arXiv subjects

Lars H. Heyen

Publications and source records attributed to Lars H. Heyen.

7 recordsLinked to original sources

arXiv2026

Rethinking Likelihood distributions: Student's t Likelihood Boosts Bayesian Neural Network Performance

In Bayesian neural networks (BNNs), variational inference is a widely adopted framework for modeling uncertainty in a distributional way, with the evidence lower bound (ELBO) serving as the standard objective function. Several distributions contribute to the ELBO loss, such as the prior, approximated posterior, and likelihood distribution. Typically, these distributions are all approximated by a Gaussian distribution, since it is easy to compute, allows for reparameterized gradients, and provides a closed-form loss for training. However, several works have highlighted that this assumption may not generally hold, posing the risk of model misspecification. Alternative distributions have been proposed for the prior specifically, while the effect of distribution choice on the likelihood distribution remains unexplored. In this work, our aim is to close this gap by investigating whether alternative assumptions for the likelihood distribution can outperform the commonly used Gaussian. We compare several likelihood distribution assumptions, such as skewed or heavy-tailed, across regression tasks on both artificial and real-world datasets using standard multilayer perceptrons (MLPs). Our findings demonstrate that Student's t yields better predictive performance than a Gaussian likelihood distribution, independent of the data distribution and MLP architecture (depth and width). In some cases, Student's t can also lead to shorter training times, while still being easy to implement.

cs.LG↗

arXiv2026

Sampling Parallelism for Fast and Efficient Bayesian Learning

Machine learning models, and deep neural networks in particular, are increasingly deployed in risk-sensitive domains such as healthcare, environmental forecasting, and finance, where reliable quantification of predictive uncertainty is essential. However, many uncertainty quantification (UQ) methods remain difficult to apply due to their substantial computational cost. Sampling-based Bayesian learning approaches, such as Bayesian neural networks (BNNs), are particularly expensive since drawing and evaluating multiple parameter samples rapidly exhausts memory and compute resources. These constraints have limited the accessibility and exploration of Bayesian techniques thus far. To address these challenges, we introduce sampling parallelism, a simple yet powerful parallelization strategy that targets the primary bottleneck of sampling-based Bayesian learning: the samples themselves. By distributing sample evaluations across multiple GPUs, our method reduces memory pressure and training time without requiring architectural changes or extensive hyperparameter tuning. We detail the methodology and evaluate its performance on a few example tasks and architectures, comparing against distributed data parallelism (DDP) as a baseline. We further demonstrate that sampling parallelism is complementary to existing strategies by implementing a hybrid approach that combines sample and data parallelism. Our experiments show near-perfect scaling when the sample number is scaled proportionally to the computational resources, confirming that sample evaluations parallelize cleanly. Although DDP achieves better raw speedups under scaling with constant workload, sampling parallelism has a notable advantage: by applying independent stochastic augmentations to the same batch on each GPU, it increases augmentation diversity and thus reduces the number of epochs required for convergence.

cs.LG↗

arXiv2025

Few is different: deciphering many-body dynamics in mesoscopic quantum gases

Emergent macroscopic descriptions of matter, such as hydrodynamics, are central to our description of complex physical systems across a wide spectrum of energy scales. The conventional understanding of these many-body phenomena has recently been shaken by a number of experimental findings. Collective behavior of matter has been observed in \emph{mesoscopic} systems, such as high-energy hadron-hadron collisions, or ultra-cold gases with only few strongly interacting fermions. In such systems, the separation of scales between macroscopic and microscopic dynamics (at the heart of any effective theory) is inapplicable. To address the conceptual challenges that arise from these observations and explore the universality of emergent descriptions of matter, the EMMI Rapid Reaction Task Force was assembled. This document summarizes the RRTF discussions on recent theoretical and experimental advances in this rapidly developing field. Leveraging technological breakthroughs in the control of quantum systems, we can now quantitatively explore what it means for a system to exhibit behavior beyond the sum of its individual parts. In particular, the report highlights how the (in)applicability of hydrodynamics and other effective theories can be probed across three principal frontiers: the size frontier, the equilibrium frontier, and the interaction frontier.

