Engineering Papers⌕ Search

SEARCH · Engineering Papers

Results for “eigenvalue analysis”

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 667 records · Page 37
All sourcesNASA NTRSDOE OSTI
DOE OSTI2024

Lattice realizations of topological defects in the critical (1+1)-d three-state Potts model

Topological/perfectly-transmissive defects play a fundamental role in the analysis of the symmetries of two dimensional conformal field theories (CFTs). In the present work, spin chain regularizations for these defects are proposed and analyzed in the case of the three-state Potts CFT. In particular, lattice versions for all the primitive defects are presented, with the remaining defects obtained from the fusion of the primitive ones. The defects are obtained by introducing modified interactions around two given sites of an otherwise homogeneous spin chain with periodic boundary condition. The various primitive defects are topological on the lattice except for one, which is topological only in the scaling limit. The lattice models are analyzed using a combination of exact diagonalization and density matrix renormalization group techniques. Low-lying energy spectra for different defect Hamiltonians as well as entanglement entropy of blocks located symmetrically around the defects are computed. The latter provides a convenient way to compute the g-function which characterizes various defects. Finally, the eigenvalues of the line operators in the “crossed channel” and fusion of different defect lines are also analyzed. The results are all in agreement with expectations from conformal field theory.

75 CONDENSED MATTER PHYSICS, SUPERCONDUCTIVITY AND↗
DOE OSTI2025

Deployment of neural-network-based neutron microscopic cross sections in the Griffin reactor physics application

The capability to utilize neural networks to predict macroscopic and microscopic cross section parametric spaces has been developed for the Griffin reactor physics application. The LibTorch interface enables Griffin's MOOSE-based materials to interact with LibTorch-trained models, allowing for the evaluation of complex macroscopic or microscopic cross section spaces, which are then used to evaluate the neutronic properties of the Griffin finite element model. This study benchmarks traditional ISOXML-formatted tabulation libraries against neural network-based models for 279 nuclides on 20,160 grid points for zero-dimensional and two-dimensional reactor models. Benchmark metrics include the fundamental mode eigenvalue, fission and absorption rates, and various temperature coefficients of reactivity (isothermal, fuel, and moderator). From the perspective of storage space, the complete set of LibTorch models uses 11 MB on disk, compared to the 10 GB for the ISOXML multigroup library that covers the same grid space. For the two-dimensional performance case considered in Griffin, the Torch model uses 97% less RAM than the reference ISOXML dataset while runtime increases by a factor of 3 when using the LibTorch model compared to the ISOXML dataset with multi-linear interpolation. The LibTorch model consistently yields errors within 0.01% for most analyzed quantities except for the temperature coefficients of reactivity where the maximum discrepancies are up to 0.3 $\frac{pcm}{K}$. Due to the neural network attempting to best predict quantities with no regard for a positive or negative bias for any given quantity, predictions may experience random fluctuations, resulting in both positive and negative errors. Future work will entail both depletion and coupled transient analysis to determine the predictive capabilities of Griffin with neural network-based cross sections.

22 GENERAL STUDIES OF NUCLEAR REACTORS↗
DOE OSTI2023

Improved Fast Reactor Capability of Griffin in FY23

Griffin is a MOOSE based reactor multiphysics analysis application jointly developed by Idaho National Laboratory and Argonne National Laboratory under the DOE-NE NEAMS program. In FY23, we enhanced capabilities required for fast reactor analysis. This effort included primarily updating the cross-section generation workflow using MC2-3 for various reactor configurations, such as homogeneous, duct-heterogeneous, ring-heterogeneous, and fully-heterogeneous geometries. In addition, we initiated the implementation of a multi-cycle depletion and shuffling capability. To support fast reactor simulation capabilities, we significantly improved the performance of the DFEM-SN-based R-Z transport solver to efficiently solve ultrafine group (over 1000 groups) transport problems. Additionally, the performance of HFEM-PN was improved by introducing red-black iteration, the cmfd acceleration technique, and various optimizations. We also completed the pin power reconstruction capability to support multiphysics simulations while identifying and addressing issues associated with SPH equivalence parameter approach. These enhanced capabilities for fast reactor core simulations, specially HFEM-PN and pin power reconstruction features, were applied to benchmark problems involving ABTR and ABR-1000. These applications showcased excellent agreement with Monte Carlo and other code solutions in terms of eigenvalue, control rod worth, and assembly and pin powers.

