Parameterized Pulse Encoding for Quantum Machine Learning

Chemical property predictions using quantum machine learning (QML) lie at the intersection of machine learning, quantum computing, and computational chemistry. QML models often use parameterized quantum circuits (PQCs) that abstract gate-level quantum operations but offer limited flexibility in adjustable parameters. To enhance QML model performance and generalizability, incorporating pulse-level operations, which lie below gate-level in the quantum computing stack, is crucial. These operations enable more adjustable parameters and finer control, potentially improving robustness against noise on real quantum hardware.

We hypothesize that integrating pulse-level operations into QML models will improve their ability to predict chemical properties like bond dissociation energies through regression tasks. The goal is to develop hybrid quantum circuit-pulse learning algorithms to optimize both pulse- and gate-level parameters. We will use quantum-mechanically calculated chemical data to train and test the models, investigating how training data size affects generalization error.

Additionally, the project will evaluate whether hybrid circuit-pulse learning offers advantages over standard quantum circuit learning in accurately predicting chemical properties. By addressing PQC limitations and exploring pulse-level operations, this work aims to advance QML’s application in computational chemistry, paving the way for more powerful and adaptable models.

Faculty Supervisor:

Viki Kumar Prasad

Student:

Partner:

Université de Bordeaux

Discipline:

Physics

Sector:

Quantum Science; Artificial Intelligence

University:

University of Calgary

Program: