MST-Kruskal( G, w )

1 / 14

# MST-Kruskal( G, w ) - PowerPoint PPT Presentation

MST-Kruskal( G, w ). 1. A   // initially A is empty 2. for each vertex v  V [ G ] // line 2-3 takes O ( V ) time 3. do Create-Set( v ) // create set for each vertex 4. sort the edges of E by nondecreasing weight w

I am the owner, or an agent authorized to act on behalf of the owner, of the copyrighted work described.

## PowerPoint Slideshow about ' MST-Kruskal( G, w )' - devin-parks

An Image/Link below is provided (as is) to download presentation

Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author.While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server.

- - - - - - - - - - - - - - - - - - - - - - - - - - E N D - - - - - - - - - - - - - - - - - - - - - - - - - -
Presentation Transcript
MST-Kruskal(G, w)

1. A   // initially A is empty

2. for each vertex v  V[G] // line 2-3 takesO(V) time

3. do Create-Set(v) // create set for each vertex

4. sort the edges of E by nondecreasing weight w

5. for each edge (u,v)  E, in order bynondecreasing weight

6. do if Find-Set(u)  Find-Set(v) // u&v on different trees

7. then A  A  {(u,v)}

8. Union(u,v)

9. return A

Total running time is O(E lg E).

Lin/Foskey/Manocha

Analysis of Kruskal
• Lines 1-3 (initialization): O(V)
• Line 4 (sorting): O(E lg E)
• Lines 6-8 (set operations): O(E log E)
• Total:O(E log E)

Lin/Foskey/Manocha

Correctness of Kruskal
• Idea: Show that every edge added is a safe edge for A
• Assume (u, v) is next edge to be added to A.
• Will not create a cycle
• Let A’ denote the tree of the forest A that contains vertex u. Consider the cut (A’, V-A’).
• This cut respects A (why?)
• and (u, v) is the light edge across the cut (why?)
• Thus, by the MST Lemma, (u,v) is safe.

Lin/Foskey/Manocha

Intuition behind Prim’s Algorithm
• Consider the set of vertices S currently part of the tree, and its complement (V-S). We have a cut of the graph and the current set of tree edges A is respected by this cut.
• Which edge should we add next? Light edge!

Lin/Foskey/Manocha

Basics of Prim ’s Algorithm
• It works by adding leaves on at a time to the current tree.
• Start with the root vertex r (it can be any vertex).At any time, the subset of edges A forms a single tree. S = vertices of A.
• At each step, a light edge connecting a vertex in S to a vertex in V- S is added to the tree.
• The tree grows until it spans all the vertices in V.
• Implementation Issues:
• How to update the cut efficiently?
• How to determine the light edge quickly?

Lin/Foskey/Manocha

Implementation: Priority Queue
• Priority queue implemented using heap can support the following operations in O(lg n) time:
• Insert (Q, u, key): Insert u with the key value key in Q
• u = Extract_Min(Q): Extract the item with minimum key value in Q
• Decrease_Key(Q, u, new_key): Decrease the value of u’s key value to new_key
• All the vertices that are not in the S (the vertices of the edges in A) reside in a priority queue Q based on a key field. When the algorithm terminates, Q is empty. A = {(v, [v]): v  V - {r}}

Lin/Foskey/Manocha

Example: Prim’s Algorithm

Lin/Foskey/Manocha

MST-Prim(G, w, r)

1. Q  V[G]

2. for each vertex u  Q // initialization: O(V) time

3. do key[u]  

4. key[r]  0 // start at the root

5.[r]  NIL // set parent of r to be NIL

6. while Q   // until all vertices in MST

7. do u  Extract-Min(Q) // vertex with lightest edge

8. for each v  adj[u]

9. do if v  Qand w(u,v) < key[v]

10. then [v]  u

11. key[v]  w(u,v) // new lighter edge out of v

12. decrease_Key(Q, v, key[v])

Lin/Foskey/Manocha

Analysis of Prim
• Extracting the vertex from the queue: O(lg n)
• For each incident edge, decreasing the key of the neighboring vertex: O(lg n) wheren = |V|
• The other steps are constant time.
• The overall running time is, where e = |E|

T(n) = uV(lg n + deg(u)lg n) = uV (1+ deg(u))lg n

= lg n (n + 2e) = O((n + e)lg n)

Essentially same as Kruskal’s: O((n+e) lg n) time

Lin/Foskey/Manocha

Correctness of Prim
• Again, show that every edge added is a safe edge for A
• Assume (u, v) is next edge to be added to A.
• Consider the cut (A, V-A).
• This cut respects A (why?)
• and (u, v) is the light edge across the cut (why?)
• Thus, by the MST Lemma, (u,v) is safe.

Lin/Foskey/Manocha

Optimization Problems
• In which a set of choices must be made in order to arrive at an optimal (min/max) solution, subject to some constraints. (There may be several solutions to achieve an optimal value.)
• Two common techniques:
• Dynamic Programming (global)
• Greedy Algorithms (local)

Lin/Foskey/Manocha

Dynamic Programming
• Similar to divide-and-conquer, it breaks problems down into smaller problems that are solved recursively.
• In contrast to D&C, DP is applicable when the sub-problems are not independent, i.e. when sub-problems share sub-subproblems. It solves every sub-subproblem just once and saves the results in a table to avoid duplicated computation.

Lin/Foskey/Manocha

Elements of DP Algorithms
• Substructure:decompose problem into smaller sub-problems. Express the solution of the original problem in terms of solutions for smaller problems.
• Table-structure:Store the answers to the sub-problem in a table, because sub-problem solutions may be used many times.
• Bottom-up computation:combine solutions on smaller sub-problems to solve larger sub-problems, and eventually arrive at a solution to the complete problem.

Lin/Foskey/Manocha

Applicability to Optimization Problems
• Optimal sub-structure (principle of optimality): for the global problem to be solved optimally, each sub-problem should be solved optimally. This is often violated due to sub-problem overlaps. Often by being “less optimal” on one problem, we may make a big savings on another sub-problem.
• Small number of sub-problems:Many NP-hard problems can be formulated as DP problems, but these formulations are not efficient, because the number of sub-problems is exponentially large. Ideally, the number of sub-problems should be at most a polynomial number.

Lin/Foskey/Manocha