Pacific Northwest National Laboratory (PNNL) has introduced a new algorithm, Picasso, that significantly reduces the computational resources required to prepare data for hybrid quantum-classical computing systems. The algorithm, which employs advanced techniques in graph coloring and clique partitioning, addresses a longstanding bottleneck in quantum information preparation by cutting down the required quantum input data by approximately 85%. The work was presented at the IEEE International Symposium on Parallel and Distributed Processing and is now publicly available on GitHub.
Picasso targets the challenge of preparing quantum input—specifically, compressing large sets of Pauli strings that represent quantum operations. In simulations of hydrogen model systems generating over two million Pauli strings and more than one trillion relational connections, the Picasso algorithm effectively grouped operations into a reduced number of cliques. It then used only a fraction of the data (about one-tenth) to compute accurate results. This allowed the team to solve problems 50 times larger than previous tools and process over 2,400 times more relationships using efficient memory management techniques such as streaming and sparsification.
The development team included high-performance computing experts S M Ferdous and Mahantesh Halappanavar, quantum researcher Bo Peng, and collaborators from North Carolina State University. The research demonstrates the critical importance of scalable preprocessing for quantum computing systems and offers a path toward handling problem sizes that require 100 to 1,000 qubits. In addition to Picasso, the team developed an AI-driven algorithm to optimize trade-offs between memory usage and data retention, further supporting efficient integration between classical and quantum computing workflows.
Read the full announcement from PNNL here.
April 23, 2025
Leave A Comment