arXiv subjects

Ivo F. Sbalzarini

Publications and source records attributed to Ivo F. Sbalzarini.

At least 37 records · Page 2Linked to original sources

arXiv2023

Entropically damped artificial compressibility for the discretization corrected particle strength exchange method in incompressible fluid mechanics

We present a consistent mesh-free numerical scheme for solving the incompressible Navier-Stokes equations. Our method is based on entropically damped artificial compressibility for imposing the incompressibility constraint explicitly, and the Discretization-Corrected Particle Strength Exchange (DC-PSE) method to consistently discretize the differential operators on mesh-free particles. We further couple our scheme with Brinkman penalization to solve the Navier-Stokes equations in complex geometries. The method is validated using the 3D Taylor-Green vortex flow and the lid-driven cavity flow problem in 2D and 3D, where we also compare our method with hr-SPH and report better accuracy for DC-PSE. In order to validate DC-PSE Brinkman penalization, we study flow past obstacles, such as a cylinder, and report excellent agreement with previous studies.

physics.flu-dyn↗

arXiv2023

An open-source pipeline for solving continuous reaction-diffusion models in image-based geometries of porous media

We present a versatile open-source pipeline for simulating inhomogeneous reaction-diffusion processes in highly resolved, image-based geometries of porous media with reactive boundaries. Resolving realistic pore-scale geometries in numerical models is challenging and computationally demanding, as the scale differences between the sizes of the interstitia and the whole system can lead to prohibitive memory requirements. The present pipeline combines a level-set method with geometry-adapted sparse block grids on GPUs to efficiently simulate reaction-diffusion processes in image-based geometries. We showcase the method by applying it to fertilizer diffusion in soil, heat transfer in porous ceramics, and determining effective diffusion coefficients and tortuosity. The present approach enables solving reaction-diffusion partial differential equations in real-world geometries applicable to porous media across fields such as engineering, environmental science, and biology.

cs.MS↗

arXiv2023

Spontaneous flow instabilities of active polar fluids in three dimensions

Active polar fluids exhibit spontaneous flow when sufficient active stress is generated by internal molecular mechanisms. This is also referred to as an active Fréedericksz transition. Experiments have revealed the existence of competing in-plane and out-of-plane instabilities in three-dimensional active matter. So far, however, a theoretical model reconciling all observations is missing. In particular, the role of boundary conditions in these instabilities still needs to be explained. Here, we characterize the spontaneous flow transition in a symmetry-preserving three-dimensional active Ericksen-Leslie model, showing that the boundary conditions select the emergent behavior. Using nonlinear numerical solutions and linear perturbation analysis, we explain the mechanism for both in-plane and out-of-plane instabilities under extensile active stress for perpendicular polarity anchoring at the boundary, whereas parallel anchoring only permits in-plane flows under contractile stress or out-of-plane wrinkling under extensile stress.

cond-mat.soft↗

arXiv2023

Learning locally dominant force balances in active particle systems

We use a combination of unsupervised clustering and sparsity-promoting inference algorithms to learn locally dominant force balances that explain macroscopic pattern formation in self-organized active particle systems. The self-organized emergence of macroscopic patterns from microscopic interactions between self-propelled particles can be widely observed nature. Although hydrodynamic theories help us better understand the physical basis of this phenomenon, identifying a sufficient set of local interactions that shape, regulate, and sustain self-organized structures in active particle systems remains challenging. We investigate a classic hydrodynamic model of self-propelled particles that produces a wide variety of patterns, like asters and moving density bands. Our data-driven analysis shows that propagating bands are formed by local alignment interactions driven by density gradients, while steady-state asters are shaped by a mechanism of splay-induced negative compressibility arising from strong particle interactions. Our method also reveals analogous physical principles of pattern formation in a system where the speed of the particle is influenced by local density. This demonstrates the ability of our method to reveal physical commonalities across models. The physical mechanisms inferred from the data are in excellent agreement with analytical scaling arguments and experimental observations.

cond-mat.soft↗

arXiv2023

Efficient Raycasting of Volumetric Depth Images for Remote Visualization of Large Volumes at High Frame Rates

We present an efficient raycasting algorithm for rendering Volumetric Depth Images (VDIs), and we show how it can be used in a remote visualization setting with VDIs generated and streamed from a remote server. VDIs are compact view-dependent volume representations that enable interactive visualization of large volumes at high frame rates by decoupling viewpoint changes from expensive rendering calculations. However, current rendering approaches for VDIs struggle with achieving interactive frame rates at high image resolutions. Here, we exploit the properties of perspective projection to simplify intersections of rays with the view-dependent frustums in a VDI and leverage spatial smoothness in the volume data to minimize memory accesses. Benchmarks show that responsive frame rates can be achieved close to the viewpoint of generation for HD display resolutions, providing high-fidelity approximate renderings of Gigabyte-sized volumes. We also propose a method to subsample the VDI for preview rendering, maintaining high frame rates even for large viewpoint deviations. We provide our implementation as an extension of an established open-source visualization library.

