An interesting problem

[Andy Armstrong]

On 3 Jan 2007, at 20:27, McGlinchy, Alistair wrote:

> Woops. The code got stripped from my previous email
>> I'd be interested in comments:

How big will the array be in practice? Exhaustively generating all  
the possible regions is going to tend towards N^4 so that's never  
going to be a very scalable solution.

Can you put any other constraints on the data or is it effectively a  
random scattering of +ve and -ve integers? Do you find to find the  
absolute best solution or is there a notion of 'good enough'?

-- 
Andy Armstrong, hexten.net