Search NASA⌕ Search

SEARCH · Search NASA

Results for “Gradient descent”

Search indexed NASA NTRS and DOE OSTI research on propulsion, heat transfer, battery materials and energy systems. Follow report and document links to the original sources.

Quote a phrase for an exact phrase match. Source license links do not imply unrestricted reuse.

At least 307 records · Page 17

An analysis of the Antarctic Halogen Occultation Experiment trace gas observations

Analysis of the version 16 Halogen Occultation Experiment (HALOE) CH4 data shows that this long-lived trace gas is well correlated with potential vorticity (PV) computed from National Meteorological Center balanced winds. Analyzing late September and October 1992 data, we show that very low CH4 values are confined to the interior of a vortex edge defined by the maximum gradient in PV. The CH4 and HF time tendency is used to estimate the descent rate in the Antarctic vortex. After removing a component of the trend correlated with the HALOE sampling pattern, we compute the lower stratosphere vertical descent rates and net heaing rates in the spring Antarctic vortex. Our computations of the spring Antarctic vortex heating rates give -0.5 to -0.1 K/day. Over the winter season, the overall lower stratospheric descent rate averages about 1.8-1.5 km/month. These computations are in line with radiative transfer estimates of the heating and descent rate. The HALOE data thus appear to be consistent with the picture of an isolated lower stratospheric Antarctic vortex.

Schoeberl, Mark R.↗

Gradient Coding With Iterative Block Leverage Score Sampling

Gradient coding is a method for mitigating straggling servers in a centralized computing network that uses erasure-coding techniques to distributively carry out first-order optimization methods. Randomized numerical linear algebra uses randomization to develop improved algorithms for large-scale linear algebra computations. In this study, we propose a method for distributed optimization that combines gradient coding and randomized numerical linear algebra. The proposed method uses a randomized ℓ 2 -subspace embedding and a gradient coding technique to distribute blocks of data to the computational nodes of a centralized network, and at each iteration the central server only requires a small number of computations to obtain the steepest descent update. The novelty of our approach is that the data is replicated according to importance scores, called block leverage scores, in contrast to most gradient coding approaches that uniformly replicate the data blocks. Furthermore, we do not require a decoding step at each iteration, avoiding a bottleneck in previous gradient coding schemes. We show that our approach results in a valid ℓ 2 -subspace embedding, and that our resulting approximation converges to the optimal solution.

97 MATHEMATICS AND COMPUTING↗

ATMOS/ATLAS-3 Observations of Long-Lived Tracers and Descent in the Antarctic Vortex in November 1994

Observations of the long-lived tracers N2O, CH4 and HF obtained by the Atmospheric Trace Molecule Spectroscopy (ATMOS) instrument in early November 1994 are used to estimate average descent rates during winter in the Antarctic polar vortex of 0.5 to 1.5 km/month in the lower stratosphere, and 2.5 to 3.5 km/month in the middle and upper stratosphere. Descent rates inferred from ATMOS tracer observations agree well with theoretical estimates obtained using radiative heating calculations. Air of mesospheric origin (N2O less than 5 ppbV) was observed at altitudes above about 25 km within the vortex. Strong horizontal gradients of tracer mixing ratios, the presence of mesospheric air in the vortex in early spring, and the variation with altitude of inferred descent rates indicate that the Antarctic vortex is highly isolated from midlatitudes throughout the winter from approximately 20 km to the stratopause. The 1994 Antarctic vortex remained well isolated between 20 and 30 km through at least mid-November.

Abrams, M. C.↗

Algorithms for Mathematical Programming with Emphasis on Bi-level Models (Final Scientific/Technical Report)

The research supported by this grant was focused primarily on first-order methods for solving large scale and structured convex optimization problems and convex relaxations of nonconvex problems. These include optimal gradient methods, operator and variable splitting methods, alternating direction augmented Lagrangian methods, and block coordinate descent methods.

97 MATHEMATICS AND COMPUTING↗

Dynamics of coronal rain and descending plasma blobs in solar prominences. I. Fully ionized case

Observations of active regions and limb prominences often show cold, dense blobs descending with an acceleration smaller than that of free fall. The dynamics of these condensations falling in the solar corona is investigated in this paper using a simple fully ionized plasma model. We find that the presence of a heavy condensation gives rise to a dynamical rearrangement of the coronal pressure that results in the formation of a large pressure gradient that opposes gravity. Eventually this pressure gradient becomes so large that the blob acceleration vanishes or even points upward. Then, the blob descent is characterized by an initial acceleration phase followed by an essentially constant velocity phase. These two stages can be identified in published time-distance diagrams of coronal rain events. Both the duration of the first stage and the velocity attained by the blob increase for larger values of the ratio of blob to coronal density, for larger blob mass, and for smaller coronal temperature. Dense blobs are characterized by a detectable density growth (up to 60% in our calculations) and by a steepening of the density in their lower part, that could lead to the formation of a shock. They also emit sound waves that could be detected as small intensity changes with periods of the order of 100 s and lasting between a few and about 10 periods. Finally, the curvature of falling paths with large radii is only relevant when a very dense blob falls along inclined magnetic field lines.