22 GENERAL STUDIES OF NUCLEAR REACTORS↗
DOE OSTI2024

Small-Signal Stability Impacts of Load and Network Dynamics on Grid-Forming Inverters

This paper presents several stability analyses for grid-forming inverters and synchronous generators considering the dynamics of transmission lines and different load models. Load models are usually of secondary importance compared to generation source models, but as the results show, they play a crucial role in stability studies with the introduction of inverter-based resources. Given inverter control time scales, the implications of considering or neglecting electromagnetic transients of the network are very relevant in the stability assessments. In this paper, we perform eigenvalue analyses for inverter-based resources and synchronous machines connected to a load and explore the effects of multiples models under different network representations. We explore maximum loadability of inverter-based resources and synchronous machines, while analyzing the effects of load and network dynamic models on small-signal stability. The results show that the network representation plays a fundamental role in the stability of the system of different load models. The resulting stability regions are significantly different depending on the source and load model considered.

grid forming inverters↗
DOE OSTI2021

Nuclear Data Sensitivity Study for the EBR-II Fast Reactor Benchmark Using SCALE with ENDF/B-VII.1 and ENDF/B-VIII.0

The EBR-II benchmark, which was recently included in the International Handbook of Evaluated Reactor Physics Benchmark Experiments, served as a basis for assessing the performance of the SCALE code system for fast reactor analyses. A reference SCALE model was developed based on the benchmark specifications. Great agreement was observed between the eigenvalue calculated with this SCALE model and the benchmark eigenvalue. To identify potential gaps and uncertainties of nuclear data for the simulation of various quantities of interest in fast spectrum systems, sensitivity and uncertainty analyses were performed for the eigenvalue, reactivity effects, and the radial power profile of EBR-II using the two most recent ENDF/B nuclear data library releases. While the nominal results are consistent between the calculations with the different libraries, the uncertainties due to nuclear data vary significantly. The major driver of observed uncertainties is the uncertainty of the 235 U ( n,γ ) reaction. Since the uncertainty of this reaction is significantly reduced in the ENDF/B-VIII.0 library compared to ENDF/B-VII.1, the obtained output uncertainties tend to be smaller in ENDF/B-VIII.0 calculations, although the decrease is partially compensated by increased uncertainties in 235 U fission and ν ¯ .

42 ENGINEERING↗
NASA NTRSDate not supplied

High-Fidelity Flight Dynamic Analysis of Transonic Truss-Braced Wing

This paper presents a high-fidelity flight dynamic analysis of the Mach 0.8 Transonic Truss-Braced Wing (TTBW). Unsteady RANS CFD simulations of the Mach 0.8 TTBW in pitch, plunge, roll, and yaw oscillations are conducted in FUN3D. The time-domain data are transformed into the frequency-domain data by Fourier series. Transfer functions of the dynamic stability derivatives are then estimated by a frequency-domain re- gression. The dynamic stability derivatives with respect to the angle of attack are determined by the regression of the unsteady aerodynamic coefficients for the plunge motion. The dynamic stability derivatives with re- spect to the pitch rate are determined by the regression of the differential unsteady aerodynamic coefficients for the pitch motion upon the removal of the angle of attack contribution by the plunge motion. Similarly, dynamic stability derivatives with respect to the angle of sideslip, roll rate, and yaw rates are determined from the frequency domain regression. The longitudinal and lateral-direction flight dynamic models of the Mach 0.8 TTBW are constructed from these dynamic stability derivatives. The eigenvalues of the aircraft modes are analyzed to determine the vehicle stability.

