Using Statistical Analysis to Fine-Tune the Results of Knapsack-Based Computational Platform Benchmarking

Abstract

In previous papers, we composed an algorithmic foundation for computational platform benchmarking of well-known exact algorithms for the Knapsack Problem. We suggested using the run time of these algorithms with fixed inputs as the performance estimates. We then derived a single performance estimate, equally impacted by each of the algorithms. Although this approach makes for a reasonable general-purpose benchmark, equalizing the impact of different algorithms is not completely legitimate, as they have different processing requirements. In this paper, we perform an in-depth analysis of algorithm operational requirements and try to fine-tune the integral estimates to describe special-purpose (e.g. data compression or encipherment/decipherment) platforms more accurately.