An official website of the United States government
Summary We construct an algebraic multigrid (AMG) based preconditioner for the reduced Hessian of a linear‐quadratic optimization problem constrained by an elliptic partial differential equation. While the preconditioner generalizes a geometric multigrid preconditioner introduced in earlier works, its construction relies entirely on a standard AMG infrastructure built for solving the forward elliptic equation, thus allowing for it to be implemented using a variety of AMG methods and standard packages. Our analysis establishes a clear connection between the quality of the preconditioner and the AMG method used. The proposed strategy has a broad and robust applicability to problems with unstructured grids, complex geometry, and varying coefficients. The method is implemented using the Hypre package and several numerical examples are presented.
more »
« less
Comparison between algebraic and matrix‐free geometric multigrid for a Stokes problem on adaptive meshes with variable viscosity
Abstract Problems arising in Earth's mantle convection involve finding the solution to Stokes systems with large viscosity contrasts. These systems contain localized features which, even with adaptive mesh refinement, result in linear systems that can be on the order of 109or more unknowns. One common approach for preconditioning to the velocity block of these systems is to apply an Algebraic Multigrid (AMG) V‐cycle (as is done in the ASPECT software, for example), however, we find that AMG is lacking robustness with respect to problem size and number of parallel processes. Additionally, we see an increase in iteration counts with refinement when using AMG. In contrast, the Geometric Multigrid (GMG) method, by using information about the geometry of the problem, should offer a more robust option.Here we present a matrix‐free GMG V‐cycle which works on adaptively refined, distributed meshes, and we will compare it against the current AMG preconditioner (Trilinos ML) used in theASPECT1software. We will demonstrate the robustness of GMG with respect to problem size and show scaling up to 114,688 cores and 217 billion unknowns. All computations are run using the open‐source, finite element librarydeal.II.2
more »
« less
- PAR ID:
- 10449167
- Publisher / Repository:
- Wiley Blackwell (John Wiley & Sons)
- Date Published:
- Journal Name:
- Numerical Linear Algebra with Applications
- Volume:
- 28
- Issue:
- 5
- ISSN:
- 1070-5325
- Format(s):
- Medium: X
- Sponsoring Org:
- National Science Foundation
More Like this
-
-
We develop multigrid methods for an elliptic distributed optimal control problem on convex domains that are robust with respect to a regularization parameter. We prove the uniform convergence of the $$W$$-cycle algorithm and demonstrate the performance of $$V$$-cycle and $$W$$-cycle algorithms through numerical experiments.more » « less
-
Abstract Accurate quantification of soil carbon fluxes is essential to reduce uncertainty in estimates of the terrestrial carbon sink. However, these fluxes vary over time and across ecosystem types and so, it can be difficult to estimate them accurately across large scales. The flux‐gradient method estimates soil carbon fluxes using co‐located measurements of soil CO2concentration, soil temperature, soil moisture and other soil properties. The National Ecological Observatory Network (NEON) provides such data across 20 ecoclimatic domains spanning the continental U.S., Puerto Rico, Alaska and Hawai‘i.We present an R software package (neonSoilFlux) that acquires soil environmental data to compute half‐hourly soil carbon fluxes for each soil replicate plot at a given terrestrial NEON site. To assess the computed fluxes, we visited six focal NEON sites and measured soil carbon fluxes using a closed‐dynamic chamber approach.Outputs from theneonSoilFluxshowed agreement with measured fluxes (R2between measured andneonSoilFluxoutputs ranging from 0.12 to 0.77 depending on calculation method used); measured outputs generally fell within the range of calculated uncertainties from the gradient method. Calculated fluxes fromneonSoilFluxaggregated to the daily scale exhibited expected site‐specific seasonal patterns.While the flux‐gradient method is broadly effective, its accuracy is highly sensitive to site‐specific inputs, including the extent to which gap‐filing techniques are used to interpolate missing sensor data and to estimates of soil diffusivity and moisture content. Future refinement and validation ofneonSoilFluxoutputs can contribute to existing databases of soil carbon flux measurements, providing near real‐time estimates of a critical component of the terrestrial carbon cycle.more » « less
-
Artificial intelligence (AI) provides versatile capabilities in applications such as image classification and voice recognition that are most useful in edge or mobile computing settings. Shrinking these sophisticated algorithms into small form factors with minimal computing resources and power budgets requires innovation at several layers of abstraction: software, algorithmic, architectural, circuit, and device-level innovations. However, improvements to system efficiency may impact robustness and vice-versa. Therefore, a co-design framework is often necessary to customize a system for its given application. A system that prioritizes efficiency might use circuit-level innovations that introduce process variations or signal noise into the system, which may use software-level redundancy in order to compensate. In this tutorial, we will first examine various methods of improving efficiency and robustness in edge AI and their tradeoffs at each level of abstraction.Then, we will outline co-design techniques for designing efficient and robust edge AI systems, using federated learning as a specific example to illustrate the effectiveness of co-design.more » « less
-
Abstract Crustal magma chambers can grow to be hundreds to thousands of cubic kilometers, potentially feeding catastrophic caldera‐forming eruptions. Smaller volume chambers are expected to erupt frequently and freeze quickly; a major outstanding question is how magma chambers ever grow to the sizes required to sustain the largest eruptions on Earth. We use a thermo‐mechanical model to investigate the primary factors that govern the extrusive:intrusive ratio in a chamber, and how this relates to eruption frequency, eruption size, and long‐term chamber growth. The model consists of three fundamental timescales: the magma injection timescaleτin, the cooling timescaleτcool, and the timescale for viscous relaxation of the crustτrelax. We estimate these timescales using geologic and geophysical data from four volcanoes (Laguna del Maule, Campi Flegrei, Santorini, and Aso) to compare them with the model. In each of these systems,τinis much shorter thanτcooland slightly shorter thanτrelax, conditions that in the model are associated with efficient chamber growth and simultaneous eruption. In addition, the model suggests that the magma chambers underlying these volcanoes are growing at rates between ~10−4and 10−2 km3/year, speeding up over time as the chamber volume increases. We find scaling relationships for eruption frequency and size that suggest that as chambers grow and volatiles exsolve, eruption frequency decreases but eruption size increases. These scaling relationships provide a good match to the eruptive history from the natural systems, suggesting that the relationships can be used to constrain chamber growth rates and volatile saturation state from the eruptive history alone.more » « less
