WalidMasri's blog

By WalidMasri, history, 9 years ago, In English

Hello everyone!

Having found the SCC of a directed graph G, how can I contract each SCC to a single node?

Thanks.

  • Vote: I like it
  • +3
  • Vote: I do not like it

| Write comment?
»
9 years ago, # |
  Vote: I like it +6 Vote: I do not like it

Suppose you know C[v] — which component contains node v.
Now you scan through edges of the original graph, let current one be x->y.
Then you should add an edge C[x]->C[y] in compressed graph if C[x] is not equal to C[y].

  • »
    »
    9 years ago, # ^ |
      Vote: I like it 0 Vote: I do not like it

    Yes, and you should also remember to mark which edges have already been added so you don't double count any edge. You can use a boolean array or an edge set

    • »
      »
      »
      9 years ago, # ^ |
        Vote: I like it 0 Vote: I do not like it

      What's so bad about keeping a multi-edge in DAG? :P

      • »
        »
        »
        »
        9 years ago, # ^ |
          Vote: I like it 0 Vote: I do not like it

        If we have to do some traversal in the compressed graph we are going to waste some operations while scanning the adjacency list of each vertex :P

    • »
      »
      »
      9 years ago, # ^ |
        Vote: I like it 0 Vote: I do not like it

      I'd rather use disjoint sets. It doesn't require too much memory and I think it's faster than an edge set :D and are VERY easy to code

      • »
        »
        »
        »
        9 years ago, # ^ |
          Vote: I like it 0 Vote: I do not like it

        How can i use disjoint sets for this purpose? I have never seen this application, i usually use DSU for Kruskal's only :P

        • »
          »
          »
          »
          »
          9 years ago, # ^ |
          Rev. 2   Vote: I like it 0 Vote: I do not like it

          Nvm, bullsh*t in rev1

          • »
            »
            »
            »
            »
            »
            9 years ago, # ^ |
              Vote: I like it 0 Vote: I do not like it

            Awesome! It's really simple, you are right, this is the easiest/most efficient option.

          • »
            »
            »
            »
            »
            »
            9 years ago, # ^ |
              Vote: I like it +15 Vote: I do not like it

            But if you have edges X->Y, Y->Z, X->Z, won't you miss one of them this way?

            • »
              »
              »
              »
              »
              »
              »
              9 years ago, # ^ |
                Vote: I like it 0 Vote: I do not like it

              omg you are right, I remember I used disjoint sets but it wasn't this way. Nvm thanks!

      • »
        »
        »
        »
        9 years ago, # ^ |
        Rev. 2   Vote: I like it 0 Vote: I do not like it

        Other valid option would be using a hash function so you can check in O(1) whether the edge exists or not! It should be easy to come up with a function that wont generate collisions since amount of edges is usually small :)

        EDIT : never used unordered set, but this may actually be the implementation of what i said above

        • »
          »
          »
          »
          »
          9 years ago, # ^ |
          Rev. 2   Vote: I like it 0 Vote: I do not like it

          UPD: I am sorry, there is nothing correct here :(

          • »
            »
            »
            »
            »
            »
            9 years ago, # ^ |
              Vote: I like it 0 Vote: I do not like it

            Won't it be more efficient supposing there won't be collisions? I'm assuming you will use map data structure to associate integers to each edge

            • »
              »
              »
              »
              »
              »
              »
              9 years ago, # ^ |
                Vote: I like it 0 Vote: I do not like it

              I am sorry, I have no idea what I was thinking at that moment but I suppose it was something wrong :(