Quantum Optimization Applications With Quark and Quapps: Bridging the Gap Between Application and Hardware [Focus: Quantum Software and its Engineering]
Solving discrete optimization problems is one of the most promising use cases on quantum computers with tremendous potential both in research and industry. We present two libraries quark and quapps which provide an easily usable, modular workflow to encode discrete optimization problems and interface them to both classical and quantum hardware. We provide a collection of already implemented optimization problems which can be easily adapted by the user and run on different solvers. From this we derive an additional instance data set that can directly be used to benchmark and test quantum hardware. In conjunction both libraries and the instance data set enable both expert and non-expert users to easily implement optimization problems, run them on classical and quantum machines and simplify benchmarking on quantum machines.
Solving discrete optimization problems is one of the most promising use cases on quantum computers with tremendous potential both in research and industry. We present two libraries quark and quapps which provide an easily usable, modular workflow to encode discrete optimization problems and interface them to both classical and quantum hardware. We provide a collection of already implemented optimization problems which can be easily adapted by the user and run on different solvers. From this we derive an additional instance data set that can directly be used to benchmark and test quantum hardware. In conjunction both libraries and the instance data set enable both expert and non-expert users to easily implement optimization problems, run them on classical and quantum machines and simplify benchmarking on quantum machines.
Quantum computing has the potential to surpass the capabilities of current classical computers when solving complex problems. Combinatorial optimization has emerged as one of the key target areas for quantum computers as problems found in this field play a critical role in many different industrial application sectors (e.g., enhancing manufacturing operations or improving decision processes). Currently, there are different types of high-performance optimization software (e.g., ILOG CPLEX and Gurobi) that support engineers and scientists in solving optimization problems using classical computers. In order to utilize quantum resources, users require domain-specific knowledge of quantum algorithms, SDKs and libraries, which can be a limiting factor for any practitioner who wants to integrate this technology into their workflows. Our goal is to add software infrastructure to a classical optimization package so that application developers can interface with quantum platforms readily when setting up their workflows. This paper presents a tool for the seamless utilization of quantum resources through a classical interface. Our approach consists of a Python library extension that provides a backend to facilitate access to multiple quantum providers. Our pipeline enables optimization software developers to experiment with quantum resources selectively and assess performance improvements of hybrid quantum-classical optimization solutions.
Quantum computing has the potential to surpass the capabilities of current classical computers when solving complex problems. Combinatorial optimization has emerged as a pivotal target area for quantum computers, as problems in this field are renowned for their complexity and resource-intensive nature. Moreover, these challenges play a critical role in various industrial sectors, including logistics, manufacturing, and finance. This thesis explores the integration of quantum computation into classical software tools as a means to potentially address combinatorial optimization problems more efficiently and effectively. This work introduces QPLEX, a Python software library that enables practitioners and researchers to implement the general mathematical formulation of a given combinatorial optimization problem once and execute it seamlessly on multiple quantum devices using various quantum algorithms. This software solution automatically adapts a general optimization model to the specific instructions utilized by the target quantum device’s SDK. It offers a versatile execution workflow capable of running gate-based hybrid quantum-classical algorithms for combinatorial optimization in a platform-agnostic manner. This approach reduces the programming overhead required for modeling and experimenting with combinatorial optimization solutions. Within this manuscript, we address and introduce the various aspects associated with the development of QPLEX in a clear and comprehensive manner. These aspects encompass the quantum algorithms and quantum hardware available in the library, along with QPLEX’s system design and implementation. Additionally, we provide a guide on how to use the library and conduct a thorough evaluation of the software solution within a specific use case as part of this thesis.
Quantum computing (QC) is anticipated to provide a speedup over classical HPC approaches for specific problems in optimization, simulation, and machine learning. With the advances in quantum computing toward practical applications, the need to analyze and compare different quantum solutions increases. While different low-level benchmarks for QC exist, these benchmarks do not provide sufficient insights into real-world application-level performance. We propose an application-centric benchmark method and the QUantum computing Application benchmaRK (QUARK) framework to foster the investigation and creation of application benchmarks for QC. This paper establishes three significant contributions: (1) it makes a case for application-level benchmarks and provides an in-depth "pen and paper" benchmark formulation of two reference problems: robot path and vehicle option optimization from the industrial domain; (2) it proposes the open-source QUARK framework for designing, implementing, executing, and analyzing benchmarks; (3) it provides multiple reference implementations for these two reference problems based on different known, and where needed, extended, classical and quantum algorithmic approaches and analyzes their performance on different types of infrastructures.
The advent of quantum algorithms has initiated a discourse on the potential for quantum speedups for optimization problems. However, several factors still hinder a practical realization of the potential benefits. These include the lack of advanced, error-free quantum hardware, the absence of accessible software stacks for seamless integration and interaction, and the lack of methods that allow us to leverage the theoretical advantages to real-world use cases. This paper works towards the creation of an accessible hybrid software stack for solving optimization problems, aiming to create a fundamental platform that can utilize quantum technologies to enhance the solving process. We introduce a novel approach that we call Hybrid Meta-Solving, which combines classical and quantum optimization techniques to create customizable and extensible hybrid solvers. We decompose mathematical problems into multiple sub-problems that can be solved by classical or quantum solvers, and propose techniques to semi-automatically build the best solver for a given problem. Implemented in our ProvideQ toolbox prototype, Meta-Solving provides interactive workflows for accessing quantum computing capabilities. Our evaluation demonstrates the applicability of Meta-Solving in industrial use cases. It shows that we can reuse state-of-the-art classical algorithms and extend them with quantum computing techniques. Our approach is designed to be at least as efficient as state-of-the-art classical techniques, while having the potential to outperform them if future advances in the quantum domain are made.
The advent of quantum algorithms has initiated a discourse on the potential for quantum speedups for optimization problems. However, several factors still hinder a practical realization of the potential benefits. These include the lack of advanced, error-free quantum hardware, the absence of accessible software stacks for seamless integration and interaction, and the lack of methods that allow us to leverage the theoretical advantages to real-world use cases. This paper works towards the creation of an accessible hybrid software stack for solving optimization problems, aiming to create a fundamental platform that can utilize quantum technologies to enhance the solving process. We introduce a novel approach that we call Hybrid Meta-Solving, which combines classical and quantum optimization techniques to create customizable and extensible hybrid solvers. We decompose mathematical problems into multiple sub-problems that can be solved by classical or quantum solvers, and propose techniques to semi-automatically build the best solver for a given problem. Implemented in our ProvideQ toolbox prototype, Meta-Solving provides interactive workflows for accessing quantum computing capabilities. Our evaluation demonstrates the applicability of Meta-Solving in industrial use cases. It shows that we can reuse state-of-the-art classical algorithms and extend them with quantum computing techniques. Our approach is designed to be at least as efficient as state-of-the-art classical techniques, while having the potential to outperform them if future advances in the quantum domain are made.
Benchmarking of quantum machine learning (QML) algorithms is challenging due to the complexity and variability of QML systems, e. g., regarding model ansatzes, data sets, training techniques, and hyper-parameter selection. The QUantum computing Application benchmaRK (QUARK) framework simplifies and standardizes benchmarking studies for quantum computing applications. Here, we propose several extensions of QUARK to include the ability to evaluate the training and deployment of quantum generative models. We describe the updated software architecture and illustrate its flexibility through several example applications: (1) We trained different quantum generative models using several circuit ansatzes, data sets, and data transformations. (2) We evaluated our models on GPU and real quantum hardware. (3) We assessed the generalization capabilities of our generative models using a broad set of metrics that capture, e. g., the novelty and validity of the generated data.