Research Mentor(s): Emily Wittrup
Authors: Rahul Hoque, Emily Wittrup, Kayvan Najarian
Although the Biomedical and Clinical Informatics Lab (BCIL) implements many classical machine learning (ML) models for classification purposes, the concept and potential applications of quantum ML (QML) including Quantum Classification, Quantum Feature Selection, and Quantum Neural Networks has not yet been explored. This study focuses on classification QML using the Python simulators PennyLane and Qiskit. These simulators provide functions and classes to perform quantum calculations that attempt to replicate outcomes of classical counterparts. When implementing the simulators on public benchmarking datasets, however, the results were unsatisfactory and inconclusive. While the simulators try to match classical results, they fall short in speed and accuracy. Running a quantum model on the iris dataset (4 features and 150 samples) produces a 78.9% accuracy score with a 90 second runtime while the classical model produces a 100% accuracy score with a 5 second runtime. Feature size also limits these simulators, since the amount of memory needed increases exponentially with the number of features in a dataset (2^n bits, n = number of features). Despite the functionality of these simulators, a usable QML model on a large scale dataset is unattainable until quantum computers are widespread.