Abstract:In this work, we establish a direct connection between generative diffusion models (DMs) and stochastic quantization (SQ). The DM is realized by approximating the reversal of a stochastic process dictated by the Langevin equation, generating samples from a prior distribution to effectively mimic the target distribution. Using numerical simulations, we demonstrate that the DM can serve as a global sampler for generating quantum lattice field configurations in two-dimensional $\phi^4$ theory. We demonstrate that DMs can notably reduce autocorrelation times in the Markov chain, especially in the critical region where standard Markov Chain Monte-Carlo (MCMC) algorithms experience critical slowing down. The findings can potentially inspire further advancements in lattice field theory simulations, in particular in cases where it is expensive to generate large ensembles.

What problem does this paper attempt to address?

The paper primarily explores the application of Diffusion Models (DMs) in lattice field theory, especially in generating quantum lattice field configurations. Specifically, the authors establish a direct connection between diffusion models and Stochastic Quantization (SQ), where diffusion models generate samples by approximating the inversion of the stochastic process governed by the Langevin equation, effectively mimicking the target distribution. The key contributions and problems addressed by the paper can be summarized as follows: 1. **Solving the sampling efficiency problem in high-dimensional systems**: In lattice field theory, although the Monte Carlo method is widely used for random sampling from the system's probability distribution, it encounters higher computational costs near critical points in the parameter space. As a generative model in machine learning, diffusion models can be trained to generate new data that follows the distribution of a given dataset, potentially improving the efficiency of Monte Carlo simulations. 2. **Reducing Markov chain autocorrelation time**: The paper demonstrates that diffusion models can significantly reduce the Markov chain autocorrelation time when generating quantum lattice field configurations in two-dimensional φ^4 theory, especially in the critical region where traditional Markov chain Monte Carlo algorithms experience critical slowing down. 3. **Enhancing the ability to handle complex weight problems**: For theories with sign problems or complex weight issues, such as quantum chromodynamics at non-zero baryon density or theories in real time rather than Euclidean time, diffusion models may offer an alternative approach, avoiding the sign problem and potentially reducing the cost of sampling from prior distributions. 4. **Establishing a connection with stochastic quantization theory**: The authors point out that the denoising process of diffusion models can be explicitly considered as part of the stochastic quantization procedure, providing a new perspective for understanding the behavior of diffusion models in the context of quantum field theory. 5. **Exploring the application of new methods**: By combining with hybrid Monte Carlo algorithms, diffusion models have been proven to act as effective global samplers, accelerating the process of reaching thermal equilibrium, shortening autocorrelation times, and thus improving sampling efficiency. In summary, the paper addresses key issues in efficiently generating samples in high-dimensional complex systems and reducing autocorrelation times in Monte Carlo simulations, while also providing new tools for dealing with theories that have sign problems or complex weights. Moreover, by linking diffusion models with stochastic quantization theory, the paper paves new paths for understanding the application of these models in quantum field theory.

Diffusion Models as Stochastic Quantization in Lattice Field Theory

Generative Diffusion Models for Lattice Field Theory

Diffusion models for lattice gauge field simulations

Open System Approach to Non-Equilibrium Dynamical Theory of Quantum Dot Systems

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

Stochastic quantization and diffusion models

Approach to Solving Spin-Boson Dynamics Via Non-Markovian Quantum Trajectories

Quantum simulation of lattice gauge theories via deterministic duality transformations assisted by measurements

Quantum Diffusion Models

Quantum circuit synthesis with diffusion models

Renormalizing Diffusion Models

Quantum-Noise-Driven Generative Diffusion Models

Discrete generative diffusion models without stochastic differential equations: a tensor network approach

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

Superdiffusive Transport in Quasi-Particle Dephasing Models

Digital Quantum Simulation of Reaction-Diffusion Systems on Lattice

Disentanglement Approach to Quantum Spin Ground States: Field Theory and Stochastic Simulation

Quantum State Generation with Structure-Preserving Diffusion Model

Anomalous Diffusion in a Dynamical Optical Lattice

Nonequilbrium physics of generative diffusion models

Quantum Cellular Automata from Lattice Field Theories