cs.GR↗

arXiv2023

A high-order fully Lagrangian particle level-set method for dynamic surfaces

We present a fully Lagrangian particle level-set method based on high-order polynomial regression. This enables closest-point redistancing without requiring a regular Cartesian mesh, relaxing the need for particle-mesh interpolation. Instead, we perform level-set redistancing directly on irregularly distributed particles by polynomial regression in a Newton-Lagrange basis on a set of unisolvent nodes. We demonstrate that the resulting particle closest-point (PCP) redistancing achieves high-order accuracy for 2D and 3D geometries discretized on highly irregular particle distributions and has better robustness against particle distortion than regression in a monomial basis. Further, we show convergence in a classic level-set benchmark case involving ill-conditioned particle distributions, and we present an application to an oscillating droplet simulation in multi-phase flow.

cs.CE↗

arXiv2023

Roadmap on Deep Learning for Microscopy

Through digital imaging, microscopy has evolved from primarily being a means for visual observation of life at the micro- and nano-scale, to a quantitative tool with ever-increasing resolution and throughput. Artificial intelligence, deep neural networks, and machine learning are all niche terms describing computational methods that have gained a pivotal role in microscopy-based research over the past decade. This Roadmap is written collectively by prominent researchers and encompasses selected aspects of how machine learning is applied to microscopy image data, with the aim of gaining scientific knowledge by improved image quality, automated detection, segmentation, classification and tracking of objects, and efficient merging of information from multiple imaging modalities. We aim to give the reader an overview of the key developments and an understanding of possibilities and limitations of machine learning for microscopy. It will be of interest to a wide cross-disciplinary audience in the physical sciences and life sciences.

physics.optics↗

arXiv2023

Adaptive weighting of Bayesian physics informed neural networks for multitask and multiscale forward and inverse problems

In this paper, we present a novel methodology for automatic adaptive weighting of Bayesian Physics-Informed Neural Networks (BPINNs), and we demonstrate that this makes it possible to robustly address multi-objective and multi-scale problems. BPINNs are a popular framework for data assimilation, combining the constraints of Uncertainty Quantification (UQ) and Partial Differential Equation (PDE). The relative weights of the BPINN target distribution terms are directly related to the inherent uncertainty in the respective learning tasks. Yet, they are usually manually set a-priori, that can lead to pathological behavior, stability concerns, and to conflicts between tasks which are obstacles that have deterred the use of BPINNs for inverse problems with multi-scale dynamics. The present weighting strategy automatically tunes the weights by considering the multi-task nature of target posterior distribution. We show that this remedies the failure modes of BPINNs and provides efficient exploration of the optimal Pareto front. This leads to better convergence and stability of BPINN training while reducing sampling bias. The determined weights moreover carry information about task uncertainties, reflecting noise levels in the data and adequacy of the PDE model. We demonstrate this in numerical experiments in Sobolev training, and compare them to analytically $ε$-optimal baseline, and in a multi-scale Lokta-Volterra inverse problem. We eventually apply this framework to an inpainting task and an inverse problem, involving latent field recovery for incompressible flow in complex geometries.

physics.comp-ph↗

arXiv2022

Global Polynomial Level Sets for Numerical Differential Geometry of Smooth Closed Surfaces

We present a computational scheme that derives a global polynomial level set parametrisation for smooth closed surfaces from a regular surface-point set and prove its uniqueness. This enables us to approximate a broad class of smooth surfaces by affine algebraic varieties. From such a global polynomial level set parametrisation, differential-geometric quantities like mean and Gauss curvature can be efficiently and accurately computed. Even 4$^{\text{th}}$-order terms such as the Laplacian of mean curvature are approximates with high precision. The accuracy performance results in a gain of computational efficiency, significantly reducing the number of surface points required compared to classic alternatives that rely on surface meshes or embedding grids. We mathematically derive and empirically demonstrate the strengths and the limitations of the present approach, suggesting it to be applicable to a large number of computational tasks in numerical differential geometry.

