Rethinking Quantum Neural Networks

Time: Mar 5th, 10am – 11am

Location: Online via Zoom (https://csun.zoom.us/j/86387751464)

Speaker: Dr. Eric R. Anschuetz, Caltech

Title: Rethinking Quantum Neural Networks

Abstract: In recent years, the prospect of using quantum devices for machine learning tasks has garnered much study. This is, in part, due to the myriad recent experiments demonstrating quantum devices which sample from distributions that are believed to be difficult to sample from classically. Recent theoretical analysis of quantum machine learning (QML) algorithms has determined that, indeed, there exist learning tasks using classical data that are more efficiently performed by QML models than classical machine learning models. However, these results come with an important caveat: QML models are generally inefficient to train. In this talk I will give an introduction to these results. I will also speak on recent methods for balancing the expressive power offered by QML models with their trainability concerns to show end-to-end quantum advantages on certain machine learning tasks.