1BP: size(I) ≤ lin(I)

Dependencies:

  1. Bin packing and Knapsack
  2. Bin packing: configuration LP

Let $I$ be a 1BP instance. Then $\Size(I) \le \lin(I)$.

Proof

Let there be $m$ distinct items. Let $b_i$ be the number of items of the $i\Th$ type. Let $s_i$ be the size of the $i\Th$ type. Let $T$ be the configuration matrix. Let $N$ be the total number of configurations. Let $x^*$ be the optimal solution to the config LP.

\begin{align} \Size(I) &= \sum_{i=1}^m b_is_i \\ &\le \sum_{i=1}^m (Tx^*)_is_i \tag{by $i\Th$ constraint of config lp} \\ &= \sum_{j=1}^N x_j^* \left( \sum_{i=1}^m T[i, j]s_i \right) \\ &\le \sum_{j=1}^N x_j^* \tag{sum of sizes in $j\Th$ bin is at most 1} \\ &= \lin(I) \end{align}

Dependency for:

  1. 1BP: Karmarkar-Karp algorithm (broken)

Info:

Transitive dependencies:

  1. Bin packing and Knapsack
  2. Bin packing: configuration LP