indeed a correct approach.

It can be improved further by using binary search on $k$, but you have to be careful with it since you want to multiply a small number of matrices (and computing $A^r$ for an arbitrary $r$, can take a lot of time using a simple multiplication algorithm)

As a side note, if you prefer using non matrix-based algorithms, then you can use BFS which will work in linear time

nir shahar
  • 11,753
  • 3
  • 17
  • 35
1

Use the Floyd-Warshall algorithm, it was designed for exactly this kind of task.

vonbrand
  • 14,204
  • 3
  • 42
  • 52