Kruskal's Algorithm

Examples

• 1

  Optimal, disjoint-set based implementation

  edit move example delete share 1

  Improvements requested by Nick Larsen♦:

  • This example is completely unclear, incomplete, or has severe formatting problems. It is unlikely to be salvageable through editing and should be removed. - jul 24 at 15:15

    Comments:

    • This example references something "above", but they are not in any particular order. It makes it look like it's been completely copy/pasted from elsewhere. - Nick Larsen

  Edit Example Delete Example

  1

  We can do two things to improve the simple disjoint-set subalgorithms:

  1. Path compression heuristic: findSet does not need to ever handle a tree with height bigger than 2. If it ends up iterating such a tree, it can link the lower nodes directly to the root, optimizing future traversals;

     subalgo findSet(v: a node):
         if v.parent != v
             v.parent = findSet(v.parent)
         return v.parent
     

  2. Height-based merging heuristic: for each node, store the height of its subtree. When merging, make the taller tree the parent of the smaller one, thus not increasing anyone's height.

     subalgo unionSet(u, v: nodes):
         vRoot = findSet(v)
         uRoot = findSet(u)
     
         if vRoot == uRoot:
             return
     
         if vRoot.height < uRoot.height:
             vRoot.parent = uRoot
         else if vRoot.height > uRoot.height:
             uRoot.parent = vRoot
         else:
             uRoot.parent = vRoot
             uRoot.height =  uRoot.height + 1
     

  This leads to O(alpha(n)) time for each operation, where alpha is the inverse of the fast-growing Ackermann function, thus it is very slow growing, and can be considered O(1) for practical purposes.

  This makes the entire Kruskal's algorithm O(m log m + m) = O(m log m), because of the initial sorting.

  edited 2 days ago

  
• 1

  Simple, disjoint-set based implementation

  edit move example delete share

  1

  The above forest methodology is actually a disjoint-set data structure, which involves three main operations:

  subalgo makeSet(v: a node):
      v.parent = v    <- make a new tree rooted at v
      
  
  subalgo findSet(v: a node):
      if v.parent == v:
          return v
      return findSet(v.parent)
  
  subalgo unionSet(v, u: nodes):
      vRoot = findSet(v)
      uRoot = findSet(u)
  
      uRoot.parent = vRoot
  
  algorithm kruskalMST''(G: a graph):
      sort G's edges by their value
      for each node n in G:
          makeSet(n)
      for each edge e in G:
          if findSet(e.first) != findSet(e.second):
              unionSet(e.first, e.second)
  

  This naive implementation leads to O(n log n) time for managing the disjoint-set data structure, leading to O(m*n log n) time for the entire Kruskal's algorithm.

  created Jul 22 at 16:45

  
• 0

  Simple, high level implementation

  edit move example delete share

  0

  Sort the edges by value and add each one to the MST in sorted order, if it doesn't create a cycle.

  algorithm kruskalMST(G: a graph)
      sort G's edges by their value
      MST = an empty graph
      for each edge e in G:
          if adding e to MST does not create a cycle:
              add e to MST
  
      return MST
  

  created Jul 22 at 16:45

  
• 0

  Simple, more detailed implementation

  edit move example delete share

  0

  In order to efficiently handle cycle detection, we consider each node as part of a tree. When adding an edge, we check if its two component nodes are part of distinct trees. Initially, each node makes up a one-node tree.

  algorithm kruskalMST'(G: a graph)
      sort G's edges by their value
      MST = a forest of trees, initially each tree is a node in the graph
      for each edge e in G:
          if the root of the tree that e.first belongs to is not the same 
          as the root of the tree that e.second belongs to:
              connect one of the roots to the other, thus merging two trees
      
      return MST, which now a single-tree forest
  

  created Jul 22 at 16:45