Engineering Papers⌕ Search

SEARCH · Engineering Papers

Results for “approximation algorithm”

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 1,531 records · Page 85
All sourcesNASA NTRSDOE OSTI
NASA NTRS1993

Generating Smooth Motions For Robotic Manipulators

In improved method for generating trajectory of robotic manipulator, each straight-line segment of trajectory composed of constant-velocity main portion sandwiched between smooth acceleration at start and smooth deceleration at finish. Algorithm implementing method computes velocity in each accelerating portion as sinusoidal function of position along line. This motion chosen for two reasons: closely approximates motion of human hand along straight-line trajectory, and provides very smooth transitions between constant-velocity portion and accelerated and decelerational end portions.

Bejczy, Antal K.↗
NASA NTRS1993

Resolution enhancement of spaceborne scatterometer data

Spaceborne wind scatterometers are designed principally to measure radar backscatter from the ocean's surface for the determination of the near-surface wind direction and speed. Although measurements of the radar backscatter are made over land, application of these measurements has been limited primarily to the calibration of the instrument. In this paper we present a method for generating enhanced resolution radar images of the earth's surface using spaceborne scatterometry. The technique is based on a new image reconstruction technique which takes advantage of the spatial overlap in scatterometer measurements made at different times to provide enhanced imaging resolution. We describe the reconstruction algorithm and demonstrate the technique using both simulated and actual Seasat-A Scatterometer (SASS) measurements. The SASS-derived images, which have approximately 4 km resolution, dramatically illustrate the resolution enhancement capability of the new technique. The technique permits utilization of both historic and contemporary scatterometer data for medium-scale monitoring of vegetation and polar ice. We discuss the tradeoff between imaging noise and resolution inherent in the technique.

Long, David G.↗
NASA NTRS1993

On l(1): Optimal decentralized performance

In this paper, the Manousiouthakis parametrization of all decentralized stabilizing controllers is employed in mathematically formulating the l(sup 1) optimal decentralized controller synthesis problem. The resulting optimization problem is infinite dimensional and therefore not directly amenable to computations. It is shown that finite dimensional optimization problems that have value arbitrarily close to the infinite dimensional one can be constructed. Based on this result, an algorithm that solves the l(sup 1) decentralized performance problems is presented. A global optimization approach to the solution of the infinite dimensional approximating problems is also discussed.

Sourlas, Dennis↗
NASA NTRS1994

Telecommunications end-to-end systems monitoring on TOPEX/Poseidon: Tools and techniques

The TOPEX/Poseidon Project Satellite Performance Analysis Team's (SPAT) roles and responsibilities have grown to include functions that are typically performed by other teams on JPL Flight Projects. In particular, SPAT Telecommunication's role has expanded beyond the nominal function of monitoring, assessing, characterizing, and trending the spacecraft (S/C) RF/Telecom subsystem to one of End-to-End Information Systems (EEIS) monitoring. This has been accomplished by taking advantage of the spacecraft and ground data system structures and protocols. By processing both the received spacecraft telemetry minor frame ground generated CRC flags and NASCOM block poly error flags, bit error rates (BER) for each link segment can be determined. This provides the capability to characterize the separate link segments, determine science data recovery, and perform fault/anomaly detection and isolation. By monitoring and managing the links, TOPEX has successfully recovered approximately 99.9 percent of the science data with an integrity (BER) of better than 1 x 10(exp 8). This paper presents the algorithms used to process the above flags and the techniques used for EEIS monitoring.

Calanche, Bruno J.↗
NASA NTRS1999

The Status of the Tropical Rainfall Measuring Mission (TRMM) after 2 Years in Orbit

