Skip to main content
10 events
when toggle format what by license comment
Aug 10, 2022 at 13:40 comment added AspiringMat @EmilJeřábek Yes, my initial comment was on just a single matching, but for the entire decomposition, you get an extra log factor in there as you mention, thanks for adding.
Aug 10, 2022 at 8:15 history edited YCor CC BY-SA 4.0
removed capitals from title
Aug 10, 2022 at 8:09 comment added Emil Jeřábek I meant $O(m\log d)$, not $O(m\log m)$.
Aug 10, 2022 at 8:03 comment added Emil Jeřábek But note that even in the power of 2 case, you need to recursively process both halves of the graph to get a decomposition. Thus, this will be time $O(m\log m)$ rather than $O(m)$. This time bound holds for general $d$ as well (finding perfect matchings in regular bipartite graphs can be done in time $O(m)$).
Aug 10, 2022 at 7:56 comment added Emil Jeřábek @AspiringMat Good idea. You can use it even if $d$ is not a power of $2$: if $d$ is even, find an Eulerian cycle as you describe, halving the degree; if $d$ is odd, find and remove a perfect matching first, making $d$ even. Then repeat.
Aug 10, 2022 at 6:47 comment added AspiringMat Specifically, find an Eulerian cycle and remove m/2 edges from B to A. You get a d/2 regular graph with m/2 edges. Repeat recursively until you have your matching
Aug 10, 2022 at 6:42 comment added AspiringMat If the graph is $d$ Regular for $d$ Power of 2, then there is a nice linear time algorithm (O($m$)) time
Aug 10, 2022 at 6:20 comment added Emil Jeřábek The standard proof of the result already gives a simple polynomial-time algorithm: find a perfect matching, remove it from the graph (which preserves its being regular bipartite), rinse and repeat until the graph is empty. I don’t know if there is anything more efficient.
S Aug 10, 2022 at 5:28 review First questions
Aug 10, 2022 at 5:53
S Aug 10, 2022 at 5:28 history asked CCC CC BY-SA 4.0