Download as txt, pdf, or txt
Download as txt, pdf, or txt
You are on page 1of 2

This algorithm finds a packing for I using at most (1 + ffl)OPT(I) + 1 buckets,

which is the bound established in theorem 28.4. All steps are at most linear in n,
except step 2, which is O(n log n), as it basically amounts to sorting J . The fact
that step 3 is linear in n was established in the previous algorithm, but note that
while f (ffi; m) is independent of n, it is exponential in

ffi

and m and thus

ffl

. Therefore, this approximation scheme polynomial but

not fully polynomial. Karmarkar and Karp came up with a an algorithm for RBP(ffi;
m) that is polynomial in

ffi

and

m as well as n. Instead of using integer linear programming, they relaxed the


condition of the x

's being integers, which results in a regular linear programming problem. The
optimal

solution to the linear programming problem may assign fractional values to the x

's, but

it was shown that by appropriate rounding of the values, a solution close to


optimal for the RBP is obtained. Since polynomial algorithms (for example, the
ellipsoid algorithm) exist for linear programming, the approximation scheme for BP
based on this result is fully polynomial.

110

You might also like