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

QI: Fachverband Quanteninformation

QI 16: Quantum Software

QI 16.7: Vortrag

Donnerstag, 12. März 2026, 11:45–12:00, BEY/0245

Encoding Numerical Data for Generative Quantum Machine Learning — •Michael Krebsbach¹, Hagen-Henrik Kowalski², Florentin Reiter¹, Ali Abedi², and Thomas Wellens¹ — ¹Fraunhofer IAF, Tullastraße 72, 79108 Freiburg — ²Bundesdruckerei GmbH, Kommandantenstraße 18, 10969 Berlin

Generative quantum machine learning has the potential to model probability distributions that are out of reach for their classical counterparts. Due to the binary nature of samples drawn from a quantum computer, many of the generative models described in the literature focus on binary data. The transition from binary to real-world data, which is typically numerical, necessitates an additional encoding step that can obscure structure in the data and hinder effective learning.

In this talk, we present our investigation into binary encodings and their effect on the training process of generative quantum machine learning algorithms. We identify situations in which the conventional approach is limited, and propose strategies that circumvent these limitations at essentially no additional cost. We test these strategies on a range of datasets and provide numerical evidence that they provide an average-case improvement over the conventional approach.

Keywords: Generative Quantum Machine Learning; Quantum circuit Born Machine (QCBM); Gray Codes; Synthetic Data Generation