Recent Advances in Optimizing Quantum Data Embedding for Machine Learning

Abstract

Quantum data embedding plays a crucial role in quantum machine learning (QML), as it determines how classical data is mapped into quantum states for further processing. In this talk, I will present our recent advances in optimizing quantum embeddings by leveraging classical deep learning techniques to enhance data separability, classification accuracy, and robustness to noise beyond the limitations of completely positive and trace-preserving maps. Following, I will explore how deterministic quantum computation with one qubit enables efficient training of quantum embeddings and its potential impact on near-term quantum devices.

Finally, I will introduce a margin-based perspective on generalisation in QML, providing a refined theoretical framework to assess and improve the learning capabilities of quantum models. These developments offer new insights into the design, implementation and evaluation of quantum embeddings, paving the way for more effective QML applications. 

Speaker

Daniel Park is an Assistant Professor at Yonsei University in Korea, specialising in quantum machine learning. He has a PhD in Physics-Quantum Information from the University of Waterloo and the Institute for Quantum Computing (2015). Prior to Yonsei, he held postdoctoral and research professor positions at KAIST and Sungkyunkwan University.