Heterogeneous quantum computing for satellite constellation optimization: solving the weightedk-clique problem

Abstract

NP-hard optimization problems scale very rapidly with problem size, becoming unsolvable with brute force methods, even with supercomputing resources. Typically, such problems have been approximated with heuristics. However, these methods still take a long time and are not guaranteed to find an optimal solution. Quantum computing offers the possibility of producing significant speed-up and improved solution quality. Current quantum annealing (QA) devices are designed to solve difficult optimization problems, but they are limited by hardware size and qubit connectivity restrictions. We present a novel heterogeneous computing stack that combines QA and classical machine learning, allowing the use of QA on problems larger than the hardware limits of the quantum device. These results represent experiments on a real-world problem represented by the weighted k-clique problem. Through this experiment, we provide insight into the state of quantum machine learning.

Quantum-Enhanced Grid of the Future: A Primer

“…Quantum computers can solve discrete combinatorial optimization problems using the properties of quantum adiabatic (QA) evolution [51]. A machine learning heterogeneous computing stack is proposed in [52] that combines QA and classical machine learning, allowing the VOLUME XX, 2017 use of QA on problems more substantial than the hardware limits of the quantum device.…”

Hybrid Annealing Method Based on subQUBO Model Extraction With Multiple Solution Instances

“…Several methods for extracting subQUBO models from an original QUBO model have been proposed [9], [10], [11]. The simplest subQUBO model extraction method randomly chooses variables from the original QUBO model [9].…”

Understanding the Continuous-Time Dynamics of Phase-Transition Nano-Oscillator-Based Ising Hamiltonian Solver

“…OMBINATORIAL optimization problems have immense real-world application including financial portfolio optimization, bio-informatics, drug discovery, cryptography, operations research, resource allocation, satellite-based target tracking, trajectory and route planning [1]- [4]. However, many of these problems belong to the non-deterministic polynomial time (NP)-hard or NP-complete complexity class, indicating an exponential increase in the resources required to solve the problem as the problem size increases.…”

“…However, many of these problems belong to the non-deterministic polynomial time (NP)-hard or NP-complete complexity class, indicating an exponential increase in the resources required to solve the problem as the problem size increases. Interestingly, many such problems can be reformulated into another physics problemfinding the ground state of an Ising model [4]- [6]. The Ising Hamiltonian describes the energy of a spin system with discrete binary spins states and a symmetric coupling matrix and is given by = − ∑ problem size of few hundred spins.…”

“…1. The real-world problems is first mapped onto the ground state search problem of an Ising model [4]- [6]. As shown in the Fig.…”