The Tropical Rainfall Measuring Mission (TRMM) satellite was launched on November 27, 1997, and data from all the instruments first became available approximately 30 days after launch. Since then, much progress has been made in the calibration of the sensors, the improvement of the rainfall algorithms, in related modeling applications and in new datasets tailored specifically for these applications. This paper reports the latest results regarding the calibration of the TRMM Microwave Imager, (TMI), Precipitation Radar (PR) and Visible and Infrared Sensor (VIRS). For the TMI, a new product is in place that corrects for a still unknown source of radiation leaking in to the TMI receiver. The PR calibration has been adjusted upward slightly (by 0.6 dBZ) to better match ground reference targets, while the VIRS calibration remains largely unchanged. In addition to the instrument calibration, great strides have been made with the rainfall algorithms as well, with the new rainfall products agreeing with each other to within less than 20% over monthly zonally averaged statistics. The TRMM Science Data and Information System (TSDIS) has responded equally well by making a number of new products, including real-time and fine resolution gridded rainfall fields available to the modeling community. The TRMM Ground Validation (GV) program is also responding with improved radar calibration techniques and rainfall algorithms to provide more accurate GV products which will be further enhanced with the new multiparameter 10 cm radar being developed for TRMM validation and precipitation studies. Progress in these various areas has, in turn, led to exciting new developments in the modeling area where Data Assimilation, and Weather Forecast models are showing dramatic improvements after the assimilation of observed rainfall fields.

Kummerow, C.↗
NASA NTRS1997

Neural Network Prediction of Failure of Damaged Composite Pressure Vessels from Strain Field Data Acquired by a Computer Vision Method

This effort used a new and novel method of acquiring strains called Sub-pixel Digital Video Image Correlation (SDVIC) on impact damaged Kevlar/epoxy filament wound pressure vessels during a proof test. To predict the burst pressure, the hoop strain field distribution around the impact location from three vessels was used to train a neural network. The network was then tested on additional pressure vessels. Several variations on the network were tried. The best results were obtained using a single hidden layer. SDVIC is a fill-field non-contact computer vision technique which provides in-plane deformation and strain data over a load differential. This method was used to determine hoop and axial displacements, hoop and axial linear strains, the in-plane shear strains and rotations in the regions surrounding impact sites in filament wound pressure vessels (FWPV) during proof loading by internal pressurization. The relationship between these deformation measurement values and the remaining life of the pressure vessels, however, requires a complex theoretical model or numerical simulation. Both of these techniques are time consuming and complicated. Previous results using neural network methods had been successful in predicting the burst pressure for graphite/epoxy pressure vessels based upon acoustic emission (AE) measurements in similar tests. The neural network associates the character of the AE amplitude distribution, which depends upon the extent of impact damage, with the burst pressure. Similarly, higher amounts of impact damage are theorized to cause a higher amount of strain concentration in the damage effected zone at a given pressure and result in lower burst pressures. This relationship suggests that a neural network might be able to find an empirical relationship between the SDVIC strain field data and the burst pressure, analogous to the AE method, with greater speed and simplicity than theoretical or finite element modeling. The process of testing SDVIC neural network analysis and some encouraging preliminary results are presented in this paper. Details are given concerning the processing of SDVIC output data such that it may be used as back propagation neural network (BPNN) input data. The software written to perform this processing and the BPNN algorithm are also discussed. It will be shown that, with limited training, test results indicate an average error in burst pressure prediction of approximately six percent,

Russell, Samuel S.↗
NASA NTRS2001

Tropical Rainfall Measuring Mission (TRMM) and the Future of Rainfall Estimation from Space

Tropical rainfall is important in the hydrological cycle and to the lives and welfare of humans. Three-fourths of the energy that drives the atmospheric wind circulation comes from the latent heat released by tropical precipitation. Recognizing the importance of rain in the tropics, NASA for the U.S.A. and NASDA for Japan have partnered in the design, construction and flight of a satellite mission to measure tropical rainfall and calculate the associated latent heat release. The Tropical Rainfall Measuring Mission (TRMM) satellite was launched on November 27, 1997, and data from all the instruments first became available approximately 30 days after launch. Since then, much progress has been made in the calibration of the sensors, the improvement of the rainfall algorithms and applications of these results to areas such as Data Assimilation and model initialization. TRMM has reduced the uncertainty of climatological rainfall in tropics by over a factor of two, therefore establishing a standard for comparison with previous data sets and climatologies. It has documented the diurnal variation of precipitation over the oceans, showing a distinct early morning peak and this satellite mission has shown the utility of precipitation information for the improvement of numerical weather forecasts and climate modeling. This paper discusses some promising applications using TRMM data and introduces a measurement concept being discussed by NASA/NASDA and ESA for the future of rainfall estimation from space.

