Portfolio Asset Identification Using Graph Algorithms on a Quantum Annealer

Abstract

Our dual objectives are to explore how commercially available quantum hardware and algorithms can solve real world problems in finance, and then to compare quantum solutions to their classical counterparts. Specifically, we use the D-Wave quantum annealing computer (D-Wave 2000Q) to address the problem of asset correlation identification for financial portfolio management. Graphical models over a natural framework to represent asset correlations. Graphs also naturally map to the quantum annealing hardware architecture developed by D-Wave. We explore how graph algorithms can be implemented on the D-Wave 2000Q machine to cluster asset correlations in order to identify various financial portfolios. Numerical experiments are conducted using four quantum/classical algorithm pairs on four real world financial time series data sets spanning 10 years. For the specific algorithms and datasets selected, the quantum solution is competitive with (and sometimes better than) the classical one. However, quantum fails to scale beyond certain levels of data dimensionality. Our study focuses on comparison of solution quality not speedup. Our results suggest specific high-potential directions for future research.