79 ASTRONOMY AND ASTROPHYSICS↗

North Pacific Cloud Feedbacks Inferred from Synoptic-Scale Dynamic and Thermodynamic Relationships

This study analyzed daily satellite cloud observations and reanalysis dynamical parameters to determine how mid-tropospheric vertical velocity and advection over the sea surface temperature gradient control midlatitude North Pacific cloud properties. Optically thick clouds with high tops are generated by synoptic ascent, but two different cloud regimes occur under synoptic descent. When vertical motion is downward during summer, extensive stratocumulus cloudiness is associated with near surface northerly wind, while frequent cloudless pixels occur with southerly wind. Examinations of ship-reported cloud types indicates that midlatitude stratocumulus breaks up as the the boundary level decouples when it is advected equatorward over warmer water. Cumulus is prevalent under conditions of synoptic descent and cold advection during winter. Poleward advection of subtropical air over colder water causes stratification of the near-surface layer that inhibits upward mixing of moisture and suppresses cloudiness until a fog eventually forms. Averaging of cloud and radiation data into intervals of 500-hPa vertical velocity and advection over the SST gradient enables the cloud response to changes in temperature and the stratification of the lower troposphere to be investigated independent of the dynamics.

Norris, Joel R.↗

N2O as a dynamical tracer in the Arctic vortex

This paper reports N2O measurements obtained by the Airborne Tunable Laser Absorption Spectrometer from 14 flights of the NASA ER-2 aircraft during the 1989 Airborne Arctic Stratospheric Expedition field campaign. In the altitude range expected for ozone loss, N2O has a long photochemical lifetime, making it an excellent tracer of lower stratospheric air motions. As in the southern hemisphere, the zonal wind speed maximum and large gradients of potential vorticity and N2O identify the vortex edge. The N2O profiles inside the vortex indicate net descent relative to outside the vortex and to the summer polar lower stratosphere. The descent of the N2O profile during the Arctic night relative to the summer profile is comparable to the downward shift in the vertical profile observed in the 1987 Antarctic winter vortex. Winter profiles at the poles are very similar above the 435 K potential temperature surface, but divergent below.

Loewenstein, M.↗

Dynamic load balancing algorithm for molecular dynamics based on Voronoi cells domain decompositions

We present a new algorithm for automatic parallel load balancing in classical molecular dynamics. It assumes a spatial domain decomposition of particles into Voronoi cells. It is a gradient method which attempts to minimize a cost function by displacing Voronoi sites associated with each processor/sub-domain along steepest descent directions. Excellent load balance has been obtained for quasi-2D and 3D practical applications, with up to 440·10 6 particles on 65,536 MPI tasks.

75 CONDENSED MATTER PHYSICS, SUPERCONDUCTIVITY AND↗

Path optimization by a variational reaction coordinate method. I. Development of formalism and algorithms

The development of algorithms to optimize reaction pathways between reactants and products is an active area of study. Existing algorithms typically describe the path as a discrete series of images (chain of states) which are moved downhill toward the path, using various reparameterization schemes, constraints, or fictitious forces to maintain a uniform description of the reaction path. The Variational Reaction Coordinate (VRC) method is a novel approach that finds the reaction path by minimizing the variational reaction energy (VRE) of Quapp and Bofill. The VRE is the line integral of the gradient norm along a path between reactants and products and minimization of VRE has been shown to yield the steepest descent reaction path. In the VRC method, we represent the reaction path by a linear expansion in a set of continuous basis functions and find the optimized path by minimizing the VRE with respect to the linear expansion coefficients. Improved convergence is obtained by applying constraints to the spacing of the basis functions and coupling the minimization of the VRE to the minimization of one or more points along the path that correspond to intermediates and transition states. The VRC method is demonstrated by optimizing the reaction path for the Müller-Brown surface and by finding a reaction path passing through 5 transition states and 4 intermediates for a 10 atom Lennard-Jones cluster.

37 INORGANIC, ORGANIC, PHYSICAL, AND ANALYTICAL CH↗

Steepest Descent

The steepest descent method has a rich history and is one of the simplest and best known methods for minimizing a function. While the method is not commonly used in practice due to its slow convergence rate, understanding the convergence properties of this method can lead to a better understanding of many of the more sophisticated optimization methods. Here, we give a short introduction and discuss some of the advantages and disadvantages of this method. Some recent results on modified versions of the steepest descent method are also discussed.

