Quantum and other physics-inspired optimization algorithms, however, are often limited to handling problems that fit the Quadratic Unconstrained Binary Optimization (QUBO) framework or one of its equivalent formulations, such as the Ising Spin model. Moreover, from an infrastructure standpoint, access to these resources is usually granted through each vendor's specific interface. Together, these aspects pose a barrier for incoming practitioners to experiment with new methods and compare platforms within a common practical scenario [4].
With this in mind, QUBO.jl[5, 6] was developed to bridge the gap between emerging optimization methods and critical industrial applications. In this talk, we would like to show how QUBO.jl makes Quantum Optimization accessible to Process System Engineering practitioners as it integrates a heterogeneous hardware and software landscape under a standard interface, providing users with a smooth modeling experience. By leveraging JuMP's extension capabilities, QUBO.jl makes it simple to access quantum and other novel devices as if they were regular optimization solvers. This makes it the ideal environment to try and explore potential applications.
QUBO.jl extends the JuMP (Julia Mathematical Programming) algebraic modeling language to provide a universal interface for multiple types of quantum and other physics-inspired solution methods in the same way that regular solvers are made available. This was made possible by the development of the QUBODrivers.jl [7] package, which provides a set of tools for rapidly integrating heterogeneous QUBO solver interfaces into the JuMP ecosystem. This common interface exempts interested experts from adapting to the specificities of each platform, especially if one considers that many PSE specialists already use JuMP daily.
On the other hand, ToQUBO.jl [8] provides an extensive set of tools for automatically reformulating mathematical models into Quadratic Unconstrained Binary Optimization (QUBO). It allows users to send more complex models they already use and are familiar with directly to quantum devices. ToQUBO performs all necessary transformations, including variable encoding, constraint penalization, and polynomial degree reduction, and maps results back to the original model, offering users a transparent interface to quantum optimization.
Its interface was integrated with DisjunctiveProgramming.jl [9], another JuMP-compatible package, to provide support for Generalized Disjunctive Programming (GDP) models [10]. This was a fundamental step in making existing models from the PSE literature automatically translatable to QUBO.
References
[1] https://arxiv.org/pdf/2310.03011
[2] https://arxiv.org/abs/2312.02279
[3] https://aiche.onlinelibrary.wiley.com/doi/10.1002/aic.17651
[4] https://arxiv.org/pdf/2409.06919v1
[5] https://github.com/JuliaQUBO/QUBO.jl
[6] https://arxiv.org/abs/2307.02577
[7] https://github.com/JuliaQUBO/QUBODrivers.jl
[8] https://github.com/JuliaQUBO/ToQUBO.jl
[9] https://github.com/hdavid16/DisjunctiveProgramming.jl
[10] https://www.sciencedirect.com/science/article/pii/B9780443288241505731