ABSTRACT

Advances in sensing, data collection, algorithms, and high-performance computing have resulted in a new paradigm of scientific discovery called the data-driven scientific discovery, where different types of data-driven models are trained based on the available data for knowledge discovery and reasoning, forecasting, and system-performance prediction. Due to the noisy and imprecise nature of data, these analyses need to be performed in the presence of uncertainty. Bayesian networks constitute one model that can be used to represent noisy and imprecise data, and that has been employed in applications ranging from atomic-level systems to cosmology, healthcare, and in various engineering domains such as transportation, manufacturing, civil infrastructure, and aerospace systems. In the last decade, there has also been tremendous interest in the field of quantum computing due to its superior computational performance over conventional computing paradigms in solving certain types of problems. This project is investigating efficient representation and simulation of Bayesian networks in the quantum-computing paradigm. The results from this project are being incorporated into STEM courses. Multiple undergraduate and graduate students are being trained as part of this project, and several short teaching modules are being developed to train high-school students in quantum computing through annual summer camps.

The proposed project investigates the fundamental question of simulating a Quantum Bayesian Network (QBN) on currently available Noisy Intermediate Scale Quantum (NISQ) devices. The proposed research is investigating a multi-pronged approach for efficient QBN simulation. First, a novel QBN representation framework through rotation angle decomposition, which has a lower analysis complexity without losing the accuracy is being investigated. Second, a mixed optimization-reinforcement learning approach for transpilation is being investigated for combined qubit placement and routing problem to efficiently map any given QBN circuit on to gate-based hardware architectures. Finally, a non-parametric statistical approach is being investigated to obtain an empirical relationship between QBN complexity and the number of quantum circuit runs required for a desired accuracy in QBN state probabilities. The proposed methods are not limited to quantum Bayesian networks but are generic and applicable to any quantum algorithm.

This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.

Please report errors in award information by writing to: awardsearch@nsf.gov.