DEEP-IN-iPI Joint Meeting

The onset of hydrodynamics in the hot medium created in relativistic heavy-ion collisions is a crucial theoretical question. A first-principle simulation requires a real-time, non-perturbative calculation of the quantum system. In this work, we perform such simulations using the tensor network method, which enables large-scale quantum many-body simulations by retaining only the most essential quantum states for collective behaviors. We focus on the massive Schwinger model, a low-dimensional analog of quantum chromodynamics (QCD), as they share important properties such as confinement and chiral symmetry breaking.

Starting from an initial state that puts a localized excitation atop the vacuum and mimics the energy deposition from colliding nuclei,  we observe hydrodynamic behavior consistent with Bjorken flow in all relevant degrees of freedom: energy density, fluid velocity, and bulk pressure. The time scale for hydrodynamic onset aligns with the thermalization time of the quantum distribution function. This simulation provides a controllable tool to test microscopic theories of hydrodynamics in a strong-coupling quantum system.

Diffusion models are a widely used method in generative AI. I will discuss the application to lattice field theory and discuss various connections with methods known from computational physics. Applications to scalar and gauge theories are presented. 

Many modern machine learning approaches focus on learning effective representations of data distributions from large datasets. Energy-based diffusion models provide an alternative viewpoint: they encode a dataset into an effective action-like functional through a learnable diffusion process. In this talk, I will demonstrate how such model can be used to investigate the explicit symmetry structure of a dataset at the level of the action itself, and how the Renormalization Group Diffusion Model (RGDM) leads to a symmetry-preserving generation trajectory.

Mapping the Milky Way’s dark matter requires moving beyond traditional, rigid dynamical models. In this talk, generative models—specifically Normalizing Flows— are used to learn the stellar phase space distribution directly from Gaia data. This approach enables a flexible, model-independent reconstruction of the Galactic gravitational potential and local dark matter density. These data-driven techniques provide a promising avenue to handle complex observational biases and what they reveal about the dark sector's influence on our Galaxy.

The rapid development of quantum computers has made them a promising new tool for many research fields, including high-energy physics. Among various research methods, quantum simulation focuses on directly simulating Hamiltonians on quantum computers. I will discuss recent developments and also the main challenges in quantum simulation for high-energy physics. I will also briefly discuss how quantum simulation can foster connections between physics and other research fields.

In recent years machine learning (ML) has started to make impact in lattice field theory (LFT), e.g. for the generation of ensembles of configurations. In this talk I will explore potential impact in the opposite direction, i.e. using theoretical physics to understand ML approaches. I will relate stochastic gradient descent to random matrix theory and then make the connection between neural networks and disordered systems, leading to a neural network phase diagram in the plane spanned by hyper parameters. I will conclude with the possible impact of our findings for practical ML applications. 

A central challenge in machine learning is to distinguish genuine structure from chance correlations in high-dimensional data. In this work, we address this issue for the perceptron, a foundational model of neural computation. Specifically, we investigate the relationship between the pattern load α and the variable selection ratio ρ for which a simple perceptron can perfectly classify P = αN random patterns by optimally selecting M = ρN variables out of N variables. While the Cover–Gardner theory establishes that a random subset of ρN dimensions can separate αN random patterns if and only if α < 2ρ, we demonstrate that optimal variable selection can surpass this bound by developing a method, based on the replica method from statistical mechanics, for enumerating the combinations of variables that enable perfect pattern classification. This not only provides a quantitative criterion for distinguishing true structure in the data from spurious regularities, but also yields the storage capacity of associative memory models with sparse asymmetric couplings.