Avoiding patterns in matrices via a small number of changes

Let ${\cal A}=\{A_1,\ldots, A_r\}$ be a partition of a set $\{1,\ldots,m\}\times\{1,\ldots, n\}$ into $r$ nonempty subsets, and $A=(a_{ij})$ be an $m\times n$ matrix. We say that $A$ has a pattern ${\cal A}$ provided that $a_{ij}=a_{i'j'}$ if and only if $(i,j),(i',j')\in A_t$ for some $t\in\{1,\ldots,r\}$. In this note we study the following function $f$ defined on the set of all $m\times n$ matrices $M$ with $s$ distinct entries: $f(M; {\cal A})$ is the smallest number of positions where the entries of $M$ need to be changed such that the resulting matrix does not have any submatrix with pattern ${\cal A}$. We give an asymptotically tight value for $$ f(m,n; s, {\cal A}) = \max\{f(M; {\cal A}): M \mbox{ is an } m\times n\mbox{ matrix with at most } s \mbox{ distinct entries}\} . $$