Aircraft Stability and Control↗
NASA NTRSDate not supplied

Stability Analysis of the Flow over a Swept Forward-Facing Step using PIV Base Flowsin a Non-Orthogonal Coordinate System

Understanding the flow of a crossflow-vortex-dominated boundary layer over a forward-facing step excrescence is necessary in order to mitigate an early transition scenario on the wings and tail of commercial aircraft. By performing BiGlobal stability analysis on the flowfield measured with high-resolution, stereographic Particle Image Velocimetry (PIV), previous work has shown that a family of unstable disturbances exists in the direct downstream vicinity of a supercritical, forward-facing step. The goal of the present paper is to improve the previously used stability approach on two fronts: 1) the stability problem is formulated in a non-orthogonal coordinate system and 2) the contribution of the out-of-plane base-flow derivatives is accounted for locally. The resulting eigen solutions feature significantly stronger growth rates and eigenfunctions that are localized above regions of reverse flow. The changes in the problem formulation furthermore establish a significant improvement in the comparison of the stability solutions with the results from the Spectral Proper Orthogonal Decomposition (SPOD) of a time-resolved measurement of the perturbation content. A large destabilizing effect by the out-of-plane base-flow derivatives is determined to be justified, despite the assumption that these derivatives are small, because the effect can be reconstructed upon using an eigenvalue-correction formula that assumes the responsible terms are infinitesimally small. The unstable perturbation mechanisms are demonstrated to have a convective nature by assessing the relation between the frequency and the out-of-plane wavenumber, which indicates that their group speed does not approach zero. The last 2 facts, that 1) the perturbations are convective and 2) that the out-of-plane base-flow-derivative terms are small, remove any qualitative suspicions that the perturbation problem is not conducive to parabolization. This opens the path to analyzing this problem with a plane-marching approach.

Boundary- Layer Transition↗
DOE OSTI2022

Impact of Increased Latent Generations on Sensitivity Calculations with SCALE

Analyses of cross section sensitivity data from systems with fissile material allow analysts to associate an importance for each material, nuclide, reaction, and neutron energy by simulating real world criticality scenarios. Although criticality safety validation efforts can be guided by the cross-section sensitivity and uncertainty data generated for a particular system, these calculations can often be computationally expensive and sometimes cumbersome without proper guidance. The TSUNAMI suite within the SCALE code package has several methods for generating sensitivity data, including multigroup and continuous energy (CE) capabilities. The release of SCALE 6.3 has three different CE methods for generating cross section sensitivity data: (1) the Iterated Fission Probability (IFP) method with the KENO Monte Carlo transport solver, (2) the IFP method with the Shift Monte Carlo transport solver, and (3) the Contributon-Linked eigenvalue sensitivity/Uncertainty estimation via Tracklength importance CHaracterization (CLUTCH) method with the KENO Monte Carlo transport solver. Although the CLUTCH method has additional parameters for generating sensitivity data files relative to the IFP method, all three methods use latent generations, which are the generations between an event (i.e., fission) and the assessment of importance based on the asymptotic population of progeny neutrons. Increasing the number of latent generations in a calculation leads to increased discrimination of the sensitivity coefficients but at the cost of the increased uncertainty associated with those generated values. Analysts must balance the accuracy of the sensitivity calculations and its uncertainty with the associated computational cost involved in generating the values. This paper discusses the impact of adjusting the latent generation parameter for a range of sensitivity values and how these changes compare with the direct perturbation values obtained from a change of ±0.5% Δ k in both benchmark and safety application models. Two benchmarks from the International Handbook of Evaluated Criticality Safety Benchmark Experiments and the MPC-32 dual purpose canister for spent nuclear fuel are used for analysis.

