Abstract:This article considers the generative modeling of the (mixed) states of quantum systems, and an approach based on denoising diffusion model is proposed. The key contribution is an algorithmic innovation that respects the physical nature of quantum states. More precisely, the commonly used density matrix representation of mixed-state has to be complex-valued Hermitian, positive semi-definite, and trace one. Generic diffusion models, or other generative methods, may not be able to generate data that strictly satisfy these structural constraints, even if all training data do. To develop a machine learning algorithm that has physics hard-wired in, we leverage mirror diffusion and borrow the physical notion of von Neumann entropy to design a new map, for enabling strict structure-preserving generation. Both unconditional generation and conditional generation via classifier-free guidance are experimentally demonstrated efficacious, the latter enabling the design of new quantum states when generated on unseen labels.

What problem does this paper attempt to address?

This paper mainly discusses the generation modeling of quantum system state (mixed state) and proposes a new method based on the structure-preserving diffusion model. Traditional diffusion models or other generation methods may not rigorously satisfy the physical constraints that quantum states must be complex conjugate symmetric, positive semidefinite, and trace-one. To address this issue, the paper utilizes mirror diffusion and von Neumann entropy to design a new mapping to ensure structure-preserving generation. Experiments show that this method is effective in both unconditional generation and classifier-guided conditional generation, the latter allowing the design of new quantum states on unseen labels. Specifically, the paper presents a machine learning algorithm that incorporates physical laws into the model to generate density matrices that are similar to the training data and strictly satisfy the structural constraints of quantum states. Unlike traditional generative models, this method does not design quantum algorithms but develops algorithms that run on classical computers to process quantum data. The research also emphasizes the significance of generating more quantum states, including designing innovative quantum states, estimating probability distributions of quantum states, and providing synthesized samples for expensive experimental data. In addition, the paper contrasts the difference between quantum state generation and quantum state tomography and points out that generative models can enhance the quantitative understanding of the underlying physical processes behind quantum systems. In conclusion, this paper aims to address the problem of generating quantum states that conform to the rules of quantum mechanics using classical computers and achieves effective modeling of quantum data through innovative algorithms.

Quantum State Generation with Structure-Preserving Diffusion Model

Quantum-Noise-Driven Generative Diffusion Models

Quantum Generative Diffusion Model: A Fully Quantum-Mechanical Model for Generating Quantum State Ensemble

Generative Quantum Machine Learning via Denoising Diffusion Probabilistic Models

Quantum Generative Models for Data Generation

Quantum Denoising Diffusion Models

Quantum circuit synthesis with diffusion models

Generative modeling assisted simulation of measurement-altered quantum criticality

Quantum sequential scattering model for quantum state learning

Trainability barriers and opportunities in quantum generative modeling

Quantum Diffusion Models

Non-Markovian Quantum State Diffusion for an Open Quantum System in Fermionic Environments

Experimental demonstration of reconstructing quantum states with generative models

A Quantum Machine Learning Algorithm Based on Generative Models

Flexible learning of quantum states with generative query neural networks

Recent Advances for Quantum Neural Networks in Generative Learning

Quantum generative adversarial network for generating discrete distribution

Toward Transparent and Controllable Quantum Generative Models

Predicting Properties of Quantum Systems with Conditional Generative Models

Enhancing Quantum Diffusion Models with Pairwise Bell State Entanglement