NCSA staff who would like to submit an item for the calendar can email newsdesk@ncsa.illinois.edu.
Quantum State Tomography with Machine Learning
Abstract: Machine learning (ML) has found broad applicability in quantum information science, where existing ML techniques are often applied without significant alterations to network architectures. However, the concept of data-centric ML suggests that incorporating domain-specific knowledge related to the underlying structure of quantum mechanics could further improve performance. Here we propose physics-inspired data-centric heuristics for ML systems used in quantum information science and demonstrate their efficacy using a neural network trained for the task of quantum state reconstruction. While some presented heuristics are translations of well-known concepts, such as the importance of diversity in training sets, we also find it is not always optimal to engineer training sets to match the expected distribution of a target scenario. Instead, performance improves when training sets are biased to be more mixed than the target distribution. We argue this is due to the heterogeneity in the number of free variables required to describe states of different purity.