73 NUCLEAR PHYSICS AND RADIATION PHYSICS↗
NASA NTRS1995

The Physics of Turbulence in the Boundary Layer

The geometry of the velocity field in a numerically simulated incompressible turbulent boundary layer over a flat plate at Re theta=670 has been studied using the invariants of the velocity gradient tensor. These invariants are computed at every grid point in the flow and used to form the discriminant. Of primary interest are those regions in the flow where the discriminant is positive; regions where, according to the characteristic equation, the eigenvalues of the velocity gradient tensor are complex. An observer moving with a frame of reference which is attached to a fluid particle lying within such a region would see a local flow pattern of the type stable-focus-stretching or unstable-focus-compressing. When the flow is visualized this way, continuous, connected, large-scale structures are revealed that extend from the point just below the buffer layer out to the beginning of the wake region. These structures are aligned with the mean shear close to the wall and arch in the cross-stream direction away from the wall. In some cases the structures observed are very similar to to the hairpin eddy vision of boundary layer structure proposed by Theodorsen. That the structure of the flow is revealed more effectively by the discriminant rather than by the vorticity is important and adds support to recent observations of the discriminant in a channel flow simulation. Of particular importance is the fact that the procedure does not require the use of an arbitrary threshold in the discriminant. Further analysis using computer flow visualization shows a high degree of spatial correlation between regions of positive discriminant, extreme negative pressure fluctuations and large instantaneous values of Reynolds shear stress.

Kline, Stephen↗
DOE OSTI2023

Deep Learning for Multigroup Cross-Section Representation in Two-Step Core Calculations

Here we investigate using deep learning, a type of machine-learning algorithm employing multiple layers of artificial neurons, for the mathematical representation of multigroup cross sections for use in the Griffin reactor multiphysics code for two-step deterministic neutronics calculations. A three-dimensional fuel element typical of a high-temperature gas reactor as well as a two-dimensional sodium-cooled fast reactor lattice are modeled using the Serpent Monte Carlo code, and multigroup macroscopic cross sections are generated for various state parameters to produce a training data set and a separate validation data set. A fully connected, feedforward neural network is trained using the open-source PyTorch machine-learning framework, and its accuracy is compared against the standard piecewise linear interpolation model. Additionally, we provide in this work a generic technique for propagating the cross-section model errors up to the k eff using sensitivity coefficients with the first-order uncertainty propagation rule. Quantifying the eigenvalue error due to the cross-section regression errors is especially practical for appropriately selecting the mathematical representation of the cross sections. We demonstrate that the artificial neural network model produces lower errors and therefore enables better accuracy relative to the piecewise linear model when the cross sections exhibit nonlinear dependencies; especially when a coarse grid is employed, where the errors can be halved by the artificial neural network. However, for linearly dependent multigroup cross sections as found for the sodium-cooled fast reactor case, a simpler linear regression outperforms deeper networks.

22 GENERAL STUDIES OF NUCLEAR REACTORS↗
DOE OSTI2021

MACHINE LEARNING BASED CHEMICAL EXPLOSIVE MODE ANALYSIS (ML-CEMA)

The software consists of a Machine Learning based Chemical Explosive Mode Analysis (ML-CEMA) tool for advanced computational flame diagnostics. CEMA, originally based on eigen-analysis of the local thermochemical system, is capable of identifying reaction fronts and limit phenomena such as auto-ignition and extinction in practical combustion systems, such as internal combustion engines and gas turbine combustors. However, the original CEMA is computationally expensive for large reaction mechanisms that are typically needed to describe fuel chemistry of practical large-hydrocarbon fuels. This novel ML-CEMA tool employs a ML technique to accelerate the eigen-analysis of the basic CEMA approach by orders of magnitude, thus making it suitable for practical fuels. In ML-CEMA,zero-dimensional (0D) reactors and one-dimensional (1D) premixed flames are first used to generate a large number of data points for neural network based ML training. The trained ML model is then used to perform CEMA prediction. This ML-CEMA tool has been demonstrated in canonical 0D and 1D configurations as well as highly-transient three-dimensional spray flames exhibiting multi-mode turbulent combustion, showing promising results. ML-CEMA, as a standalone tool, can be used for computationally-efficient diagnostics of massive datasets generated from both experiments and simulations. For example, based on spatially resolved measurements of a small set of reactive scalars(such as temperature, hydroxyl radical and formaldehyde), ML-CEMA can effectively identify flame fronts and rare events. ML-CEMA also provides a robust online or offline flame feature detection tool. When used for on-the-fly simulations, ML-CEMA further enables zone-adaptive combustion modeling, in which the predicted eigenvalue is used as a robust mode indicator for judicious assignment of locally-valid combustion models. This ML-CEMA based zone-adaptive model can lead to substantial computational cost savings when used for large-scale simulations of multi-mode combustion systems. Third Party Code Web Page to Download Code Web Page Location of Third Party License