math.NA↗

arXiv2022

Corvo: Visualizing CellxGene Single-Cell Datasets in Virtual Reality

The CellxGene project has enabled access to single-cell data in the scientific community, providing tools for browsed-based no-code analysis of more than 500 annotated datasets. However, single-cell data requires dimensional reduction to visualize, and 2D embedding does not take full advantage of three-dimensional human spatial understanding and cognition. Compared to a 2D visualization that could potentially hide gene expression patterns, 3D Virtual Reality may enable researchers to make better use of the information contained within the datasets. For this purpose, we present \emph{Corvo}, a fully free and open-source software tool that takes the visualization and analysis of CellxGene single-cell datasets to 3D Virtual Reality. Similar to CellxGene, Corvo takes a no-code approach for the end user, but also offers multimodal user input to facilitate fast navigation and analysis, and is interoperable with the existing Python data science ecosystem. In this paper, we explain the design goals of Corvo, detail its approach to the Virtual Reality visualization and analysis of single-cell data, and briefly discuss limitations and future extensions.

cs.HC↗

arXiv2022

Learning deterministic hydrodynamic equations from stochastic active particle dynamics

We present a principled data-driven strategy for learning deterministic hydrodynamic models directly from stochastic non-equilibrium active particle trajectories. We apply our method to learning a hydrodynamic model for the propagating density lanes observed in self-propelled particle systems and to learning a continuum description of cell dynamics in epithelial tissues. We also infer from stochastic particle trajectories the latent phoretic fields driving chemotaxis. This demonstrates that statistical learning theory combined with physical priors can enable discovery of multi-scale models of non-equilibrium stochastic processes characteristic of collective movement in living systems.

cond-mat.soft↗

arXiv2021

Parallel Discrete Convolutions on Adaptive Particle Representations of Images

We present data structures and algorithms for native implementations of discrete convolution operators over Adaptive Particle Representations (APR) of images on parallel computer architectures. The APR is a content-adaptive image representation that locally adapts the sampling resolution to the image signal. It has been developed as an alternative to pixel representations for large, sparse images as they typically occur in fluorescence microscopy. It has been shown to reduce the memory and runtime costs of storing, visualizing, and processing such images. This, however, requires that image processing natively operates on APRs, without intermediately reverting to pixels. Designing efficient and scalable APR-native image processing primitives, however, is complicated by the APR's irregular memory structure. Here, we provide the algorithmic building blocks required to efficiently and natively process APR images using a wide range of algorithms that can be formulated in terms of discrete convolutions. We show that APR convolution naturally leads to scale-adaptive algorithms that efficiently parallelize on multi-core CPU and GPU architectures. We quantify the speedups in comparison to pixel-based algorithms and convolutions on evenly sampled data. We achieve pixel-equivalent throughputs of up to 1 TB/s on a single Nvidia GeForce RTX 2080 gaming GPU, requiring up to two orders of magnitude less memory than a pixel-based implementation.

cs.CV↗

arXiv2021

Inverse-Dirichlet Weighting Enables Reliable Training of Physics Informed Neural Networks

We characterize and remedy a failure mode that may arise from multi-scale dynamics with scale imbalances during training of deep neural networks, such as Physics Informed Neural Networks (PINNs). PINNs are popular machine-learning templates that allow for seamless integration of physical equation models with data. Their training amounts to solving an optimization problem over a weighted sum of data-fidelity and equation-fidelity objectives. Conflicts between objectives can arise from scale imbalances, heteroscedasticity in the data, stiffness of the physical equation, or from catastrophic interference during sequential training. We explain the training pathology arising from this and propose a simple yet effective inverse-Dirichlet weighting strategy to alleviate the issue. We compare with Sobolev training of neural networks, providing the baseline of analytically $\boldsymbolε$-optimal training. We demonstrate the effectiveness of inverse-Dirichlet weighting in various applications, including a multi-scale model of active turbulence, where we show orders of magnitude improvement in accuracy and convergence over conventional PINN training. For inverse modeling using sequential training, we find that inverse-Dirichlet weighting protects a PINN against catastrophic forgetting.

cs.LG↗

arXiv2021

A Mathematical Definition of Particle Methods

We provide a formal definition for a class of algorithms known as "particle methods". Particle methods are used in scientific computing. They include popular simulation methods, such as Discrete Element Methods (DEM), Molecular Dynamics (MD), Particle Strength Exchange (PSE), and Smoothed Particle Hydrodynamics (SPH), but also particle-based image processing methods, point-based computer graphics, and computational optimization algorithms using point samples. All of these rest on a common concept, which we here formally define. The presented definition of particle methods makes it possible to distinguish what formally constitutes a particle method, and what not. It also enables us to define different sub-classes of particle methods that differ with respect to their computational complexity and power. Our definition is purely formal, independent of any application. After stating the definition, we therefore illustrate how several well-known particle methods can be formalized in our framework, and we show how the formal definition can be used to formulate novel particle methods for non-canonical problems.

