I don't think you can avoid a brute force approach
Although I agree that the general problem is likely NP complete, there are probably at least some things one might do,singly or in combination, to pick out more or less likely solutions (and hence improve the likelihood that the first of N solutions contains something close to the optimal solution - i.e. we limit he number of solutions tried to N since we don't have infinite computing resources). I don't have the time to really think these through (or relearn my linear algebra), but someone else might like to pick up the ball on these ideas:
Best, beth
Update: in response to JavaFan's "I don't see how that could help" note below, added clarification of reason for picking more or less likely solutions.
In reply to Re^3: "Divide" challenge
by ELISHEVA
in thread "Divide" challenge
by grizzley
For: | Use: | ||
& | & | ||
< | < | ||
> | > | ||
[ | [ | ||
] | ] |