Xu, Chao↗
DOE OSTI2020

Band gap renormalization, carrier mobilities, and the electron-phonon self-energy in crystalline naphthalene

Organic molecular crystals are expected to feature appreciable electron-phonon interactions that influence their electronic properties at zero and finite temperature. Here, we report first-principles calculations and an analysis of the electron-phonon self-energy in naphthalene crystals. We compute the zero-point renormalization and temperature dependence of the fundamental band gap, and the resulting scattering lifetimes of electronic states near the valence- and conduction-band edges employing density functional theory. Further, our calculated phonon renormalization of the GW-corrected quasiparticle band structure predicts a fundamental band gap of 5 eV for naphthalene at room temperature, in good agreement with experiments. From our calculated phonon-induced electron lifetimes, we obtain the temperature-dependent mobilities of electrons and holes in good agreement with experimental measurements at room temperature. Finally, we show that an approximate energy self-consistent computational scheme for the electron-phonon self-energy leads to the prediction of strong satellite bands in the electronic band structure. We find that a single calculation of the self-energy can reproduce the self-consistent results of the band gap renormalization and electrical mobilities for naphthalene, provided that the on-the-mass-shell approximation is used, i.e., if the self-energy is evaluated at the bare eigenvalues.

75 CONDENSED MATTER PHYSICS, SUPERCONDUCTIVITY AND↗
DOE OSTI2021

Electronic structure of chromium trihalides beyond density functional theory

In this work, we explore the electronic band structure of freestanding monolayers of chromium trihalides Cr X 3 , X = Cl, Br, I, within an advanced ab initio theoretical approach based on the use of Green's function functionals. We compare the local density approximation with the quasiparticle self-consistent GW (QS GW ) approximation and its self-consistent extension ( QS G W ^ ) by solving the particle-hole ladder Bethe-Salpeter equations to improve the effective interaction W . We show that, at all levels of theory, the valence band consistently changes shape in the sequence Cl → Br → I , and the valence band maximum shifts from the M point to the Γ point. By analyzing the dynamic and momentum-dependent self-energy, we show that QS G W ^ adds to the localization of the systems in comparison with QS GW , thereby leading to a narrower band and reduced amount of halogens in the valence band manifold. Further analysis shows that X = Cl is most strongly correlated, and X = I is least correlated (most bandlike) as the hybridization between Cr d and X p enhances in the direction Cl → Br → I . For CrBr 3 and CrI 3 , we observe remarkable differences between the QS GW and QS G W ^ valence band structures, while their eigenfunctions are very similar. We show that weak perturbations, like moderate strain, weak changes to the d - p hybridization, and adding small U , can flip the valence band structures between these two solutions in these materials.

