On the application of adiabatic quantum computing in machine learning

• Grant number: 24/18269-0
• Support Opportunities: Scholarships in Brazil - Program to Stimulate Scientific Vocations
• Start date: March 03, 2025
• End date: April 10, 2025
• Field of knowledge: Engineering - Electrical Engineering - Telecommunications
• Principal Investigator: Leonardo Tomazeli Duarte
• Grantee: Pedro Faria Albuquerque
• Host Institution: Faculdade de Ciências Aplicadas (FCA). Universidade Estadual de Campinas (UNICAMP). Limeira , SP, Brazil

Abstract

A theme that has been gaining prominence in the scientific literature is the development of artificial intelligence (AI) algorithms-especially machine learning-adapted for execution on quantum computers, with the goal of benefiting from potential computational performance gains provided by quantum computing. One of the central challenges in this area is that the quantum and classical paradigms of computation are quite distinct in terms of information representation and processing, requiring a reformulation of existing AI algorithms. Additionally, as another challenge, there is more than one quantum computing paradigm, each based on processing and information representation structures that are not necessarily similar.In the present project, we will conduct a study on the formulation of machine learning methods adapted to the quantum adiabatic computing paradigm. This paradigm, which relies on the so-called adiabatic theorem, can be understood as a single-instruction computer associated with solving an unconstrained quadratic binary optimization (QUBO) problem. Therefore, the formulation of machine learning algorithms for execution on adiabatic computers requires adapting the optimization problem associated with a given learning task to a QUBO formulation. In our research, this task will be investigated in the context of two classic problems in supervised learning: regression and classification.The project will consist of a bibliographic study, followed by implementation in Python and numerical experiments to verify the performance of the QUBO formulation in the context of classification.