EAGER: QAC-QSA: Can classical machine learning beat variational quantum algorithms at their own game?

点击翻译按钮获取中文摘要

Professor Francesco Evangelista of Emory University is supported by an award from the Chemical Theory, Models and Computation Program in the Division of Chemistry to develop algorithms that combine machine learning and quantum computing. Quantum computers carry out computations using the principles of quantum mechanics and these quantum computers may have an enormous advantage over classical computers in many tasks. A particularly promising application of quantum computers is the simulation of electrons and nuclei: the elementary constituents of atoms, molecules, and materials. Many open questions regarding the properties and the ways molecules react cannot be answered even using the fastest supercomputers. Quantum computers can potentially address even the most challenging chemistry problems. Realizing this potential requires the development of practical quantum algorithms for molecular simulations. Professor Evangelista develops methods that combine classical machine learning with quantum algorithms to create more efficient ways to perform molecular simulations. This project's broader impacts include organizing a winter school to train a broad and diverse generation of researchers and educators in quantum computing. The project also supports the development of open-source computer codes that implement these new algorithms.This project explores adaptive versions of the variational quantum eigensolver method. These approaches have been demonstrated to produce very compact quantum circuits. While successful in this regard, the current approaches are impractical in applications based on near-term quantum computers due to the high number of measurements they require. A new strategy is pursued based on machine learning to avoid the high measurement cost of current adaptive variational quantum algorithms. The selection of a compact quantum circuit for variational quantum algorithms is formulated as a game in which the goal is to find the best variational solution with fewer quantum gates. More fundamentally, this project explores ways to generate machine-learned quantum circuits optimal for a specific instance of a computational problem. Therefore, it could lead to an approach broadly applicable to other problems in quantum information science, where compact quantum circuits are sought.This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.