This thesis covers neural quantum states applied to computational chemistry and variational Monte Carlo methods, synthesizing the work on scalable neural quantum states for quantum chemistry, natural evolution strategies, overcoming scalability barriers in VMC, quantum-inspired variational algorithms for PDEs, and meta-learning for variational Monte Carlo. The thesis provides a unified framework spanning quantum chemistry, combinatorial optimization, and financial derivative pricing, with detailed scalability analysis and GPU parallelization strategies.
One-Sentence Summary
PhD thesis synthesizing neural quantum states for scientific computing, covering variational Monte Carlo for quantum chemistry, combinatorial optimization, and financial derivative pricing.
Key Contributions
Comprehensive treatment of neural quantum states for scientific computing
Unified framework spanning quantum chemistry, combinatorial optimization, and PDEs
Scalability analysis and GPU parallelization strategies for VMC
Cite This Paper If...
You are looking for a comprehensive reference on neural quantum states for scientific computing, including applications to computational chemistry, combinatorial optimization, and financial derivative pricing. This thesis provides a unified treatment of the theory, algorithms, and scalability considerations for VMC-based approaches with neural-network wavefunctions.
Frequently Asked Questions
What topics does this thesis cover?
The thesis synthesizes research on neural quantum states applied to three domains: ab-initio quantum chemistry, combinatorial optimization (Max-Cut), and financial derivative pricing (multi-asset Black-Scholes). It covers theory, algorithms, and large-scale implementation strategies.
Which published papers are included in this thesis?
The thesis draws from several published works including: Scalable Neural Quantum States Architecture for Quantum Chemistry (ML:ST 2023), Natural Evolution Strategies and Variational Monte Carlo (ML:ST 2021), Overcoming Barriers to Scalability in VMC (SC 2021), Quantum-Inspired Variational Algorithms for PDEs (Quantitative Finance 2024), and Meta Variational Monte Carlo.
What is the main computational contribution?
The thesis presents GPU parallelization strategies that overcome the scalability bottleneck in variational quantum Monte Carlo by replacing MCMC sampling with autoregressive exact sampling, enabling distributed-memory and GPU parallelism for problems with up to ten thousand dimensions.
How does this work connect quantum computing and machine learning?
Neural quantum states use neural networks to parameterize quantum wavefunctions, combining the expressiveness of deep learning with the variational Monte Carlo framework from quantum physics. The thesis shows this hybrid approach can tackle problems in chemistry, optimization, and finance that are intractable for classical methods.
Where can I find the full thesis?
The thesis is available through the University of Michigan Deep Blue repository. Check the PDF link above for the direct download (link pending).
Keywords
neural quantum states scientific computingneural quantum states computational chemistryneural quantum states financevariational Monte Carlo chemistry financescientific computing neural wavefunctions
@phdthesis{zhao2022neural,
title={Neural Quantum States for Scientific Computing},
author={Zhao, Tianchen},
school={University of Michigan},
year={2022}
}
Recommended Citation Contexts
Neural quantum states provide a unified variational framework connecting quantum chemistry, physics simulation, and scientific computing applications (Zhao, 2022).