It would be right if we have two edges (d,g) and (c,h).

]]>I was originally going to use integer knapsack (MP10), but convinced myself yesterday that since the G&J book says to use partition, that would work just as well.

I’ll do integer knapsack next week so I can fix this. Thanks!

]]>If you just set X_1=X_2 as the original partition instance you will also count numbers x such that x+x=2x=B/2 (i.e. it is allowed to sum some numbers multiple times to create one partition).

From what I can see the paper reduces from subset sum with repetition where you do not have this problem and can apply straight-forward binary search.

]]>