Rostock 2019 – wissenschaftliches Programm

Bereiche | Tage | Auswahl | Suche | Aktualisierungen | Downloads | Hilfe

Q: Fachverband Quantenoptik und Photonik

Q 3: Quantum Information (Quantum Computing) I

Q 3.6: Vortrag

Montag, 11. März 2019, 12:00–12:15, S HS 001 Chemie

Classifying images with hierarchical quantum neural networks — •Andrea Skolik and Martin Leib — Volkswagen Data:Lab, Ungererstr. 69, 80805 München

With noisy intermediate scale quantum devices now being available, algorithms that don't require fully error corrected quantum computers are receiving increased attention. One particularly promising area in this respect is quantum machine learning, where noise is even conjectured to be beneficial, based on findings in the classical counterpart. In this work, we investigate parametrized quantum circuits based on an image classification task. An image is classified according to the measurement of a designated qubit after the variational circuit acted on an initial state that is generated based on information of the respective image. Starting with random values we optimize the gate parameters in a classical external learning loop such that the training data gets classified correctly. We perform extensive numerical studies to investigate the capabilities of this nascent technique, and show that even quantum circuits with a modest size of parameters can achieve up to 95% classification accuracy on a set of handwritten digits.