Engineering Papers⌕ Search

SEARCH · Engineering Papers

Results for “model weights”

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 919 records · Page 51
All sourcesNASA NTRSDOE OSTI
NASA NTRS1993

Steady induction effects in geomagnetism. Part 1B: Geomagnetic estimation of steady surficial core motions: A non-linear inverse problem

The problem of estimating a steady fluid velocity field near the top of Earth's core which induces the secular variation (SV) indicated by models of the observed geomagnetic field is examined in the source-free mantle/frozen-flux core (SFI/VFFC) approximation. This inverse problem is non-linear because solutions of the forward problem are deterministically chaotic. The SFM/FFC approximation is inexact, and neither the models nor the observations they represent are either complete or perfect. A method is developed for solving the non-linear inverse motional induction problem posed by the hypothesis of (piecewise, statistically) steady core surface flow and the supposition of a complete initial geomagnetic condition. The method features iterative solution of the weighted, linearized least-squares problem and admits optional biases favoring surficially geostrophic flow and/or spatially simple flow. Two types of weights are advanced radial field weights for fitting the evolution of the broad-scale portion of the radial field component near Earth's surface implied by the models, and generalized weights for fitting the evolution of the broad-scale portion of the scalar potential specified by the models.

Voorhies, Coerte V.↗
DOE OSTI2021

Model-based real-time surface heat flux and temperature estimation for the DIII-D tokamak

A control-oriented model for monitoring of wall power flux densities on the DIII-D tokamak has been successfully implemented and validated experimentally. Future reactors will have to withstand severe steady state high heat flux loads on plasma-facing components (PFCs). Due to the difficulty of directly-measuring local heat fluxes on these components, monitoring and protection of PFCs during the plasma discharge can benefit from simplified physics-based real-time (RT) functional models to estimate and guide heat load control. As a first step into the development, a control-oriented model for monitoring of wall power flux densities and temperatures on DIII-D tokamak has been successfully implemented. The paper discusses the experimental demonstration and comparison of the 2-D model-based wall heat flux algorithm on the DIII-Dinner wall limiter (IWL) against infra-red camera heat flux measurements for limited plasma configurations. The paper also reports on the benchmarking of the field line tracing environment, SMITER, developed at ITER organization on DIII-D tokamak against experimental IR diagnostic data and the derivation of the component shaping weighting factors for the 2-D model-based approach. Here, the extension of the model-based approach for surface temperature estimation on the DIII-D IWL is also presented.

70 PLASMA PHYSICS AND FUSION TECHNOLOGY↗
DOE OSTI2017

A Method to Alleviate the Long History Problem Encountered in Monte Carlo Simulations via Weight Window Variance Reduction

When the weight window is implemented in large and complex models, streaming path and deep penetration could lead to a long history problem that takes a disproportionate amount of time for the accomplishment of Monte Carlo simulations and exerts a detrimental effect on the efficiency of parallel computing. An approach set the limitation of weight window splitting was proposed for the alleviation of long history problem. Tests were conducted on a streaming dog-leg geometry model and a 3D model of the Chinese Fusion Engineering Testing Reactor. The results show that a suitable parameter in the alleviation method could optimize both the performance of variance reduction and the efficiency of parallel calculation, which makes long history problem tractable.

70 PLASMA PHYSICS AND FUSION TECHNOLOGY↗
NASA NTRS2016

Full Gradient Solution to Adaptive Hybrid Control

This paper focuses on the adaptation mechanisms in adaptive hybrid controllers. Most adaptive hybrid controllers update two filters individually according to the filtered-reference least mean squares (FxLMS) algorithm. Because this algorithm was derived for feedforward control, it does not take into account the presence of a feedback loop in the gradient calculation. This paper provides a derivation of the proper weight vector gradient for hybrid (or feedback) controllers that takes into account the presence of feedback. In this formulation, a single weight vector is updated rather than two individually. An internal model structure is assumed for the feedback part of the controller. The full gradient is equivalent to that used in the standard FxLMS algorithm with the addition of a recursive term that is a function of the modeling error. Some simulations are provided to highlight the advantages of using the full gradient in the weight vector update rather than the approximation.

