Engineering Papers⌕ Search

SEARCH · Engineering Papers

Results for “NEURAL NETWORK”

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 955 records · Page 53

SwitchX : Gmin-Gmax Switching for Energy-efficient and Robust Implementation of Binarized Neural Networks on ReRAM Xbars

Memristive crossbars can efficiently implement Binarized Neural Networks (BNNs) wherein the weights are stored in high-resistance states (HRS) and low-resistance states (LRS) of the synapses. We propose SwitchX mapping of BNN weights onto ReRAM crossbars such that the impact of crossbar non-idealities, that lead to degradation in computational accuracy, are minimized. Essentially, SwitchX maps the binary weights in such a manner that a crossbar instance comprises of more HRS than LRS synapses. We find BNNs mapped onto crossbars with SwitchX to exhibit better robustness against adversarial attacks than the standard crossbar mapped BNNs, the baseline. Finally, we combine SwitchX with state-aware training (that further increases the feasibility of HRS states during weight mapping) to boost the robustness of a BNN on hardware. We find that this approach yields stronger defense against adversarial attacks than adversarial training, a state-of the-art software defense. We perform experiments on a VGG16 BNN with benchmark datasets (CIFAR-10, CIFAR-100 and TinyImagenet) and use Fast Gradient Sign Method (ϵ = 0.05 to 0.3) and Projected Gradient Descent (ϵ = $\frac{2}{255}$ to $\frac{32}{255}$, α = $\frac{2}{255}$) adversarial attacks. We show that SwitchX combined with state-aware training can yield upto ~35% improvements in clean accuracy and ~6–16% in adversarial accuracies against conventional BNNs. Furthermore, an important by-product of SwitchX mapping is increased crossbar power savings, owing to an increased proportion of HRS synapses, which is furthered with state-aware training. We obtain upto ~21–22% savings in crossbar power consumption for state-aware trained BNN mapped via SwitchX on 16 × 16 and 32 × 32 crossbars using the CIFAR-10 and CIFAR-100 datasets.

97 MATHEMATICS AND COMPUTING↗

Full-dimensional and reduced-dimensional calculations of initial state-selected reaction probabilities studying the H + CH{sub 4} → H{sub 2} + CH{sub 3} reaction on a neural network PES

Initial state-selected reaction probabilities of the H + CH{sub 4} → H{sub 2} + CH{sub 3} reaction are calculated in full and reduced dimensionality on a recent neural network potential [X. Xu, J. Chen, and D. H. Zhang, Chin. J. Chem. Phys. 27, 373 (2014)]. The quantum dynamics calculation employs the quantum transition state concept and the multi-layer multi-configurational time-dependent Hartree approach and rigorously studies the reaction for vanishing total angular momentum (J = 0). The calculations investigate the accuracy of the neutral network potential and study the effect resulting from a reduced-dimensional treatment. Very good agreement is found between the present results obtained on the neural network potential and previous results obtained on a Shepard interpolated potential energy surface. The reduced-dimensional calculations only consider motion in eight degrees of freedom and retain the C{sub 3v} symmetry of the methyl fragment. Considering reaction starting from the vibrational ground state of methane, the reaction probabilities calculated in reduced dimensionality are moderately shifted in energy compared to the full-dimensional ones but otherwise agree rather well. Similar agreement is also found if reaction probabilities averaged over similar types of vibrational excitation of the methane reactant are considered. In contrast, significant differences between reduced and full-dimensional results are found for reaction probabilities starting specifically from symmetric stretching, asymmetric (f{sub 2}-symmetric) stretching, or e-symmetric bending excited states of methane.

71 CLASSICAL AND QUANTUM MECHANICS, GENERAL PHYSIC↗

Accelerated Charged Particle Tracking with Graph Neural Networks on FPGAs

We develop and study FPGA implementations of algorithms for charged particle tracking based on graph neural networks. The two complementary FPGA designs are based on OpenCL, a framework for writing programs that execute across heterogeneous platforms, and hls4ml, a high-level-synthesis-based compiler for neural network to firmware conversion. We evaluate and compare the resource usage, latency, and tracking performance of our implementations based on a benchmark dataset. We find a considerable speedup over CPU-based execution is possible, potentially enabling such algorithms to be used effectively in future computing workflows and the FPGA-based Level-1 trigger at the CERN Large Hadron Collider.

46 INSTRUMENTATION RELATED TO NUCLEAR SCIENCE AND ↗

AutoTandemML: Active Learning Enhanced Tandem Neural Networks for Inverse Design Problems

Inverse design in science and engineering involves determining optimal design parameters that achieve desired performance outcomes, a process often hindered by the complexity and high dimensionality of design spaces, leading to significant computational costs. To tackle this challenge, we propose a novel hybrid approach that combines active learning with Tandem Neural Networks to enhance the efficiency and effectiveness of solving inverse design problems. Active learning allows to selectively sample the most informative data points, reducing the required dataset size without compromising accuracy. We investigate this approach using three benchmark problems: airfoil inverse design, photonic surface inverse design, and scalar boundary condition reconstruction in diffusion partial differential equations. We demonstrate that integrating active learning with Tandem Neural Networks outperforms standard approaches across the benchmark suite, achieving better accuracy with fewer training samples.