cond-mat.quant-gas↗

arXiv2025

Jigsaw: Training Multi-Billion-Parameter AI Weather Models with Optimized Model Parallelism

AI-based methods have revolutionized atmospheric forecasting, with recent successes in medium-range forecasting spurring the development of climate foundation models. Accurate modeling of complex atmospheric dynamics at high spatial resolutions and longer lead times requires large neural networks and gigabyte-sized data samples, making accelerator memory and I/O-bandwidth the bottlenecks for model training. We introduce WeatherMixer, a multi-layer-perceptron-based architecture whose workload scales linearly with input size, allowing the model to learn global weather phenomena at accuracies similar to numerical weather prediction. To cope with the computational demand, we propose Jigsaw, a novel model parallelization scheme that employs both domain and tensor parallelism, eliminating memory redundancy. Jigsaw exceeds state-of-the-art performance in strong scaling in compute-communication-limited systems and achieves superscalar weak scaling in I/O-bandwidth-limited systems. We scale training to 256 GPUs, reaching peak performances of 9 and 11 PFLOPs, 23% and 28% of theoretical peaks, achieving 68% and 72% scaling efficiency versus 51% without model parallelism.

cs.LG↗

arXiv2025

Emergent interaction-driven elliptic flow of few fermionic atoms

Hydrodynamics provides a successful framework to effectively describe the dynamics of complex many-body systems ranging from subnuclear to cosmological scales by introducing macroscopic quantities such as particle densities and fluid velocities. According to textbook knowledge, it requires coarse graining over microscopic constituents to define a macroscopic fluid cell, which is large compared to the interparticle spacing and the mean free path. In addition, the entire system must consist of many such fluid cells. In high energy heavy ion collisions, hydrodynamic behaviour is inferred from the observation of elliptic flow. Here, we demonstrate the emergence of elliptic flow in a system of few strongly interacting atoms. In our system a hydrodynamic description is a priori not applicable, as all relevant length scales, i.e. the system size, the inter-particle spacing, and the mean free path are comparable. The single particle resolution, deterministic control over particle number and interaction strength in our experiment allow us to explore the boundaries between a microscopic description and a hydrodynamic framework in unprecedented detail.

cond-mat.quant-gas↗

arXiv2024

Quantum corrections at second order in derivatives to the dynamics of small non-relativistic fluids

To capture the dynamics of macroscopic non-relativistic fluids consisting of very many atoms, it is typically sufficient to truncate the gradient expansion at order zero, leading to ideal fluid dynamics, or at order one, leading to the Navier-Stokes theory. For mesoscopic fluids consisting of a small number of atoms, second-order corrections can become significant. We investigate here specifically superfluids at vanishing temperature, and identify relevant second-order terms of quantum origin that contribute already in a static situation. The general form of these terms arises from an extension of the Gross-Pitaevskii theory. In the context of density functional theory, they are named after C. von Weizsäcker. We assess the influence of these terms on numerical solutions of second-order fluid dynamic equations for the expansion of a mesoscopic ultra-cold Fermi gas released from an anisotropic harmonic trap in two spatial dimensions.

cond-mat.quant-gas↗

arXiv2021

How many particles do make a fluid? Qualifying collective behavior in expanding ultracold gases

Collective phenomena in quantum many-body systems are often described in terms of hydrodynamics, an appropriate framework when the involved particle numbers are effectively macroscopic. We propose to use experiments on expanding clouds of few and many interacting cold atoms to investigate the emergence of hydrodynamics as a function of particle number. We consider gases confined in two-dimensional elliptically-deformed traps, and we employ the manifestation of elliptic flow as an indicator of collective behavior. We quantify the response of the gas to the deformation of the trapping potential, and show how such information can be used to establish how many atoms are needed for the system to develop a degree of collectivity comparable to that expected in the hydrodynamic limit. This method permits one, in particular, to exploit observations made in expanding atomic gases to shed light on the apparent hydrodynamic behaviour of mesoscopic systems of quarks and gluons formed in the scattering of light ions in high-energy collider experiments.

cond-mat.quant-gas↗