Bean, Jacob↗
NASA NTRS2017

Full Gradient Solution to Adaptive Hybrid Control

This paper focuses on the adaptation mechanisms in adaptive hybrid controllers. Most adaptive hybrid controllers update two filters individually according to the filtered reference least mean squares (FxLMS) algorithm. Because this algorithm was derived for feedforward control, it does not take into account the presence of a feedback loop in the gradient calculation. This paper provides a derivation of the proper weight vector gradient for hybrid (or feedback) controllers that takes into account the presence of feedback. In this formulation, a single weight vector is updated rather than two individually. An internal model structure is assumed for the feedback part of the controller. The full gradient is equivalent to that used in the standard FxLMS algorithm with the addition of a recursive term that is a function of the modeling error. Some simulations are provided to highlight the advantages of using the full gradient in the weight vector update rather than the approximation.

Bean, Jacob↗
DOE OSTI2015

A SIMPLE METHOD FOR MODELING COLLISION PROCESSES IN PLASMAS WITH A KAPPA ENERGY DISTRIBUTION

We demonstrate that a nonthermal distribution of particles described by a kappa distribution can be accurately approximated by a weighted sum of Maxwell–Boltzmann distributions. We apply this method to modeling collision processes in kappa-distribution plasmas, with a particular focus on atomic processes important for solar physics. The relevant collision process rate coefficients are generated by summing appropriately weighted Maxwellian rate coefficients. This method reproduces the rate coefficients for a kappa distribution to an estimated accuracy of better than 3%. This is equal to or better than the accuracy of rate coefficients generated using “reverse-engineering” methods, which attempt to extract the needed cross sections from the published Maxwellian rate coefficient data and then reconvolve the extracted cross sections with the desired kappa distribution. Our approach of summing Maxwellian rate coefficients is easy to implement using existing spectral analysis software. Moreover, the weights in the sum of the Maxwell–Boltzmann distribution rate coefficients can be found for any value of the parameter κ, thereby enabling one to model plasmas with a time-varying κ. Tabulated Maxwellian fitting parameters are given for specific values of κ from 1.7 to 100. We also provide polynomial fits to these parameters over this entire range. Several applications of our technique are presented, including the plasma equilibrium charge state distribution (CSD), predicting line ratios, modeling the influence of electron impact multiple ionization on the equilibrium CSD of kappa-distribution plasmas, and calculating the time-varying CSD of plasmas during a solar flare.

79 ASTRONOMY AND ASTROPHYSICS↗
NASA NTRS1987

Anticipation of the landing shock phenomenon in flight simulation

An aircraft landing may be described as a controlled crash because a runway surface is intercepted. In a simulation model the transition from aerodynamic flight to weight on wheels involves a single computational cycle during which stiff differential equations are activated; with a significant probability these initial conditions are unrealistic. This occurs because of the finite cycle time, during which large restorative forces will accompany unrealistic initial oleo compressions. This problem was recognized a few years ago at Ames Research Center during simulation studies of a supersonic transport. The mathematical model of this vehicle severely taxed computational resources, and required a large cycle time. The ground strike problem was solved by a described technique called anticipation equations. This extensively used technique has not been previously reported. The technique of anticipating a significant event is a useful tool in the general field of discrete flight simulation. For the differential equations representing a landing gear model stiffness, rate of interception and cycle time may combine to produce an unrealistic simulation of the continuum.

Mcfarland, Richard E.↗
NASA NTRS2020

AERoBOND Project Summary

Under NASA’s Convergent Aeronautics Solutions (CAS) project, the Adhesive-Free Bonding of Complex Composites (AERoBOND) project investigated off-stoichiometric epoxy polymers for fast, reliable assembly of epoxy matrix composite structures. The project goal was to demonstrate feasibility of the AERoBOND joining method by demonstrating mechanical properties greater than 80% of conventional co-cured materials while reducing structure weight by 1%. The project consisted of three convergent research areas: material and process development, systems analysis, and material and process modeling. Material and process development was the largest component of AERoBOND with approximately 6 FTE and 1WYE of support to formulate and characterize new resins, prepare carbon fiber prepregs, fabricate laminates, measure mechanical properties, analyze failure results, and select material and process improvements. The systems analysis activity estimated the potential reduction in part count and aircraft weight by comparing models of composite wing boxes with no fasteners (co-cured structure), fasteners in major joints (co-cured stringers), and fasteners in all joints. The materials and process modeling activity included a molecular model of the AERoBOND materials system to predict mechanical properties of resins with offset stoichiometry and a process model to predict the effect of resin formulation and processing conditions on the extent of mixing and degree of cure in a finished joint. As the number of airline passenger trips doubles in the next 20 years (IATA/Tourism Economics Air Passenger Forecasts, April 2019), the increased demand for new commercial aircraft is now the single greatest technical challenge to the airframe manufacturing industry. To meet efficiency requirements, new aircraft must be fabricated primarily from high performance structural composites, but manufacturing processes are inherently slow with the largest bottleneck attributed to assembly and installation of fasteners (NASA/TM–2019-220428). Manufactures of commercial transport aircraft are compelled to install more than 100,000 redundant fasteners into bonded joints to prevent failures due to unpredictable weak bonds. In structural adhesive bonds, the interface between adherend and adhesive is nearly two-dimensional making it susceptible to minute quantities of contamination, which can cause weak bonds. Currently, bond strength assessment is only possible through destructive testing (i.e., breaking the joint). For these reasons, regulatory organizations such as the Federal Aviation Administration (FAA) often require redundant load paths in secondary-bonded, primary-structures to alleviate concerns with bond performance. The AERoBOND process enables reflow of matrix resin during assembly to eliminate the material discontinuity at the interface, thereby eliminating the dependence of mechanical performance on interfacial adhesion. The AERoBOND joint is equivalent to the interlaminar region obtained during a co-cure process, so joint performance depends on the cohesive properties of the matrix resin. Conventional co-cured structures, although too costly and complex for large-scale manufacturing, are trusted by manufacturers and regulators, and are certified for flight with few or no redundant fasteners.Systems analysis performed on a composite wing model at the scale of a single-aisle commercial transport aircraft indicated that >20,000 redundant fasteners per wing could be eliminated by implementing the AERoBOND joining method. A total weight reduction of 15% was predicted in a wing box by eliminating fasteners and thinning components that must no longer support localized fastener loads and accommodate fastener dimensions. Interlaminar shear fracture toughness measured by the end-notched flexure test was greater than 1 kJ/m2 (nearly 140% of the co-cured benchmark property), which is greatly in excess of the project goals for mechanical properties. Testing was planned to measure interlaminar tensile fracture toughness as well as interlaminar tensile and shear strengths using the same AERoBOND configuration, but was delayed due to closure of LaRC facilities during the COVID-19 pandemic. The AERoBOND process model is partially validated and available for experimental use. It allows the user to input AERoBOND process parameters such as material composition, laminate configuration, and cure cycle to predict the final cure state of the AERoBOND joint. A preliminary, multi-scale material model was developed to predict AERoBOND joint mechanical properties (stiffness and strength) based on the cure state of the joint provided by the process model. The timing for transition of this technology within NASA is excellent as NASA initiates new enduring projects to address composites manufacturing rate challenges. AERoBOND technology is well suited to AAVP/AATT objectives for rapid manufacturing of a composite wing. A minimal effort (1 FTE/$15k procurement/0 WYE) is proposed in FY21 to continue a minor mechanical testing effort and maintain a SAA with ASX composites to develop commercial quality prepreg material. An RFI with the composites industry is suggested to quantify the technology gap between the current TRL and the TRL needed for transition to industry. A moderate effort [3-4 FTE/$150k/1 WYE (~$115k)] is proposed in FY22 for the “high rate composites manufacturing” project currently in planning. The partnership with ASX Composites will be expanded to produce material for sub-element/element-scale “panel-off” activities. Industry partnerships with airframe manufacturers is an expected component to explore damage tolerance and environmental stability. Further development of multi-scale modeling tools (process model, meso-scale model, and molecular model) is planned to enhance and deliver tools for rapid manufacturing infusion.

Frank Louis Palmieri↗
DOE OSTI2016

A Multi-Band Uncertainty Set Based Robust SCUC With Spatial and Temporal Budget Constraints

The dramatic increase of renewable energy resources in recent years, together with the long-existing load forecast errors and increasingly involved price sensitive demands, has introduced significant uncertainties into power systems operation. In order to guarantee the operational security of power systems with such uncertainties, robust optimization has been extensively studied in security-constrained unit commitment (SCUC) problems, for immunizing the system against worst uncertainty realizations. However, traditional robust SCUC models with single-band uncertainty sets may yield over-conservative solutions in most cases. This paper proposes a multi-band robust model to accurately formulate various uncertainties with higher resolution. By properly tuning band intervals and weight coefficients of individual bands, the proposed multi-band robust model can rigorously and realistically reflect spatial/temporal relationships and asymmetric characteristics of various uncertainties, and in turn could effectively leverage the tradeoff between robustness and economics of robust SCUC solutions. The proposed multi-band robust SCUC model is solved by Benders decomposition (BD) and outer approximation (OA), while taking the advantage of integral property of the proposed multi-band uncertainty set. In addition, several accelerating techniques are developed for enhancing the computational performance and the convergence speed. Numerical studies on a 6-bus system and the modified IEEE 118-bus system verify the effectiveness of the proposed robust SCUC approach for enhancing uncertainty modeling capabilities and mitigating conservativeness of the robust SCUC solution.

24 POWER TRANSMISSION AND DISTRIBUTION↗
DOE OSTI2024

3D Multiresolution Velocity Model Fusion with Probability Graphical Models

ABSTRACT The variability in spatial resolution of seismic velocity models obtained via tomographic methodologies is attributed to many factors, including inversion strategies, ray-path coverage, and data integrity. Integration of such models, with distinct resolutions, is crucial during the refinement of community models, thereby enhancing the precision of ground-motion simulations. Toward this goal, we introduce the probability graphical model (PGM), combining velocity models with heterogeneous resolutions and nonuniform data point distributions. The PGM integrates data relations across varying resolution subdomains, enhancing detail within low-resolution (LR) domains by utilizing information and prior knowledge from high-resolution (HR) subdomains through a maximum posterior problem. Assessment of efficacy, utilizing both 2D and 3D velocity models—consisting of synthetic checkerboard models and a fault-zone model from Ridgecrest, California—demonstrates noteworthy improvements in accuracy, compared to state-of-the-art fusion techniques. Specifically, we find reductions of 30% and 44% in computed travel-time residuals for 2D and 3D models, respectively, as compared to conventional smoothing techniques. Unlike conventional methods, the PGM’s adaptive weight selection facilitates preserving and learning details from complex, nonuniform HR models and applies the enhancements to the LR background domain.

Geochemistry & Geophysics↗
NASA NTRS1997

Parametric Flutter Analysis of the TCA Configuration and Recommendation for FFM Design and Scaling

The current HSR Aeroelasticity plan to design, build, and test a full span, free flying transonic flutter model in the TDT has many technical obstacles that must be overcome for a successful program. One technical obstacle is the determination of a suitable configuration and point in the sky to use in setting the scaling point for the ASE models program. Determining this configuration and point in the sky requires balancing several conflicting requirements, including model buildability, tunnel test safety, and the ability of the model to represent the flutter mechanisms of interest. As will be discussed in detail in subsequent sections, the current TCA design exhibits several flutter mechanisms of interest. It has been decided that the ASE models program will focus on the low frequency symmetric flutter mechanism, and will make no attempt to investigate high frequency flutter mechanisms. There are several reasons for this choice. First, it is believed that the high frequency flutter mechanisms are similar in nature to classical wing bending/torsion flutter, and therefore there is more confidence that this mechanism can be predicted using current techniques. The low frequency mode, on the other hand, is a highly coupled mechanism involving wing, body, tail, and engine motion which may be very difficult to predict. Second, the high frequency flutter modes result in very small weight penalties (several hundred pounds), while suppression of the low frequency mechanism inside the flight envelope causes thousands of pounds to be added to the structure. In order to successfully test the low frequency flutter mode of interest, a suitable starting configuration and point in the sky must be identified. The configuration and point in the sky must result in a wind tunnel model that (1) represents the low-frequency wing/body/engine/empennage flutter mechanisms that are unique to HSCT configurations, (2) flutters at an acceptably low frequency in the tunnel, (3) flutters at an acceptably low dynamic pressure in the tunnel, (4) allows sufficient weight for model buildability without inordinately high cost, and (5) has significant separation between the target flutter mechanism and other, potentially catastrophic, flutter mechanisms.

Baker, Myles↗
DOE OSTI2025

Vehicle-Cycle Inventory for Type C School Buses & Intra-City Transit Buses

This report documents the new inventory incorporated into the Research and Development version of Greenhouse gases, Regulated Emissions, and Energy use in Technologies (R&D GREET) 2025 model for the vehicle cycle of Type C school buses and intracity transit buses. The transportation sector contributes significantly to the United States’ energy consumption and resultant emissions (EPA, 2025a). However, public transit plays an important role in mitigating these impacts because it consumes a relatively low amount of energy per passenger (Congressional Budget Office, 2022). Public transit is widely used in the United States; more than 500,000 school buses (EPA, 2025b) and ~75,000 service buses (American Public Transportation Association, 2025) operate in the nation. These are primarily internal combustion engine vehicles (ICEVs) powered by diesel. Original equipment manufacturers (OEMs) are making efforts to electrify U.S. bus fleets by using batteries as a propulsion system to replace internal combustion engines. Electrification can reduce tailpipe emissions, such as particulate matter (with a diameter ≤10 µm [PM 10 ] and with a diameter ≤2.5 µm [PM 2.5 ]) and nitrogen oxides (Jonas et al., 2025; Martinez and Samaras, 2024; EPA, 2025b; Wayne et al., 2009). Hence, any energy and emission impact analysis of public transit must consider both conventional ICEVs and upcoming electric vehicle (EV) options for the school and transit buses that dominate this landscape. To understand the detailed environmental impact profiles of ICEV and EV school and transit buses, it is necessary to conduct a thorough analysis covering both vehicle manufacturing and vehicle use stages. The current literature lacks a detailed vehicle-cycle inventory for school and transit buses, which makes this kind of comparison difficult. To overcome this gap, we developed a comprehensive vehicle-cycle model for school and transit buses in Argonne’s R&D GREET 2025 model. The model is flexible in handling user inputs for key assumptions, such as component weights and material compositions, upstream energy sources for material processing, and vehicle operating parameters, to understand their impacts on energy use and emissions for both school and transit buses. This report is organized as follows: Section 2 provides details on the modeling approach and vehicle specifications (weights and composition of different vehicle components, and vehicle operating parameters) for both school and transit buses. Section 3 provides details on vehicle assembly, disposal, and recycling (ADR) approaches for the two buses. Section 4 includes details about their incorporation into the R&D GREET model.

32 ENERGY CONSERVATION, CONSUMPTION, AND UTILIZATI↗
DOE OSTI2015

Ensemble-type numerical uncertainty information from single model integrations

We suggest an algorithm that quantifies the discretization error of time-dependent physical quantities of interest (goals) for numerical models of geophysical fluid dynamics. The goal discretization error is estimated using a sum of weighted local discretization errors. The key feature of our algorithm is that these local discretization errors are interpreted as realizations of a random process. The random process is determined by the model and the flow state. From a class of local error random processes we select a suitable specific random process by integrating the model over a short time interval at different resolutions. The weights of the influences of the local discretization errors on the goal are modeled as goal sensitivities, which are calculated via automatic differentiation. The integration of the weighted realizations of local error random processes yields a posterior ensemble of goal approximations from a single run of the numerical model. From the posterior ensemble we derive the uncertainty information of the goal discretization error. This algorithm bypasses the requirement of detailed knowledge about the models discretization to generate numerical error estimates. The algorithm is evaluated for the spherical shallow-water equations. For two standard test cases we successfully estimate the error of regional potential energy, track its evolution, and compare it to standard ensemble techniques. The posterior ensemble shares linear-error-growth properties with ensembles of multiple model integrations when comparably perturbed. The posterior ensemble numerical error estimates are of comparable size as those of a stochastic physics ensemble.

71 CLASSICAL AND QUANTUM MECHANICS, GENERAL PHYSIC↗
DOE OSTI2024

Monitoring covariance in multivariate time series: Comparing machine learning and statistical approaches

Abstract In complex systems with multiple variables monitored at high‐frequency, variables are not only temporally autocorrelated, but they may also be nonlinearly related or exhibit nonstationarity as the inputs or operation changes. One approach to handling such variables is to detrend them prior to monitoring and then apply control charts that assume independence and stationarity to the residuals. Monitoring controlled systems is even more challenging because the control strategy seeks to maintain variables at prespecified mean levels, and to compensate, correlations among variables may change, making monitoring the covariance essential. In this paper, a vector autoregressive model (VAR) is compared with a multivariate random forest (MRF) and a neural network (NN) for detrending multivariate time series prior to monitoring the covariance of the residuals using a multivariate exponentially weighted moving average (MEWMA) control chart. Machine learning models have an advantage when the data's structure is unknown or may change. We design a novel simulation study with nonlinear, nonstationary, and autocorrelated data to compare the different detrending models and subsequent covariance monitoring. The machine learning models have superior performance for nonlinear and strongly autocorrelated data and similar performance for linear data. An illustration with data from a reverse osmosis process is given.

Weix, Derek↗
DOE OSTI2022

FLOW Estimation and Rose Superposition (FLOWERS): an integral approach to engineering wake models

Abstract. Annual energy production (AEP) is often the objective function in wind plant layout optimization studies. The conventional method to compute AEP for a wind farm is to first evaluate power production for each discrete wind direction and speed using either computational fluid dynamics simulations or engineering wake models. The AEP is then calculated by weighted-averaging (based on the wind rose at the wind farm site) the power produced across all wind directions and speeds. We propose a novel formulation for time-averaged wake velocity that incorporates an analytical integral of a wake deficit model across every wind direction. This approach computes the average flow field more efficiently, and layout optimization is an obvious application to exploit this benefit. The clear advantage of this new approach is that the layout optimization produces solutions with comparable AEP performance yet is completed 2 orders of magnitude faster. The analytical integral and the use of a Fourier expansion to express the wind speed and wind direction frequency create a relatively smooth solution space for the gradient-based optimizer to excel in comparison to the existing weighted-averaging power calculation.

17 WIND ENERGY↗
DOE OSTI2025

Resin Testing and Modeling for Optimal Composite Processing

Polymer composites have properties such as high strength and stiffness, low weight, good thermal and chemical stability, as well as impact and abrasion resistance that make them ideal for high-performance applications. The chemistries of these materials are continuously improving, so determining their properties is vital for successfully producing them and achieving the desired results. Multiple methods can be employed to monitor characteristics such as heat flow, weight, dimension, and modulus as a function of time and temperature. By analyzing this information, models can be developed to predict outcomes of parameters not tested for. In one application, materials proposed for wet filament winding and the production of high pressure vessels can be analyzed to verify they will have the necessary low viscosity for good fiber wetting and long pot life for the extended handling inherent to this process. Such data about a prospective system provides valuable information on how that material could ultimately be processed to yield the desired part.

36 MATERIALS SCIENCE↗
DOE OSTI2017

Exploring precipitation pattern scaling methodologies and robustness among CMIP5 models

Pattern scaling is a well-established method for approximating modeled spatial distributions of changes in temperature by assuming a time-invariant pattern that scales with changes in global mean temperature. We compare two methods of pattern scaling for annual mean precipitation (regression and epoch difference) and evaluate which method is better in particular circumstances by quantifying their robustness to interpolation/extrapolation in time, inter-model variations, and inter-scenario variations. Both the regression and epoch-difference methods (the two most commonly used methods of pattern scaling) have good absolute performance in reconstructing the climate model output, measured as an area-weighted root mean square error. We decompose the precipitation response in the RCP8.5 scenario into a CO 2 portion and a non-CO 2 portion. Extrapolating RCP8.5 patterns to reconstruct precipitation change in the RCP2.6 scenario results in large errors due to violations of pattern scaling assumptions when this CO 2 -/non-CO 2 -forcing decomposition is applied. As a result, the methodologies discussed in this paper can help provide precipitation fields to be utilized in other models (including integrated assessment models or impacts assessment models) for a wide variety of scenarios of future climate change.

54 ENVIRONMENTAL SCIENCES↗
DOE OSTI2025

SineKAN: Kolmogorov-Arnold Networks using sinusoidal activation functions

Recent work has established an alternative to traditional multi-layer perceptron neural networks in the form of Kolmogorov-Arnold Networks (KAN). The general KAN framework uses learnable activation functions on the edges of the computational graph followed by summation on nodes. The learnable edge activation functions in the original implementation are basis spline functions (B-Spline). Here, we present a model in which learnable grids of B-Spline activation functions are replaced by grids of re-weighted sine functions (SineKAN). We evaluate numerical performance of our model on a benchmark vision task. We show that our model can perform better than or comparable to B-Spline KAN models and an alternative KAN implementation based on periodic cosine and sine functions representing a Fourier Series. Further, we show that SineKAN has numerical accuracy that could scale comparably to dense neural networks (DNNs). Compared to the two baseline KAN models, SineKAN achieves a substantial speed increase at all hidden layer sizes, batch sizes, and depths. Current advantage of DNNs due to hardware and software optimizations are discussed along with theoretical scaling. Additionally, properties of SineKAN compared to other KAN implementations and current limitations are also discussed.

Reinhardt, Eric↗