Two Classification Algorithms in Quantum Machine Learning
Seminar
QUEST Center event
Yes
Speaker
Zohim Chandani and Graham Enos (Rigetti)
Date
09/03/2022 - 19:00 - 18:00Add to Calendar
2022-03-09 18:00:00
2022-03-09 19:00:00
Two Classification Algorithms in Quantum Machine Learning
The case for utilizing quantum circuits for supervised learning is generally derived from accessing the exponentially large Hilbert space dimension that computation with qubits yields. It is hoped that encoding data into this space may yield advantages in classification margins previously unreachable by classical methods if the feature map used is hard to simulate classically. We discuss two classification algorithms, namely quantum neural networks and quantum kernel methods, both of which utilize a classical optimization routine. We introduce a novel gradient estimator (GSPSA) suitable for implementation on NISQ devices and show its advantages over SPSA which is popular in the literature. Moreover, we explore the idea of deriving a gram matrix for support vector machines via a quantum feature map encoding, the problems associated with the exponential scaling of Hilbert space, and how one can overcome them with projected quantum kernels.
http://zoom.us/j/88022048688
Department of Physics
physics.dept@mail.biu.ac.il
Asia/Jerusalem
public
Place
http://zoom.us/j/88022048688
Abstract
The case for utilizing quantum circuits for supervised learning is generally derived from accessing the exponentially large Hilbert space dimension that computation with qubits yields. It is hoped that encoding data into this space may yield advantages in classification margins previously unreachable by classical methods if the feature map used is hard to simulate classically. We discuss two classification algorithms, namely quantum neural networks and quantum kernel methods, both of which utilize a classical optimization routine. We introduce a novel gradient estimator (GSPSA) suitable for implementation on NISQ devices and show its advantages over SPSA which is popular in the literature. Moreover, we explore the idea of deriving a gram matrix for support vector machines via a quantum feature map encoding, the problems associated with the exponential scaling of Hilbert space, and how one can overcome them with projected quantum kernels.
Last Updated Date : 01/03/2022