97 MATHEMATICS AND COMPUTING↗

A General Method for Solving Systems of Non-Linear Equations

The method of steepest descent is modified so that accelerated convergence is achieved near a root. It is assumed that the function of interest can be approximated near a root by a quadratic form. An eigenvector of the quadratic form is found by evaluating the function and its gradient at an arbitrary point and another suitably selected point. The terminal point of the eigenvector is chosen to lie on the line segment joining the two points. The terminal point found lies on an axis of the quadratic form. The selection of a suitable step size at this point leads directly to the root in the direction of steepest descent in a single step. Newton's root finding method not infrequently diverges if the starting point is far from the root. However, the current method in these regions merely reverts to the method of steepest descent with an adaptive step size. The current method's performance should match that of the Levenberg-Marquardt root finding method since they both share the ability to converge from a starting point far from the root and both exhibit quadratic convergence near a root. The Levenberg-Marquardt method requires storage for coefficients of linear equations. The current method which does not require the solution of linear equations requires more time for additional function and gradient evaluations. The classic trade off of time for space separates the two methods.

Nachtsheim, Philip R.↗

Northern hemisphere nitrous oxide morphology during the 1989 AASE and the 1991-1992 AASE 2 campaigns

Nitrous oxide vertical profiles and latitudinal distributions for the 1989 Airborne Antarctic Ozone Experiment (AASE) and 1992 AASE 2 northern polar winters are developed from the ATLAS N2O dataset, using both potential temperature and pressure as vertical coordinates. Morphologies show strong descent occuring poleward of the polar jet. The AASE 2 morphology shows a mid latitude 'surf zone', characterized by strong horizontal mixing, and a horizontal gradient south of 30 deg N due to the sub-tropical jet. These features are similar to those produced by two-dimensional photochemical models which include coupling between transport, radiation, and chemistry.

Podolske, James R.↗

Northern Hemisphere Nitrous Oxide Morphology during the 1989 AASE and the 1991-1992 AASE 2 Campaigns

Nitrous oxide vertical profiles and latitudinal distributions for the 1989 AASE and 1992 AASE II northern polar winters are developed from the ATLAS N2O dataset, using both potential temperature and pressure as vertical coordinates. Morphologies show strong descent occurring poleward of the polar jet. The AASE II morphology shows a mid latitude 'surf zone,' characterized by strong horizontal mixing, and a horizontal gradient south of 30 deg N due to the sub-tropical jet. These features are similar to those produced by two-dimensional photochemical models which include coupling between transport, radiation, and chemistry.

Podolske, James R.↗

Randomized Preconditioned Solvers for Strong Constraint 4D-Var Data Assimilation

The Strong Constraint 4D Variational (SC-4DVAR) data assimilation method is widely used in climate and weather applications. SC-4DVAR involves solving a minimization problem to compute the maximum a posteriori estimate, which we tackle using the Gauss-Newton method. The computation of the descent direction is expensive since it involves the solution of a large-scale and potentially ill-conditioned linear system, solved using the preconditioned conjugate gradient (PCG) method. Here, to address this cost, we efficiently construct scalable preconditioners using three different randomization techniques, which all rely on a certain low-rank structure involving the Gauss-Newton Hessian. The proposed techniques come with theoretical guarantees on the condition number, and at the same time, are amenable to parallelization. We also develop an adaptive approach to estimate the sketch size and choose between the reuse or recomputation of the preconditioner. We demonstrate the performance and effectiveness of our methodology on two representative model problems—the Burgers and barotropic vorticity equation—showing a drastic reduction in both the number of PCG iterations and the number of Gauss-Newton Hessian products after including the preconditioner construction cost.

Gauss-Newton↗

Determination of atmospheric properties for STS-1 aerothermodynamic investigations

A procedure for determining an approximation to the freestream atmospheric properties along the Shuttle entry trajectory is presented. Meteorological data as input is obtained by rawinsondes from surface to 70 km, and meteorological spheres from 60-90 km, launched from Hawaii and California. The Langley Atmospheric Information Retrieval System (LAIRS) developed to approximate the atmospheric freestream properties along the flight path, is outlined, noting temperature and wind data are interpolated in altitude, while gradients and diurnal and semidiurnal coefficients are taken from the COSPAR reference atmosphere. The data are input to a model to project temperature profiles for the Shuttle descent, and the input atmospheric parameters are listed. Efforts are continuing in order to correct discrepancies in the generated profiles for regions below 3 km.

Price, J. M.↗

Rocket measurements of electron density irregularities during MAC/SINE

Four Super Arcas rockets were launched at the Andoya Rocket Range, Norway, as part of the MAC/SINE campaign to measure electron density irregularities with high spatial resolution in the cold summer polar mesosphere. They were launched as part of two salvos: the turbulent/gravity wave salvo (3 rockets) and the EISCAT/SOUSY radar salvo (one rocket). In both salvos meteorological rockets, measuring temperature and winds, were also launched and the SOUSY radar, located near the launch site, measured mesospheric turbulence. Electron density irregularities and strong gradients were measured by the rocket probes in the region of most intense backscatter observed by the radar. The electron density profiles (8 to 4 on ascent and 4 on descent) show very different characteristics in the peak scattering region and show marked spatial and temporal variability. These data are intercompared and discussed.

Ulwick, J. C.↗