cs.DS↗

arXiv2021

STENCIL-NET: Data-driven solution-adaptive discretization of partial differential equations

Numerical methods for approximately solving partial differential equations (PDE) are at the core of scientific computing. Often, this requires high-resolution or adaptive discretization grids to capture relevant spatio-temporal features in the PDE solution, e.g., in applications like turbulence, combustion, and shock propagation. Numerical approximation also requires knowing the PDE in order to construct problem-specific discretizations. Systematically deriving such solution-adaptive discrete operators, however, is a current challenge. Here we present STENCIL-NET, an artificial neural network architecture for data-driven learning of problem- and resolution-specific local discretizations of nonlinear PDEs. STENCIL-NET achieves numerically stable discretization of the operators in an unknown nonlinear PDE by spatially and temporally adaptive parametric pooling on regular Cartesian grids, and by incorporating knowledge about discrete time integration. Knowing the actual PDE is not necessary, as solution data is sufficient to train the network to learn the discrete operators. A once-trained STENCIL-NET model can be used to predict solutions of the PDE on larger spatial domains and for longer times than it was trained for, hence addressing the problem of PDE-constrained extrapolation from data. To support this claim, we present numerical experiments on long-term forecasting of chaotic PDE solutions on coarse spatio-temporal grids. We also quantify the speed-up achieved by substituting base-line numerical methods with equation-free STENCIL-NET predictions on coarser grids with little compromise on accuracy.

math.NA↗

arXiv2020

Letter to the Editor and Comments on: Intravital dynamic and correlative imaging reveals diffusion-dominated canalicular and flow-augmented ductular bile flux

Bile, the central metabolic product of the liver, is secreted by hepatocytes into bile canaliculi (BC), tubular subcellular structures of 0.5-2 $μ$m diameter which are formed by the apical membranes of juxtaposed hepatocytes. BC interconnect to build a highly ramified 3D network that collects and transports bile towards larger interlobular bile ducts (IBD). The transport mechanism of bile is of fundamental interest and the current text-book model of "osmotic canalicular bile flow", i.e. enforced by osmotic water influx, has recently been quantitatively verified based on flux measurements and BC contractility (Meyer et al., 2017) but challenged in the article entitled "Intravital dynamic and correlative imaging reveals diffusion-dominated canalicular and flow-augmented ductular bile flux" (Vartak et al., 2020). We feel compelled to share a series of arguments that question the key conclusions of this study (Vartak et al., 2020), particularly because it risks to be misinterpreted and misleading for the field in view of potential clinical applications. We summarized the arguments in the following 8 points.

q-bio.TO↗

arXiv2020

Learning physically consistent mathematical models from data using group sparsity

We propose a statistical learning framework based on group-sparse regression that can be used to 1) enforce conservation laws, 2) ensure model equivalence, and 3) guarantee symmetries when learning or inferring differential-equation models from measurement data. Directly learning $\textit{interpretable}$ mathematical models from data has emerged as a valuable modeling approach. However, in areas like biology, high noise levels, sensor-induced correlations, and strong inter-system variability can render data-driven models nonsensical or physically inconsistent without additional constraints on the model structure. Hence, it is important to leverage $\textit{prior}$ knowledge from physical principles to learn "biologically plausible and physically consistent" models rather than models that simply fit the data best. We present a novel group Iterative Hard Thresholding (gIHT) algorithm and use stability selection to infer physically consistent models with minimal parameter tuning. We show several applications from systems biology that demonstrate the benefits of enforcing $\textit{priors}$ in data-driven modeling.

cs.LG↗

arXiv2020

A robustness measure for singular point and index estimation in discretized orientation and vector fields

The identification of singular points or topological defects in discretized vector fields occurs in diverse areas ranging from the polarization of the cosmic microwave background to liquid crystals to fingerprint recognition and bio-medical imaging. Due to their discrete nature, defects and their topological charge cannot depend continuously on each single vector, but they discontinuously change as soon as a vector changes by more than a threshold. Considering this threshold of admissible change at the level of vectors, we develop a robustness measure for discrete defect estimators. Here, we compare different template paths for defect estimation in discretized vector or orientation fields. Sampling prototypical vector field patterns around defects shows that the robustness increases with the length of template path, but less so in the presence of noise on the vectors. We therefore find an optimal trade-off between resolution and robustness against noise for relatively small templates, except for the "single pixel" defect analysis, which cannot exclude zero robustness. The presented robustness measure paves the way for uncertainty quantification of defects in discretized vector fields.

cs.CV↗