Performance of the Extended Ising Machine for the Quadratic Knapsack Problem

TL;DR

This paper evaluates the extended Ising machine's ability to solve the quadratic knapsack problem, demonstrating its advantages over traditional Ising models and other solvers in handling constrained quadratic optimization.

Contribution

The paper introduces the extended Ising machine (EIM) for quadratic knapsack problems and compares its performance with existing methods, showing its improved effectiveness.

Findings

EIM outperforms conventional Ising models in solving QKP.

EIM shows superior results compared to commercial exact and heuristic solvers.

The approach effectively handles constraints in quadratic optimization problems.

Abstract

The extended Ising machine (EIM) enhances conventional Ising models, which handle only binary quadratic forms by allowing constraints through real-valued dependent variables. We address the quadratic knapsack problem (QKP), hard to solve using Ising machines when formulated as a quadratic unconstrained binary optimization (QUBO). We demonstrated the EIM's superiority by comparing it with the conventional Ising model-based approach, a commercial exact solver, and a state-of-the-art heuristic solver for QKP.