Tensor Networks for Data Science: Stability, Efficiency and Explainability at Scale
As stated by the researchers: “Tensor networks represent exponentially large data structures in an efficient manner. This compressed representation can lead to dramatic improvements in storage and processing. Because of this, they have become a central tool in physics and quantum computation. While tensor networks are expected to play just as important a role in data science, their adoption has been hampered by some conceptual and technical obstacles. By combining insight from physics, data science, and life science, this project will overcome these obstacles and establish tensor networks as the go-to tool for very large-scale data modelling. Furthermore, since tensor networks possess a clean mathematical structure, they typically provide explainable models – in contrast to modern deep learning. We will apply the new methods to key problems in machine learning, chemometrics and quantum computation.”