Machine learning for quantum information and computing
Doktorsavhandling, 2023
This compilation thesis explores the merger of machine learning, quantum information, and computing. Inspired by the successes of neural networks and gradient-based learning, the thesis explores how such ideas can be adapted to tackle complex problems that arise during the modeling and control of quantum systems, such as quantum tomography with noisy data or optimizing quantum operations, by incorporating physics-based constraints. We also discuss the Bayesian estimation of a quantum state with uncertainty estimates using physically meaningful priors.
Classical machine learning could inspire new quantum-computing algorithms. One such idea is presented to extend the capabilities of variational quantum algorithms using implicit differentiation, enabling straightforward computation of physically interesting quantities on a quantum computer as a gradient. Implicit differentiation also leads to a novel method to generate multipartite entangled quantum states and allows hyperparameter tuning of quantum machine learning algorithms.
Several new experiments were possible due to the theoretical and numerical techniques developed in the thesis — robust generation of a Gottesman- Kitaev-Preskill and cubic phase state in a 3D cavity, fast process tomography of a new family of superconducting gates with known noise, efficient process tomography of a physical operation implementing a logical gate on a bosonic error-correction code, and the reconstruction of a photoelectron’s quantum state.
Classical machine learning could inspire new quantum-computing algorithms. One such idea is presented to extend the capabilities of variational quantum algorithms using implicit differentiation, enabling straightforward computation of physically interesting quantities on a quantum computer as a gradient. Implicit differentiation also leads to a novel method to generate multipartite entangled quantum states and allows hyperparameter tuning of quantum machine learning algorithms.
Several new experiments were possible due to the theoretical and numerical techniques developed in the thesis — robust generation of a Gottesman- Kitaev-Preskill and cubic phase state in a 3D cavity, fast process tomography of a new family of superconducting gates with known noise, efficient process tomography of a physical operation implementing a logical gate on a bosonic error-correction code, and the reconstruction of a photoelectron’s quantum state.
quantum machine learning
quantum process tomography
quantum information
Machine learning
generative neural networks
variational quantum algorithms
quantum state tomography
optimization
quantum computing
Bayesian estimation
Författare
Shahnawaz Ahmed
Chalmers, Mikroteknologi och nanovetenskap, Tillämpad kvantfysik
Quantum State Tomography with Conditional Generative Adversarial Networks ()
Physical Review Letters,;Vol. 127(2021)
Artikel i vetenskaplig tidskrift
Classification and reconstruction of optical quantum states with deep neural networks ()
Physical Review Research,;Vol. 3(2021)
Artikel i vetenskaplig tidskrift
Robust Preparation of Wigner-Negative States with Optimized SNAP-Displacement Sequences
PRX Quantum,;Vol. 3(2022)
Artikel i vetenskaplig tidskrift
Pulse-level noisy quantum circuits with QuTiP
Quantum,;Vol. 6(2022)
Artikel i vetenskaplig tidskrift
Gradient-Descent Quantum Process Tomography by Learning Kraus Operators
Physical Review Letters,;Vol. 130(2023)
Artikel i vetenskaplig tidskrift
Extensive characterization and implementation of a family of three-qubit gates at the coherence limit
npj Quantum Information,;Vol. 9(2023)
Artikel i vetenskaplig tidskrift
Transmon qubit readout fidelity at the threshold for quantum error correction without a quantum-limited amplifier
npj Quantum Information,;Vol. 9(2023)
Artikel i vetenskaplig tidskrift
Quantum physics has revolutionized our understanding of the natural world, stretching the boundaries of classical physics to encompass the mysterious behaviors of particles at the smallest scales. Quantum computers promise a leap in information processing capabilities by harnessing quantum-mechanical effects that can overcome the limitations of classical digital computing as they scale to a size where quantum effects start to dominate. However, to build such quantum computers, we must overcome many challenges that demand precise characterization and enhanced control over quantum devices. Meeting these challenges calls for creating sophisticated models for quantum systems that are fitted with large amounts of noisy experimental data.
Machine learning is a way to push the boundaries of computer programming with data-driven learning, obviating the need for explicit programming. Modern machine learning algorithms have achieved remarkable feats, from identifying objects in images to generating entirely new images from text descriptions and emulating intelligence through conversations, all accomplished by processing large amounts of data.
The convergence of machine learning, quantum information, and computing presents an exciting frontier. We can construct a bridge between these domains by framing the challenges of modeling and characterization in quantum systems as machine-learning problems. At the same time, machine-learning-inspired concepts can lead to new ideas for quantum algorithms. We explore the two aspects of this merger between machine learning, quantum information, and computing.
Machine learning is a way to push the boundaries of computer programming with data-driven learning, obviating the need for explicit programming. Modern machine learning algorithms have achieved remarkable feats, from identifying objects in images to generating entirely new images from text descriptions and emulating intelligence through conversations, all accomplished by processing large amounts of data.
The convergence of machine learning, quantum information, and computing presents an exciting frontier. We can construct a bridge between these domains by framing the challenges of modeling and characterization in quantum systems as machine-learning problems. At the same time, machine-learning-inspired concepts can lead to new ideas for quantum algorithms. We explore the two aspects of this merger between machine learning, quantum information, and computing.
Fundament
Grundläggande vetenskaper
Ämneskategorier
Atom- och molekylfysik och optik
Annan fysik
Datavetenskap (datalogi)
ISBN
978-91-7905-915-6
Doktorsavhandlingar vid Chalmers tekniska högskola. Ny serie: 5381
Utgivare
Chalmers