75 CONDENSED MATTER PHYSICS, SUPERCONDUCTIVITY AND↗
NASA NTRS1995

Temporal variation of meandering intensity and domain-wide lateral oscillations of the Gulf Stream

The path of the Gulf Stream exhibits two modes of variability: wavelike spatial meanders associated with instability processes and large-sale lateral shifts of the path presumably due to atmospheric forcing. The objectives of this study are to examine the temporal variation of the intensity of spatial meandering in the stream, to characterize large-scale lateral oscillations in the stream's path, and to study the correlation betwen these two dynamically distinct modes of variability. The data used for this analysis are path displacemets ofthe Gulf Stream between 75 deg and 60 deg W obtained from AVHRR-derived (Advanced Very High Resolution Radiometer) infrared images for the period April 1982 through December 1989. Meandering intensity, measured by the spatial root-mean-sqaure displacement of the stream path, displays a 9-month dominant periodicity which is persistent through the study period. The 9-month fluctuation in meandering intensity may be related to the interaction of Rosseby waves with the stream. Interannual variation of meandering intensity is also found to be significant, with meandering being mich more intense during 1985 than it was in 1987. Annual variation, however,is weak and not well-defined.The spatially averaged position of the stream, which reflects nonmeandering large-scale lateral oscillations of the stream path, is dominated by an annual cycle. On average, the mean position is farthest north in November and farthest south in April. The first empirical orthogonal function mode of the space-time path displacements represents lateral oscillatins that are in-phase over the space-time domain. Interannual oscillations are also observed and are found to be weaker than the annual oscillation. The eigenvalue of the first mode indicates that about 21.5% of the total space-time variability of the stream path can be attibuted to domain-wide lateral oscillation. The correlation between meandering intensity and domain-wide lateral oscillations is very weak.

Lee, Tong↗
NASA NTRS2001

Cross-Axis Proportional Gains Used to Control Gyroscopic Effects in a Magnetic- Bearing-Supported Flywheel

For magnetic-bearing-supported high-speed machines with significant gyroscopic effects, it is necessary to stabilize both forward and backward tilt whirling modes. Instability or the low damping of these modes can prevent the attainment of desired shaft speeds. Previous work elsewhere showed that cross-axis derivative gain in the magnetic bearing control law can improve the stability of the forward whirl mode, but it is commonly recognized that derivative gains amplify high-frequency noise and increase the required control effort. At the NASA Glenn Research Center, it has been shown previously that a simple cross-axis proportional gain can add stability (without adding noise) to either forward whirl or backward whirl, depending on the sign of the gain, but that such a gain destabilizes the other mode. It has been predicted by Glenn analysis that both modes can be stabilized by cross-axis proportional gains by utilizing the large-frequency separation of the two modes at speeds where the gyroscopic effects are significant. We use a modal controller that decouples the tilt and center-of-mass-translation modes. Only the tilt modes exhibit speed-dependent gyroscopic effects. The key to controlling them by the present method is to stabilize the backward whirl tilt mode with the appropriate sign of cross-axis proportional gain in the control law, but to include a low-pass filter on that gain term to restrict its effect only to the low-frequency backward-whirl mode. A second cross-axis term with the opposite sign and a high-pass filter stabilizes the forward whirl, which can have a frequency one or two orders of magnitude higher than the backward whirl, permitting very independent action of the two terms. Because the physical gyroscopic torques are proportional to the spinning speed of the shaft, it is convenient to gain-schedule the cross-axis control terms by making them proportional to shaft speed. This has the added benefit of avoiding a somewhat awkward zero-speed splitting of the tilt-mode eigenvalues.

Brown, Gerald V.↗
DOE OSTI2022

A numerical approach for detecting switch-like bistability in mass action chemical reaction networks with conservation laws

