The Barcelona Supercomputing Center – Centro Nacional de Supercomputacion (BSC-CNS) and Fujitsu Limited will sign a dual collaboration agreement on April 19 to promote personalized medicine through the exploitation of clinical data and to advance quantum simulation technologies using tensor networks(1).
Based on this agreement, the two parties will start joint research in May 2023.
The first collaboration project aims to position BSC and Fujitsu at the forefront of a new field that is key to enabling precision medicine: the ability to exploit different types of data to be used in the clinic, from molecular features in the genome to large scale features in X-ray images. In this way, the two parties will not only contribute to improving disease detection rates, but also to reducing the burden on doctors when diagnosing diseases. Great efforts are being made to make clinical data available at both the national and European levels, but the development of technologies to fully exploit such data remains in its early stages.
This project combines BSC’s Life Sciences department expertise in natural language processing of medical records, genomics, and multi-layer networks with Fujitsu’s existing research in genomics AI, large scale causal discovery, computer vision and HPC high speed computing technology. The two parties aim to create a next generation large-scale multimodal AI technology for precision medicine by realizing medical data with large-scale graph structure leveraging these respective strengths. Another primary goal of the collaboration is the development of digital twins in biomedicine, using genomics, medical and imaging data as input for models of biological processes and cellular interactions.
Quantum computing simulation
The second collaborative initiative focuses on the simulation of quantum circuits using tensor networks. The simulation of quantum computers offers the possibility to design, develop, and test novel quantum algorithms under conditions not available yet in experimental devices.
Expanding the scale of quantum circuit calculations represents an ongoing challenge, as current quantum simulators must double memory when increasing the size of a quantum circuit for 1 qubit.
To address this issue, the two parties will utilize tensor networks to reduce the computational complexity of quantum circuits, realizing a quantum simulator that can perform large-scale quantum circuit calculations with the same memory capacity as before and allowing simulations comparable in size to the best current quantum devices.
In this project, BSC and Fujitsu will develop new high-performance computing (HPC) tensor network methods suitable for Fujitsu systems and other modern architectures. In a second phase, the results will be applied to relevant industrial customer problems, including a comprehensive study of potential applications of quantum circuit simulation.
Mateo Valero, director of BSC, said: “This dual agreement with Fujitsu, which is the culmination of years of mutual collaboration, allows us to advance research in two important areas such as personalized medicine and quantum computing. We hope that this joint research will result in new technologies that can ultimately benefit society.”
Fujitsu Limited SEVP, CTO & CPO, Vivek Mahajan comments: “We are excited to collaborate with BSC to accelerate R&D on multimodal AI and quantum simulators. We will build on this joint research program to further strengthen our lineup of advanced computing and AI technologies and develop new practical applications. Fujitsu will actively promote joint research on a global level to contribute to the realization of a sustainable society and take the lead in a sustainable technology development.”
(1) Tensor networks :
The product of tensors (vectors, matrices, etc.) in the form of a network. Also used for the simulation of quantum circuits, where various algorithms for contractions have been proposed for optimization and speeding up the computation.