Kakar, Ramesh↗
NASA NTRS2023

Efficient Calibration of Expensive Computational Models

Accounting for uncertainty when calibrating expensive computational models is a common challenge faced by scientists and engineers. Often Bayesian techniques are adopted to estimate a probability density function over the model parameters given noisy empirical data. The methods used to perform this type of probabilistic calibration are computationally prohibitive in that they require a large number of evaluations of the expensive model. In these cases, surrogate modeling -- that is, using a fast-to-evaluate, lower fidelity stand-in for the original computational model -- may be the only option to alleviate this computational burden. However, the upfront cost of generating training data to build a surrogate model can itself be expensive. As such, it is important to be judicious when selecting training points at which the full-fidelity model is evaluated. Here, an active learning approach is proposed that enables efficient selection of training points using approximate samples of the calibrated parameter probability density function. In this way, the training points can be concentrated in regions where the calibration algorithm requires high model accuracy.

active learning↗
NASA NTRS2023

Using Satellite Data to Characterize Land Surface Processes in Morocco

This study endeavors to produce a comprehensive land cover map for Morocco, addressing the absence of such a detailed map in the country. Our research encompasses ecological and climatic aspects specific to Morocco, while the methods used can be adapted to various regions and countries, considering their unique climatic conditions and land cover types. A combination of MODIS and Landsat datasets was employed to create a 5 km resolution Land Use and Land Cover (LULC) map for the entire nation. The process involved the aggregation and advanced processing of these datasets using surface processes algorithms. The resulting LULC map is the first of its kind for Morocco, shedding light on land cover distribution nationwide. It shows that approximately 13.5% of the country is covered by forests, predominantly in the Atlas and Rif mountains, Rabat–Sale, and the southern regions. Grasslands occupy over 16% of the study area, mainly in the north-east and west. Urban areas, including major cities like Casablanca, Rabat, and Marrakech, span nearly 3400 km². Moreover, large areas of shrublands and bare lands are evident across the country, while agricultural lands account for almost 20% of the national territory, mainly in the interior plains and north-western Atlantic coast. This study forms a crucial basis for ecological and climatic research in Morocco and serves as a valuable reference for various disciplines such as agriculture, natural resource management, and climate modeling. The mapping of biophysical parameters for each land cover class is a key feature of our research, and these parameters will be instrumental in a subsequent study examining the impact of urban development on surface climate in Morocco. Overall, our study underscores the importance of understanding biophysical parameters in addressing environmental and societal challenges.

Mohammed Thaiki↗
NASA NTRS2005

Evolutionary Computing for Low-thrust Navigation

The development of new mission concepts requires efficient methodologies to analyze, design and simulate the concepts before implementation. New mission concepts are increasingly considering the use of ion thrusters for fuel-efficient navigation in deep space. This paper presents parallel, evolutionary computing methods to design trajectories of spacecraft propelled by ion thrusters and to assess the trade-off between delivered payload mass and required flight time. The developed methods utilize a distributed computing environment in order to speed up computation, and use evolutionary algorithms to find globally Pareto-optimal solutions. The methods are coupled with two main traditional trajectory design approaches, which are called direct and indirect. In the direct approach, thrust control is discretized in either arc time or arc length, and the resulting discrete thrust vectors are optimized. In the indirect approach, a thrust control problem is transformed into a costate control problem, and the initial values of the costate vector are optimized. The developed methods are applied to two problems: 1) an orbit transfer around the Earth and 2) a transfer between two distance retrograde orbits around Europa, the closest to Jupiter of the icy Galilean moons. The optimal solutions found with the present methods are comparable to other state-of-the-art trajectory optimizers and to analytical approximations for optimal transfers, while the required computational time is several orders of magnitude shorter than other optimizers thanks to an intelligent design of control vector discretization, advanced algorithmic parameterization, and parallel computing.

