A Knowledge-Based System for Managing Hardware Dependency and Reproducibility in Quantum Machine Learning Workflows

In this paper, we emphasize the importance of having a Knowledge Graph-based system for managing hardware dependency and reproducibility in Quantum Machine Learning (QML) workflows. QML applications can benefit from integration with a Knowledge Graph (KG) to effectively organize, contextualize, and scale information for complex problem-solving using Quantum Computing (QC) techniques. QC-based Quantum Machine Learning (QML) is emerging as a field % based on Quantum Computing techniques for solving complex computational problems that are challenging for classical (i.e., non-quantum) systems. However, reproducibility and benchmarking against classical machine learning (CML) models remain challenging due to the varied and evolving quantum hardware and computational techniques, as well as the intricate nature of the datasets. By leveraging KGs to recognize and abstract beyond these variables, QML can use CML approaches more generally. This can extend to handling heterogeneous, interconnected datasets, particularly in domains that require spatiotemporal and relational modeling. Using a QML application as a use-case example from environmental analytics, we demonstrate that this approach enhances interpretability, scalability, and adaptability across various quantum computing hardware, quantum algorithms, and applications across different domains.