Skip to main content
11 events
when toggle format what by license comment
Jan 10, 2016 at 14:39 answer added user85022 timeline score: 2
Jan 8, 2016 at 10:17 vote accept user85022
Jan 8, 2016 at 2:08 comment added Tony Huynh Given $M_n$, you can define a bipartite graph $G_n$ with bipartition $A$ and $B$ where $a_i$ is adjacent to $b_j$ if $M_{i,j}=0$. Your question then becomes what is the minimum number of complete bipartite subgraphs needed to partition the edges of $G_n$. You can do this for any square $0$-$1$ matrix, so if you are interested in matrices besides $M_n$, then Googling biclique covering number will turn up a lot of useful links.
Jan 8, 2016 at 0:52 history edited user85022 CC BY-SA 3.0
example with M_{15}
Jan 7, 2016 at 23:49 answer added Douglas Zare timeline score: 10
Jan 7, 2016 at 23:33 comment added user85022 @DouglasZare That would be great to have such a divide-and-conquer method! I do have a lower bound, but I must admit it is quite bad. For $M_7$, if we look at the entries (1,2), (2,3), (4,7) and (7,1), none can be in the same submatrix of another, so we have at least 4 different submatrices. But even for greater $n$, I am not able to find more than 4 entries verifying this property.
Jan 7, 2016 at 23:13 comment added Douglas Zare My guess is that some sort of divide-and-conquer gives a cover with $O(\log n)$ submatrices. What lower bounds do you have?
S Jan 7, 2016 at 23:11 history suggested user85022 CC BY-SA 3.0
better example
Jan 7, 2016 at 23:00 review Suggested edits
S Jan 7, 2016 at 23:11
Jan 7, 2016 at 21:46 review First posts
Jan 7, 2016 at 22:09
Jan 7, 2016 at 21:42 history asked user85022 CC BY-SA 3.0