Typically the way that optimization is you only need the subproblem solution from the previous row, so the rows before that you can ignore (and thrown away) by

Typically the way that optimization is you only need the subproblem solution from the previous row, so the rows before that you can ignore (and thrown away) by keeping only 2n entries. This assumes you don't need to reconstruct the path though.