An official website of the United States government
Modeling unsteady, fast transient, and advection-dominated physics problems is a pressing challenge for physics-aware deep learning (PADL). The physics of complex systems is governed by large systems of partial differential equations (PDEs) and ancillary constitutive models with nonlinear structures, as well as evolving state fields exhibiting sharp gradients and rapidly deforming material interfaces. Here, we investigate an inductive bias approach that is versatile and generalizable to model generic nonlinear field evolution problems. Our study focuses on the recent physics-aware recurrent convolutions (PARC), which incorporates a differentiator-integrator architecture that inductively models the spatiotemporal dynamics of generic physical systems. We extend the capabilities of PARC to simulate unsteady, transient, and advection-dominant systems. The extended model, referred to as PARCv2, is equipped with differential operators to model advection-reaction-diffusion equations, as well as a hybrid integral solver for stable, long-time predictions. PARCv2 is tested on both standard benchmark problems in fluid dynamics, namely Burgers and Navier-Stokes equations, and then applied to more complex shock-induced reaction problems in energetic materials. We evaluate the behavior of PARCv2 in comparison to other physics-informed and learning bias models and demonstrate its potential to model unsteady and advection-dominant dynamics regimes.
more »
« less
This content will become publicly available on March 1, 2026
Laplace Meets Moreau: Smooth Approximation to Infimal Convolutions Using Laplace’s Method
We study approximations to the Moreau envelope—and infimal convolutions more broadly—based on Laplace’s method, a classical tool in analysis which ties certain integrals to suprema of their integrands. We believe the connection between Laplace’s method and infimal convolutions is generally deserving of more attention in the study of optimization and partial differential equations, since it bears numerous potentially important applications, from proximal-type algorithms to Hamilton-Jacobi equations.
more »
« less
- Award ID(s):
- 2110745
- PAR ID:
- 10636467
- Publisher / Repository:
- Proceedings of Machine Learning Research
- Date Published:
- Journal Name:
- Journal of machine learning research
- ISSN:
- 1532-4435
- Format(s):
- Medium: X
- Sponsoring Org:
- National Science Foundation
More Like this
-
-
Applications of neural networks like MLPs and ResNets in temporal data mining has led to improvements on the problem of time series classification. Recently, a new class of networks called Temporal Convolution Networks (TCNs) have been proposed for various time series tasks. Instead of time invariant convolutions they use temporally causal convolutions, this makes them more constrained than ResNets but surprisingly good at generalization. This raises an important question: How does a network with causal convolution solve these tasks when compared to a network with acausal convolutions? As the first attempt at answering these questions, we analyze different architectures through a lens of representational subspace similarity. We demonstrate that the evolution of input representations in the layers of TCNs is markedly different from ResNets and MLPs. We find that acausal networks are prone to form groupings of similar layers and TCNs on the other hand learn representations that are much more diverse throughout the network. Next, we study the convergence properties of internal layers across different architecture families and discover that the behaviour of layers inside Acausal network is more homogeneous when compared to TCNs. Our extensive empirical studies offer new insights into internal mechanisms of convolution networks in the domain of time series analysis and may assist practitioners gaining deeper understanding of each network.more » « less
-
Native electrospray mass spectrometry is a powerful method for determining the native stoichiometry of many polydisperse multi-subunit biological complexes, including multi-subunit protein complexes and lipid-bound transmembrane proteins. However, when polydispersity results from incorporation of multiple copies of two or more different subunits, it can be difficult to analyze subunit stoichiometry using conventional mass spectrometry analysis methods, especially when m / z distributions for different charge states overlap in the mass spectrum. It was recently demonstrated by Marty and co-workers (K. K. Hoi, et al. , Anal. Chem. , 2016, 88 , 6199–6204) that Fourier Transform (FT)-based methods can determine the bulk average lipid composition of protein-lipid Nanodiscs assembled with two different lipids, but a detailed statistical description of the composition of more general polydisperse two-subunit populations is still difficult to achieve. This results from the vast number of ways in which the two types of subunit can be distributed within the analyte ensemble. Here, we present a theoretical description of three common classes of heterogeneity for mixed-subunit analytes and demonstrate how to differentiate and analyze them using mass spectrometry and FT methods. First, we first describe FT-based analysis of mass spectra corresponding to simple superpositions, convolutions, and multinomial distributions for two or more different subunit types using model data sets. We then apply these principles with real samples, including mixtures of single-lipid Nanodiscs in the same solution (superposition), mixed-lipid Nanodiscs and copolymers (convolutions), and isotope distribution for ubiquitin (multinomial distribution). This classification scheme and the FT method used to study these analyte classes should be broadly useful in mass spectrometry as well as other techniques where overlapping, periodic signals arising from analyte mixtures are common.more » « less
-
Machine learning methods are increasingly being employed as surrogate models in place of computationally expensive and slow numerical integrators for a bevy of applications in the natural sciences. However, while the laws of physics are relationships between scalars, vectors and tensors that hold regardless of the frame of reference or chosen coordinate system, surrogate machine learning models are not coordinate-free by default. We enforce coordinate freedom by using geometric convolutions in three model architectures: a ResNet, a Dilated ResNet and a UNet. In numerical experiments emulating two-dimensional compressible Navier–Stokes, we see better accuracy and improved stability compared with baseline surrogate models in almost all cases. The ease of enforcing coordinate freedom without making major changes to the model architecture provides an exciting recipe for any convolutional neural network-based method applied to an appropriate class of problems. This article is part of the theme issue ‘Partial differential equations in data science’.more » « less
-
Abstract This paper addresses the overdetermined problem of inverting the n -dimensional cone (or Compton) transform that integrates a function over conical surfaces in R n . The study of the cone transform originates from Compton camera imaging, a nuclear imaging method for the passive detection of gamma-ray sources. We present a new identity relating the n -dimensional cone and Radon transforms through spherical convolutions with arbitrary weight functions. This relationship, which generalizes a previously obtained identity, leads to various inversion formulas in n -dimensions under a mild assumption on the geometry of detectors. We present two such formulas along with the results of their numerical implementation using synthetic phantoms. Compared to our previously discovered inversion techniques, the new formulas are more stable and simpler to implement numerically.more » « less
