Kathleen Hamilton
Research Scientist, Computational Sciences & Engineering Division
Oak Ridge National Laboratory
I am a research staff member in the Quantum Computational Science Group. Since joining Oak Ridge National Laboratory in 2016 I have worked on several projects focusing on the development of algorithms, methods and applications for near-term, next-generation processors (e.g. quantum computers and neuromorphic computers).
Currently I am focused on hybrid workflows that use classical and next-generation platforms. This work allows me to leverage my diverse research background in quantum computing, discrete mathematics, and correlated dynamics. What I find fascinating about my research is discovering how different components of hybrid workflows interact, and how this can be used to build scalable applications that are noise robust. In addition to understanding the capabilities of near-term hardware, my work has also resulted in demonstrations of real-world applications such as epidemic modeling.
Presentation Title:
NISQ2NISQ: Self-Supervised Learning and the Role of Classical AI in Quantum Data Post-Processing Workflows
Presentation Abstract:
As the capabilities of near-term quantum hardware continue to develop, there is the potential to distributed computing workflows that scale beyond small qubit sizes. However, true fault tolerance is still out of reach for most quantum processors, and thus noise mitigation strategies are vital and need to address three key challenges. First: quantum hardware noise is difficult to fully characterize due to its many sources and temporal dependence. Second: many end users rely on remote access, and may not be able to exhaustively characterize a system before sending compute jobs. Third: to ensure moderate longevity, we need methods that can operate on 100 noisy qubits.
Similar challenges are found in the medical imaging field, and self-supervised learning has had great success in training denoising models. In this talk I will present results adapting such training paradigms to noisy quantum data and discuss the potential for incorporating such trained models in larger workflows.
