Loading...
Opportunities for use of quantum spin glasses as reservoir in time series prediction
Shiekh, Kylee N.
Shiekh, Kylee N.
Citations
Altmetric:
Advisor
Editor
Date
Date Issued
2024
Date Submitted
Collections
Research Projects
Organizational Units
Journal Issue
Embargo Expires
Abstract
This thesis investigates the paradigm of Quantum Reservoir Computing (QRC), a cutting-edge framework that employs fixed quantum circuits to improve neural network performance. Rooted in the principles of Reservoir Computing, this model of QRC diverges by integrating a quantum reservoir capable of adapting to multivariable time series data without recurrent circuit modifications between training sessions. Reservoir Computing, a subset of machine learning, centers on harnessing dynamic systems, or reservoirs, to process information and simplify computations. Within this framework, the quantum reservoir exhibits intrinsic adaptability to complex data structures, enabling seamless encoding of intricate temporal relationships within multivariable datasets. Furthermore, this research contextualizes Quantum Computing, a revolutionary field leveraging quantum mechanics principles to revolutionize computation. Quantum phenomena, such as superposition and entanglement, underpin the principles of quantum computing. The current state of quantum computing explores harnessing these phenomena to develop powerful computational frameworks, yet remains in a nascent stage due to challenges in achieving and maintaining quantum coherence and scalability. This study showcases the potential efficacy of the proposed QRC model in multivariable time series prediction, demonstrating its capability in handling intricate dependencies within complex datasets. By integrating the quantum spin glass model as the quantum reservoir, this research underscores the adaptability of such reservoirs in encoding temporal relationships and capturing intricate data dynamics. This work accentuates the advantages of Quantum Reservoir Computing over conventional quantum machine learning frameworks, which rely on backpropogation and gradient descent, with more efficient use of computational resources. This work contributes to the evolving landscape of Quantum Reservoir Computing, shedding light on successful implementations while delineating avenues for refinement. By exploring successful applications and contextualizing within the evolving field of quantum computing, this thesis guides future directions for the refinement and expansion of quantum computing frameworks, specifically in the realm of multivariable time series data processing.
Associated Publications
Rights
Copyright of the original work is retained by the author.