A Quantum Annealer for Subset Feature Selection and the Classification of Hyperspectral Images

Hyperspectral images (HSIs) showing objects belonging to several distinct target classes are characterized by dozens of spectral bands being available. However, some these redundant and/or noisy, and hence, selecting highly informative trustworthy for each class is a vital step classification saving internal storage space; then the selected termed band subset. We use mutual information (MI)-based method select subset given two additional binary quantum classifiers, namely boost (Qboost) plus (Qboost-Plus) classifier, classify two-label dataset pose both our MI-based selection problem classifiers as quadratic unconstrained optimization (QUBO) problem. Such solvable with help conventional techniques. QUBO an NP-hard global problem, it worthwhile applying annealer. Thus, we adapted HSI be run on D-Wave After class, employ In addition, provide novel multilabel classifier exploiting error-encoding output code when using classifiers. As real-world in Earth observation, used well-known AVIRIS Indian Pine, north-western Indiana, USA. can demonstrate that annealer selects Pine HSI. also prove generate correct two- their high accuracy such having been produced classifiers-and even better instances.