Cute :)

[denny]

Biological imaging algorithm also solves Sudoku puzzles

Comments

[kelemvor.livejournal.com]

*Brain hurts from the hard science*
Reminds me of the DNA-sequencing software that also filters astonishing quanities of spam.

[simont]

... whereas my problem with that article is that the science isn't hard enough; I'd have liked to see a link to the paper or the program or at least some technical detail about exactly what the algorithm is and how you apply it to Sudoku. Googling around the subject gets me a vague idea of the shape of the thing, only with some important details missing.

[dennyd.livejournal.com]

http://www.cecm.sfu.ca/~veit/references/elser03.pdf isn't gory enough for you regarding the algorithm? I can't find anything about application to games design though, which is what I assume caught your interest.

[simont]

Indeed. The key paragraph in that paper, for me, is at the end of the third page:

The fixed point property of the difference map makes no reference to the detailed forms of the maps. These maps are key to the second property: the attractive nature of the fixed points. In an iterative solution method a fixed point is useless unless it is attractive; moreover, the greater the basin of attraction, the more effective is the method.

In other words, in order to apply this algorithm to Sudoku, you need a pair of maps, each of which converts an arbitrary square of numbers into one which satisfies one of the Sudoku constraints, in such a way that applying the maps in alternation is likely to converge. There's plenty of scope there for choosing entirely the wrong kind of map, and it isn't the least bit clear to me how you tell in advance whether a given map is going to work.