optimization↗
DOE OSTI2024

Development of Segregated Thermal-Hydraulics Solvers in MOOSE

The simulation of fluid flows is an essential part of the design and analysis of nuclear systems. Algorithms able to simulate flows at different fidelity levels are available in the Multiphysics Object-Oriented Simulation Environment (MOOSE) and MOOSE-based applications such as Pronghorn \cite{novak2018pronghorn}, Pronghorn-Subchannel, RELAP-7, and SAM. Currently, significant effort is being invested in the development of coarse-mesh Computational Fluid Dynamics (CFD) capabilities within MOOSE and Pronghorn for the simulation of Generation IV nuclear reactors. Traditionally, the solution algorithms in MOOSE have relied on Newton or quasi-Newton methods (such as the preconditioned Jacobian-free Newton-Krylov method) where residuals and Jacobians (or approximations thereof) are constructed. Both Newton and quasi-Newton methods require the solution of a linear system at each nonlinear Newton iteration with the Jacobian as the system matrix. The Jacobian contains blocks originating from all variables in the problem (i.e., for thermal-hydraulics at least pressure, velocities, and temperature). Due to the formulation of the problem in a general multiphysics setting on unstructured mesh, creating a good preconditioner for the linear system can be challenging, thus many fluid applications have utilized direct solver-based methods such as LU factorization. However, with increasing system size and complexity in multi-dimensional problems, the direct solution of linear systems becomes computationally expensive both in execution time and and memory. For this reason, recent effort has focused on adapting segregated solution algorithms for CFD problems in MOOSE. These algorithms use fixed-point iteration between segregated systems whose assembly and preconditioning are easier those of the monolithic system. Initial results show that the segregated solution algorithm outperforms the monolithic approach in terms of memory usage and for large 3D problems in terms of CPU time as well.

42 ENGINEERING↗
DOE OSTI2024

Development of Segregated Thermal-Hydraulics Solvers in MOOSE

The simulation of fluid flows is an essential part of the design and analysis of nuclear systems. Algorithms able to simulate flows at different fidelity levels are available in the Multiphysics Object-Oriented Simulation Environment (MOOSE) and MOOSE-based applications such as Pronghorn \cite{novak2018pronghorn}, Pronghorn-Subchannel, RELAP-7, and SAM. Currently, significant effort is being invested in the development of coarse-mesh Computational Fluid Dynamics (CFD) capabilities within MOOSE and Pronghorn for the simulation of Generation IV nuclear reactors. Traditionally, the solution algorithms in MOOSE have relied on Newton or quasi-Newton methods (such as the preconditioned Jacobian-free Newton-Krylov method) where residuals and Jacobians (or approximations thereof) are constructed. Both Newton and quasi-Newton methods require the solution of a linear system at each nonlinear Newton iteration with the Jacobian as the system matrix. The Jacobian contains blocks originating from all variables in the problem (i.e., for thermal-hydraulics at least pressure, velocities, and temperature). Due to the formulation of the problem in a general multiphysics setting on unstructured mesh, creating a good preconditioner for the linear system can be challenging, thus many fluid applications have utilized direct solver-based methods such as LU factorization. However, with increasing system size and complexity in multi-dimensional problems, the direct solution of linear systems becomes computationally expensive both in execution time and and memory. For this reason, recent effort has focused on adapting segregated solution algorithms for CFD problems in MOOSE. These algorithms use fixed-point iteration between segregated systems whose assembly and preconditioning are easier those of the monolithic system. Initial results show that the segregated solution algorithm outperforms the monolithic approach in terms of memory usage and for large 3D problems in terms of CPU time as well.

42 ENGINEERING↗
DOE OSTI2021

Filament-induced breakdown spectroscopy signal enhancement using optical wavefront control

Filament-induced breakdown spectroscopy (FIBS) is an analytical method that holds significant promise for remote sensing. In FIBS, it is desirable to maximize the intensity and improve the reproducibility of a selected spectroscopic feature. Here, we demonstrate the use of wavefront control in conjunction with a genetic algorithm in FIBS of metallic copper and show that this approach can increase the efficacy of filament-induced breakdown signal production. Through wavefront optimization, we enhance the intensity of a chosen characteristic spectroscopic feature of copper by a factor of approximately 3 when performing filamentation over a 1-meter distance. The relative standard deviation of signal intensity is reduced from ~17% to ~12% after optimization. We find that modification of the wavefront by introduction of astigmatism and coma maximizes the signal intensity, and these aberrations appear in two distinct trials of the genetic algorithm. We compare these findings to previous work on multiple filament control and discuss the possible mechanisms that lead to signal enhancement associated with both the beam amplitude and phase profile. A broader use of wavefront control with feedback may improve the performance of FIBS in remote sensing applications.

