A $m \times n$ table is filled with signs $"+"$ and $"-"$. A table is called irreducible if one cannot reduce it to the table filled with $"+"$, applying the following operations (as many times as one wishes). $a)$ It is allowed to flip all the signs in a row or in a column. Prove that an irreducible table contains an irreducible $2\times 2$ sub table. $b)$ It is allowed to flip all the signs in a row or in a column or on a diagonal (corner cells are diagonals of length $1$). Prove that an irreducible table contains an irreducible $4\times 4$ sub table.
Problem
Source: ToT 2003 SA-7
Tags: induction, linear algebra, matrix, combinatorics unsolved, combinatorics
03.07.2011 15:26
(I tried my best) a) Suppose every 2x2 sub table is reducible. Then note that every row is either exactly the same or every cell is different to the row above. So it is easy to perform operations such that everything becomes "+". b) First we show that every $3\times n$ table is reducible. We just have to show we can change any cell without changing the rest. We can change every cell in the first column, so removing the first column and we have a $3\times (n-1)$ table. Do this a few more times. Hence every cell can be changed independently. Now we show that every $m\times n$ table that contains only reducible 4x4 board is reducible, by induction. The base case is $m=n=4$ which is obvious. Now suppose it is true for $(m-1) \times n$, then we show for $m\times n$. First we make the first $m-1$ rows "+". Note that in a reducible 4x4 board, the number of "+" in the 8 cells on the edges is always even. Thus in the last row of the $m\times n$ board, the $n-2$ edges are all "+" or "-". Thus we can change the entire board into all "+". By induction it is true for all boards and we are done.
03.07.2011 17:30
The exact same idea works for part b:
28.01.2015 11:08
Lokitos, I think I understand your reasoning, but if I'm correctly understanding it, what you prove in part a is that a bijection between legal combinations and tables containing only reducible 2x2 subtables exists, and what you don't prove is exactly what you need to prove, namely that those tables are reducible. There exist numerous bijections of the kind you proved to exist, but their existence as far as I'm aware doesn't prove that there exists a bijection which sends each table to a combination that reduces it.
28.01.2015 12:27
TT Fall 2003, Senior A/ Problem 7. See attachment for solution(official).
Attachments:
TT2003F_SAsolutions.pdf (96kb)