Why not based on cost-effective greedy? Obviously right ah?
Think on the line, assume now selected as new items: Maybe \ (A \) cost little or simply \ (0 \) , but its volume is very small, the overall impact is also negligible; perhaps \ (B \) greater cost but it's very, very large volume, simply imagine infinitely good, so overall it is extremely cost-effective approaches \ (B \) a significant influence. At this time, the former is clearly better. Since local optima are guaranteed not to talk about the global optimum it?
This requires the introduction of the half.