ML-optimization of tensor contraction operations (TCOs) in the PyBEST software package

The main goal of this project is to integrate advanced machine learning techniques into the open-source PyBEST quantum chemistry software package to substantially accelerate quantum chemical calculations through the automated selection of optimal computational strategies. The project specifically targets the AI-driven optimization of tensor contraction operations (TCOs), which constitute the primary computational bottleneck in many quantum chemistry methods. By enabling data-driven prediction of the most efficient contraction schemes and execution pathways across different problem sizes and hardware configurations, the proposed approach will significantly reduce time-to-solution, enhance scalability on modern GPU architectures, and improve computational resource efficiency, while preserving numerical accuracy and scientific reliability.

Faculty Supervisor:

Stijn De Baerdemacker

Student:

Partner:

Nicolaus Copernicus University in Torun

Discipline:

Computer science

Sector:

Education

University:

University of New Brunswick

Program: