Given a connected directed acyclic graph $G(V, E)$, is there an algorithm for changing a spanning tree to a minimum spanning tree through a series of edge swaps?

We can use Prim's or Kruskal's algorithm to produce the minimum spanning tree from scratch, but I think we can find an asymptotically more efficient algorithm for finding a minimum spanning tree if we know for instance that we have some constant $k$ more edges than we have vertices, so $|E| = |V| + k$.


Are you familiar with Matroids? A Matroid is a construct M(G, I) with a ground set $G$ and independent set $I \subseteq 2^{G}$. Graphic matroids are defined with $G = E$, the edge set of the graph; and independence is acyclicity. That is, if a subset of edges is acyclic, it is in $I$.

Greedy basis algorithms are based on this edge-substitution procedure. In fact, this is how the proofs of correctness of Prim's, Kruskal's, and Dijkstra's algorithms are done. It is important to note that a Matroid allows us to deal with graphs in the same way that we deal with vector spaces. Hence, the term "basis" is quite accurate here. It refers to a maximally independent set.

So we will adapt a standard greedy basis algorithm. Start by sorting $G$, the ground set. Now for each $e \in G$, check if $e \in T$. If it is in $T$, discard $e$. Otherwise, check to see which edge $e^{\prime} \in T$ is incident to the same two vertices as $e$.


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.