Mathematics Stack Exchange is a question and answer site for people studying math at any level and professionals in related fields. Join them; it only takes a minute:

Sign up
Here's how it works:
  1. Anybody can ask a question
  2. Anybody can answer
  3. The best answers are voted up and rise to the top

There is a net of $N$ almost-straight paths on an aerial map. Some of them intersect with another. At the points of intersection there are possibly a "mis-tie", which is expressed as a number, which is zero if the paths are properly "tied" at the intersection. All those numbers are incorporated into a $N\times N$ matrix $A$, where the cell $A_{ij}$ contains a value of a "mis-tie" at the point of intersection of paths $i$ and $j$. A "mis-tie" is signed, so $A_{ij}=-A{ji}$ (that is, if $i$ is "above" $j$ then $j$ is "below" $i$). There is "nothing" in cells $A_{ij}$ if paths $i$ and $j$ do not intersect.

I'd like to minimize (in some sense) the mis-ties by shifting the paths "up" and "down", that is, by adding|subtracting a number from an entire row|column of a matrix $A$ (not affecting the "nothing" cells).

How can I calculate the optimal shifts to all of the paths to minimize the mis-ties? For example, to minimize $\sum_{ij: A_{ij}\neq 0} (A_{ij})^2$ ?

share|cite|improve this question
up vote 1 down vote accepted

You want to minimize

$$ \sum_{ij}c_{ij}(A_{ij}+s_i-s_j)^2\;, $$

where $c_{ij}\in\{0,1\}$, $c_{ij}=c_{ji}$ and $A_{ij}=-A_{ji}$. Setting the derivative with respect to $s_k$ to zero yields

$$ \sum_jc_{kj}(A_{kj}+s_k-s_j)-\sum_ic_{ik}(A_{ik}+s_i-s_k)=2\sum_jc_{kj}(A_{kj}+s_k-s_j)=0\;. $$

This is a linear system of equations for the $s_k$ that you can solve. If all the $c_{kj}$ were $1$, you could perform the sum over $j$ and calculate the $s_k$ explicitly, but for general $c_{kj}$ I don't see a way to make special use of the binary character of the matrix.

share|cite|improve this answer
What do you think about solving it iteratively, taking averages along a row|column as an initial $s^1_i$, recalculating $A_{ij} \rightarrow A^1_{ij}$ then taking averages again as $s^2_i$, recalculating $A^1_{ij} \rightarrow A^2_{ij}$ etc. - would it converge to optimal? – mbaitoff Dec 7 '12 at 4:20

Your Answer


By posting your answer, you agree to the privacy policy and terms of service.

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