2024 R&D 100 Award for MAQ: Machine Learning on Adiabatic Quantum Computers

ORNL-developed MAQ is the first software that enables developers to leverage the speed of quantum computing to train machine learning algorithms significantly faster than using classical computers without affecting accuracy. Any application, from artificial intelligence to industrial systems, that depends on machine learning now can benefit from faster training leading to rapid deployment.

The MAQ software speeds up the training of machine learning models using a type of quantum computer called an adiabatic quantum computer, or AQC. Quantum computers, especially AQCs, are good at solving complex problems faster. The MAQ library shows that AQCs can train machine learning models much quicker than regular computers, for example, up to five times faster for certain tasks compared to classical methods.

In addition to speeding up training, MAQ includes a special method for embedding problems onto quantum computers, which is more efficient than previous methods. MAQ’s methods could benefit many fields, such as science, business and healthcare, by making data analysis faster and more effective. In fact, the MAQ library methods have a direct impact on industrial and societal applications. Machine learning significantly impacts both industries and society by streamlining operations, enhancing decision-making and improving quality of life. In industry, it boosts efficiency through predictive maintenance, supply chain optimization and personalized services across sectors such as healthcare, retail and manufacturing.

Societally, machine learning advances healthcare diagnostics, personalized education, environmental conservation and public safety. It also supports sustainable agriculture, urban planning and accessibility for people with disabilities. MAQ methods have potential to revolutionize machine learning applications by accelerating the training process.

The research was funded by ORNL's Laboratory Directed Research and Development SEED program through the DOE Office of Science.

The developers are Prasanna Date, Kathleen Hamilton, Robert Patton, Travis Humble and Thomas Potok, all from ORNL.