Hybrid Quantum-Classical Processing Workflows in Modular Supercomputing Architectures for Data-Intensive Earth Observation Applications

The project is a proof-of-concept study harnessing the power of advanced high performance and quantum computing systems for big data machine learning applications in remote sensing data analysis. The obtained workflows aim at implementing scalability and addressing the growing rate of higher resolution and shorter revisit time information acquisition in the context of Earth Observation. The initial phase of the project consists in benchmarking single computing technologies. Advancements in the usage of the D-Wave Advantage Quantum Annealer JUPSI for machine learning have been reached. An algorithm called Quantum Multiclass Support Vector Machine has been developed, which consists of a single optimization step performed bythe quantum annealer for performing multiclass classification. A key aspect is that only a subset of training examples is used by the quantum annealing algorithm. The obtained quantum annealer solutions are then combined to maximize the accuracy on the whole training set.

Reference: doi: 10.1109/IGARSS46834.2022.9883963