42 ENGINEERING↗
DOE OSTI2021

A Frequency-Domain-Based Algorithm for Detecting Microseismicity Using Dense Surface Seismic Arrays

We propose a new frequency-domain-based algorithm for detecting small-magnitude seismic events using dense surface seismic arrays. Our proposed method takes advantage of the high energy carried by S waves, and approximate known source locations, which are used to rotate the horizontal components to obtain the maximum amplitude. By surrounding the known source area with surface geophones, we achieve a favorable geometry for locating the detected seismic events with the backprojection method. To test our new detection method, we used a dense circular array, consisting of 151 5 Hz three-component geophones, over a 5 km aperture that was in operation at the Utah Frontier Observatory for Research in Geothermal Energy (FORGE) in southcentral Utah. We apply the new detection method during a small-scale test injection phase at FORGE, and during an aftershock sequence of an Mw 4.1 earthquake located ~30 km north of the geophone array, within the Black Rock volcanic field. We are able to detect and locate microseismic events (Mw<0) during injections, despite the high level of anthropogenic activity, and several aftershocks that are missing from the regional catalog. By comparing our method with known algorithms that operate both in the time and frequency domain, we show that our proposed method performs better in the case of the FORGE injection monitoring, and equally well for the off-array aftershock sequence. Our new method has the potential to improve microseismic event detections even in extremely noisy environments, and the proposed location scheme serves as a direct discriminant between true and false detections.

Geochemistry & Geophysics↗
NASA NTRS2016

Soil Moisture Active Passive Mission L4_SM Data Product Assessment (Version 2 Validated Release)

During the post-launch SMAP calibration and validation (Cal/Val) phase there are two objectives for each science data product team: 1) calibrate, verify, and improve the performance of the science algorithm, and 2) validate the accuracy of the science data product as specified in the science requirements and according to the Cal/Val schedule. This report provides an assessment of the SMAP Level 4 Surface and Root Zone Soil Moisture Passive (L4_SM) product specifically for the product's public Version 2 validated release scheduled for 29 April 2016. The assessment of the Version 2 L4_SM data product includes comparisons of SMAP L4_SM soil moisture estimates with in situ soil moisture observations from core validation sites and sparse networks. The assessment further includes a global evaluation of the internal diagnostics from the ensemble-based data assimilation system that is used to generate the L4_SM product. This evaluation focuses on the statistics of the observation-minus-forecast (O-F) residuals and the analysis increments. Together, the core validation site comparisons and the statistics of the assimilation diagnostics are considered primary validation methodologies for the L4_SM product. Comparisons against in situ measurements from regional-scale sparse networks are considered a secondary validation methodology because such in situ measurements are subject to up-scaling errors from the point-scale to the grid cell scale of the data product. Based on the limited set of core validation sites, the wide geographic range of the sparse network sites, and the global assessment of the assimilation diagnostics, the assessment presented here meets the criteria established by the Committee on Earth Observing Satellites for Stage 2 validation and supports the validated release of the data. An analysis of the time average surface and root zone soil moisture shows that the global pattern of arid and humid regions are captured by the L4_SM estimates. Results from the core validation site comparisons indicate that "Version 2" of the L4_SM data product meets the self-imposed L4_SM accuracy requirement, which is formulated in terms of the ubRMSE: the RMSE (Root Mean Square Error) after removal of the long-term mean difference. The overall ubRMSE of the 3-hourly L4_SM surface soil moisture at the 9 km scale is 0.035 cubic meters per cubic meter requirement. The corresponding ubRMSE for L4_SM root zone soil moisture is 0.024 cubic meters per cubic meter requirement. Both of these metrics are comfortably below the 0.04 cubic meters per cubic meter requirement. The L4_SM estimates are an improvement over estimates from a model-only SMAP Nature Run version 4 (NRv4), which demonstrates the beneficial impact of the SMAP brightness temperature data. L4_SM surface soil moisture estimates are consistently more skillful than NRv4 estimates, although not by a statistically significant margin. The lack of statistical significance is not surprising given the limited data record available to date. Root zone soil moisture estimates from L4_SM and NRv4 have similar skill. Results from comparisons of the L4_SM product to in situ measurements from nearly 400 sparse network sites corroborate the core validation site results. The instantaneous soil moisture and soil temperature analysis increments are within a reasonable range and result in spatially smooth soil moisture analyses. The O-F residuals exhibit only small biases on the order of 1-3 degrees Kelvin between the (re-scaled) SMAP brightness temperature observations and the L4_SM model forecast, which indicates that the assimilation system is largely unbiased. The spatially averaged time series standard deviation of the O-F residuals is 5.9 degrees Kelvin, which reduces to 4.0 degrees Kelvin for the observation-minus-analysis (O-A) residuals, reflecting the impact of the SMAP observations on the L4_SM system. Averaged globally, the time series standard deviation of the normalized O-F residuals is close to unity, which would suggest that the magnitude of the modeled errors approximately reflects that of the actual errors. The assessment report also notes several limitations of the "Version 2" L4_SM data product and science algorithm calibration that will be addressed in future releases. Regionally, the time series standard deviation of the normalized O-F residuals deviates considerably from unity, which indicates that the L4_SM assimilation algorithm either over- or under-estimates the actual errors that are present in the system. Planned improvements include revised land model parameters, revised error parameters for the land model and the assimilated SMAP observations, and revised surface meteorological forcing data for the operational period and underlying climatological data. Moreover, a refined analysis of the impact of SMAP observations will be facilitated by the construction of additional variants of the model-only reference data. Nevertheless, the “Version 2” validated release of the L4_SM product is sufficiently mature and of adequate quality for distribution to and use by the larger science and application communities.

