Quantum Eigenfaces

I co-authored a research paper that introduces a quantum implementation of a seminal machine learning algorithm for data classification. We had the privilege of presenting our findings at the IEEE International Conference on Quantum Computing and Engineering (QCE) in September 2023.

Authors

Armando Bellante, William Bonvini, Stefano Vanerio, Stefano Zanero.

Abstract

We propose a quantum machine learning algorithm for data classification, inspired by the seminal computer vision approach of eigenfaces for face recognition. The algorithm enhances nearest neighbor/centroid classifiers with concepts from principal component analysis, enabling the automatic detection of outliers and finding use in anomaly detection domains beyond face recognition. Assuming classical input data, we formalize how to implement the algorithm using a quantum random access memory and state-of-the-art quantum linear algebra, discussing the complexity of performing the classification algorithm on a fault-tolerant quantum device. The asymptotic time complexity analysis shows that the quantum classification algorithm can be more efficient than its classical counterpart. We showcase an application of this algorithm for face recognition and image classification datasets with anomalies, obtaining promising results for the running time parameters. This work contributes to the growing field of quantum machine learning applications, and the algorithm’s simplicity makes it easily adoptable by future quantum machine learning practitioners.