Supervised learning with quantum-enhanced feature spaces.

Area of Science:

• Quantum Computing
• Machine Learning
• Supervised Learning

Background:

• Kernel methods in machine learning, such as support vector machines (SVMs), face challenges with large feature spaces and computationally expensive kernel function estimations.
• Quantum computing offers potential computational speed-ups by exploiting exponentially large quantum state spaces through entanglement and interference.
• Bridging quantum computing and machine learning is crucial for addressing complex computational problems.

Purpose of the Study:

• To propose and experimentally implement two novel quantum algorithms for supervised learning classification tasks.
• To explore the use of quantum-enhanced feature spaces for potential quantum advantage in machine learning.
• To investigate the application of noisy intermediate-scale quantum (NISQ) computers to machine learning.

Main Methods:

• Experimental implementation of two quantum algorithms on a superconducting processor.
• Utilizing the quantum state space as an enhanced feature space, accessible efficiently only on a quantum computer.
• One method employs a quantum variational classifier using variational quantum circuits, analogous to classical SVMs.
• The second method involves a quantum kernel estimator that computes kernel functions on the quantum computer to optimize a classical SVM.

Main Results:

• Successful experimental implementation of two quantum algorithms for classification.
• Demonstration of quantum-enhanced feature spaces as a pathway to quantum advantage.
• Development of tools for applying NISQ devices to machine learning problems.

Conclusions:

• Quantum algorithms can effectively address limitations in classical machine learning, particularly for large feature spaces.
• Quantum-enhanced feature spaces offer a promising approach for achieving quantum advantage in machine learning.
• The developed methods provide practical tools for leveraging NISQ computers in machine learning applications.