Quantum Feature Maps for the NISQ Era Workshop

April 16, 2021

 

The idea of a quantum feature map, or a circuit that maps a classical datapoint into the computational space of a quantum computer, has only recently gained the attention of researchers. The motivation for using such feature maps is provided by the success of classical kernel methods, where an embedding of the data into a (often higher-dimensional) feature space makes the data easier to analyze. Quantum computing provides a natural extension, as it can be understood as the efficient indirect manipulation of objects (quantum states) in an exponentially-sized Hilbert space. Many existing kernel methods, including support vector machines, support vector regression, kernel regularized least squares, kernel principal component analysis, and more may benefit from quantum- enhanced feature spaces. A number of quantum feature maps that can be executed on near-term quantum computers have been proposed and experimentally demonstrated with the potential for quantum advantage stemming from the classical hardness of evaluating such feature maps classically. 

However, the potential for a quantum computational advantage with quantum feature maps is still poorly understood, since the computational hardness does not necessarily imply an advantage for an application. At the same time, classical computational hardness is a prerequisite for quantum advantage, further limiting the class of ”interesting” feature maps to consider. The goal of this workshop is to provide an overview of current challenges in constructing applicable feature maps and an understanding of the state-of-the-art, as well as start conversations about how these challenges can be addressed under the constraints imposed by NISQ hardware. 

All times in Central Time 

Welcome and Talks 
10:55 – 11:00 am  Opening Remarks 
11:00 – 11:30 am  Why data encoding determines the power of quantum machine learning models
Maria Schuld (Xanadu)
 11:30 am – 12:00 pm  Quantum feature maps and generative adversarial networks
Seth Lloyd (MIT)
12:00 – 12:30 pm  The power of data and simple methods for assessing the possibility of quantum advantage in learning
Jarrod McClean (Google)
12:30 – 1:00 pm  A rigorous and robust quantum speed-up in supervised machine learning 
Kristan Temme (IBM)
Break 

1:00 – 1:15 pm 

1:15 – 1:45 pm  Quantum machine learning using Gaussian processes with performant quantum kernels
Matt Otten (HRL)
1:45 – 2:15 pm  Machine learning of high-dimensional data on a noisy quantum processor
Evan Peters (University of Waterloo)
2:15 – 3:00 pm  Panel discussion 
Ruslan Shaydulin, Moderator 

 

Organizing Committee


Jeffrey Larson (Argonne)
Matt Otten (HRL)
Ruslan Shaydulin (Argonne)