SMAP L4_SM↗
DOE OSTI2023

Quantum algorithms for geologic fracture networks

Abstract Solving large systems of equations is a challenge for modeling natural phenomena, such as simulating subsurface flow. To avoid systems that are intractable on current computers, it is often necessary to neglect information at small scales, an approach known as coarse-graining. For many practical applications, such as flow in porous, homogenous materials, coarse-graining offers a sufficiently-accurate approximation of the solution. Unfortunately, fractured systems cannot be accurately coarse-grained, as critical network topology exists at the smallest scales, including topology that can push the network across a percolation threshold. Therefore, new techniques are necessary to accurately model important fracture systems. Quantum algorithms for solving linear systems offer a theoretically-exponential improvement over their classical counterparts, and in this work we introduce two quantum algorithms for fractured flow. The first algorithm, designed for future quantum computers which operate without error, has enormous potential, but we demonstrate that current hardware is too noisy for adequate performance. The second algorithm, designed to be noise resilient, already performs well for problems of small to medium size (order 10–1000 nodes), which we demonstrate experimentally and explain theoretically. We expect further improvements by leveraging quantum error mitigation and preconditioning.

58 GEOSCIENCES↗
NASA NTRS1998

Multi-Component Diffusion with Application To Computational Aerothermodynamics

The accuracy and complexity of solving multicomponent gaseous diffusion using the detailed multicomponent equations, the Stefan-Maxwell equations, and two commonly used approximate equations have been examined in a two part study. Part I examined the equations in a basic study with specified inputs in which the results are applicable for many applications. Part II addressed the application of the equations in the Langley Aerothermodynamic Upwind Relaxation Algorithm (LAURA) computational code for high-speed entries in Earth's atmosphere. The results showed that the presented iterative scheme for solving the Stefan-Maxwell equations is an accurate and effective method as compared with solutions of the detailed equations. In general, good accuracy with the approximate equations cannot be guaranteed for a species or all species in a multi-component mixture. 'Corrected' forms of the approximate equations that ensured the diffusion mass fluxes sum to zero, as required, were more accurate than the uncorrected forms. Good accuracy, as compared with the Stefan- Maxwell results, were obtained with the 'corrected' approximate equations in defining the heating rates for the three Earth entries considered in Part II.

Sutton, Kenneth↗
DOE OSTI2022

GAAF: Searching Activation Functions for Binary Neural Networks Through Genetic Algorithm

Binary neural networks (BNNs) show promising utilization in cost and power-restricted domains such as edge devices and mobile systems. This is due to its significantly less computation and storage demand, but at the cost of degraded performance. To close the accuracy gap, in this paper we propose to add a complementary activation function (AF) ahead of the sign based binarization, and rely on the genetic algorithm (GA) to automatically search for the ideal AFs. These AFs can help extract extra information from the input data in the forward pass, while allowing improved gradient approximation in the backward pass. Fifteen novel AFs are identified through our GA-based search, while most of them show improved performance (up to 2.54% on ImageNet) when testing on different datasets and network models. Interestingly, periodic functions are identified as a key component for most of the discovered AFs, which rarely exist in human designed AFs. Our method offers a novel approach for designing general and application-specific BNN architecture.

59 BASIC BIOLOGICAL SCIENCES↗