Theoretical analysis of signaling pathways can provide a substantial amount of insight into their function. One particular area of research considers signaling pathways capable of assuming two or more stable states given the same amount of signaling ligand. This phenomenon of bistability can give rise to switch-like behavior, a mechanism that governs cellular decision making. Investigation of whether or not a signaling pathway can confer bistability and switch-like behavior, without knowledge of specific kinetic rate constant values, is a mathematically challenging problem. Recently a technique based on optimization has been introduced, which is capable of finding example parameter values that confer switch-like behavior for a given pathway. Although this approach has made it possible to analyze moderately sized pathways, it is limited to reaction networks that presume a uniterminal structure. It is this limited structure we address by developing a general technique that applies to any mass action reaction network with conservation laws. In this paper we developed a generalized method for detecting switch-like bistable behavior in any mass action reaction network with conservation laws. The method involves (1) construction of a constrained optimization problem using the determinant of the Jacobian of the underlying rate equations, (2) minimization of the objective function to search for conditions resulting in a zero eigenvalue, (3) computation of a confidence level that describes if the global minimum has been found and (4) evaluation of optimization values, using either numerical continuation or directly simulating the ODE system, to verify that a bistability region exists. The generalized method has been tested on three motifs known to be capable of bistability. We have developed a variation of an optimization-based method for the discovery of bistability, which is not limited to uniterminal chemical reaction networks. Successful completion of the method provides an S-shaped bifurcation diagram, which indicates that the network acts as a bistable switch for the given optimization parameters.

37 INORGANIC, ORGANIC, PHYSICAL, AND ANALYTICAL CH↗
DOE OSTI2022

From Event Data to Wind Power Plant DQ Admittance and Stability Risk Assessment

This paper presents a dynamic event data-based stability risk assessment method for power grids with high penetrations of inverter-based resources (IBRs). This method relies on obtaining the IBRs' DQ admittance through dynamic event data and computing the system's eigenvalues based on the admittance models. Two critical technologies are employed in this research, including time-domain and frequency-domain data fitting and dq-frame voltage and current signal derivation. The first technology is key to obtaining the s-domain expressions from the transient response data, and the s-domain DQ admittance model from the frequency-domain measurements. The second technology is key to obtaining the dq-frame voltage and current signals from either the three-phase instantaneous measurements or the phasor measurement unit (PMU) data. The method is illustrated using data generated from a Type-4 wind power plant modeled in PSCAD. This paper demonstrates the technical feasibility of the proposed approach.

17 WIND ENERGY↗
NASA NTRS2010

Fast Solution in Sparse LDA for Binary Classification

An algorithm that performs sparse linear discriminant analysis (Sparse-LDA) finds near-optimal solutions in far less time than the prior art when specialized to binary classification (of 2 classes). Sparse-LDA is a type of feature- or variable- selection problem with numerous applications in statistics, machine learning, computer vision, computational finance, operations research, and bio-informatics. Because of its combinatorial nature, feature- or variable-selection problems are NP-hard or computationally intractable in cases involving more than 30 variables or features. Therefore, one typically seeks approximate solutions by means of greedy search algorithms. The prior Sparse-LDA algorithm was a greedy algorithm that considered the best variable or feature to add/ delete to/ from its subsets in order to maximally discriminate between multiple classes of data. The present algorithm is designed for the special but prevalent case of 2-class or binary classification (e.g. 1 vs. 0, functioning vs. malfunctioning, or change versus no change). The present algorithm provides near-optimal solutions on large real-world datasets having hundreds or even thousands of variables or features (e.g. selecting the fewest wavelength bands in a hyperspectral sensor to do terrain classification) and does so in typical computation times of minutes as compared to days or weeks as taken by the prior art. Sparse LDA requires solving generalized eigenvalue problems for a large number of variable subsets (represented by the submatrices of the input within-class and between-class covariance matrices). In the general (fullrank) case, the amount of computation scales at least cubically with the number of variables and thus the size of the problems that can be solved is limited accordingly. However, in binary classification, the principal eigenvalues can be found using a special analytic formula, without resorting to costly iterative techniques. The present algorithm exploits this analytic form along with the inherent sequential nature of greedy search itself. Together this enables the use of highly-efficient partitioned-matrix-inverse techniques that result in large speedups of computation in both the forward-selection and backward-elimination stages of greedy algorithms in general.

Moghaddam, Baback↗