97 MATHEMATICS AND COMPUTING↗

Fast inference of deep neural networks in FPGAs for particle physics

Here, recent results at the Large Hadron Collider (LHC) have pointed to enhanced physics capabilities through the improvement of the real-time event processing techniques. Machine learning methods are ubiquitous and have proven to be very powerful in LHC physics, and particle physics as a whole. However, exploration of the use of such techniques in low-latency, low-power FPGA (Field Programmable Gate Array) hardware has only just begun. FPGA-based trigger and data acquisition systems have extremely low, sub-microsecond latency requirements that are unique to particle physics. We present a case study for neural network inference in FPGAs focusing on a classifier for jet substructure which would enable, among many other physics scenarios, searches for new dark sector particles and novel measurements of the Higgs boson. While we focus on a specific example, the lessons are far-reaching. A companion compiler package for this work is developed based on High-Level Synthesis (HLS) called hls4ml to build machine learning models in FPGAs. The use of HLS increases accessibility across a broad user community and allows for a drastic decrease in firmware development time. We map out FPGA resource usage and latency versus neural network hyperparameters to identify the problems in particle physics that would benefit from performing neural network inference with FPGAs. For our example jet substructure model, we fit well within the available resources of modern FPGAs with a latency on the scale of 100 ns.

46 INSTRUMENTATION RELATED TO NUCLEAR SCIENCE AND ↗

Effective cosmic density field reconstruction with convolutional neural network

ABSTRACT We present a cosmic density field reconstruction method that augments the traditional reconstruction algorithms with a convolutional neural network (CNN). Following previous work, the key component of our method is to use the reconstructed density field as the input to the neural network. We extend this previous work by exploring how the performance of these reconstruction ideas depends on the input reconstruction algorithm, the reconstruction parameters, and the shot noise of the density field, as well as the robustness of the method. We build an eight-layer CNN and train the network with reconstructed density fields computed from the Quijote suite of simulations. The reconstructed density fields are generated by both the standard algorithm and a new iterative algorithm. In real space at z = 0, we find that the reconstructed field is 90 per cent correlated with the true initial density out to $k\sim 0.5 \, \mathrm{ h}\, \rm {Mpc}^{-1}$, a significant improvement over $k\sim 0.2 \, \mathrm{ h}\, \rm {Mpc}^{-1}$ achieved by the input reconstruction algorithms. We find similar improvements in redshift space, including an improved removal of redshift space distortions at small scales. We also find that the method is robust across changes in cosmology. Additionally, the CNN removes much of the variance from the choice of different reconstruction algorithms and reconstruction parameters. However, the effectiveness decreases with increasing shot noise, suggesting that such an approach is best suited to high density samples. This work highlights the additional information in the density field beyond linear scales as well as the power of complementing traditional analysis approaches with machine learning techniques.

Astronomy & Astrophysics↗

Electronic neural network for dynamic resource allocation

A VLSI implementable neural network architecture for dynamic assignment is presented. The resource allocation problems involve assigning members of one set (e.g. resources) to those of another (e.g. consumers) such that the global 'cost' of the associations is minimized. The network consists of a matrix of sigmoidal processing elements (neurons), where the rows of the matrix represent resources and columns represent consumers. Unlike previous neural implementations, however, association costs are applied directly to the neurons, reducing connectivity of the network to VLSI-compatible 0 (number of neurons). Each row (and column) has an additional neuron associated with it to independently oversee activations of all the neurons in each row (and each column), providing a programmable 'k-winner-take-all' function. This function simultaneously enforces blocking (excitatory/inhibitory) constraints during convergence to control the number of active elements in each row and column within desired boundary conditions. Simulations show that the network, when implemented in fully parallel VLSI hardware, offers optimal (or near-optimal) solutions within only a fraction of a millisecond, for problems up to 128 resources and 128 consumers, orders of magnitude faster than conventional computing or heuristic search methods.

Thakoor, A. P.↗

What’s the Difference? The Potential for Convolutional Neural Networks for Transient Detection without Template Subtraction

Abstract We present a study of the potential for convolutional neural networks (CNNs) to enable separation of astrophysical transients from image artifacts, a task known as “real–bogus” classification, without requiring a template-subtracted (or difference) image, which requires a computationally expensive process to generate, involving image matching on small spatial scales in large volumes of data. Using data from the Dark Energy Survey, we explore the use of CNNs to (1) automate the real–bogus classification and (2) reduce the computational costs of transient discovery. We compare the efficiency of two CNNs with similar architectures, one that uses “image triplets” (templates, search, and difference image) and one that takes as input the template and search only. We measure the decrease in efficiency associated with the loss of information in input, finding that the testing accuracy is reduced from ∼96% to ∼91.1%. We further investigate how the latter model learns the required information from the template and search by exploring the saliency maps. Our work (1) confirms that CNNs are excellent models for real–bogus classification that rely exclusively on the imaging data and require no feature engineering task and (2) demonstrates that high-accuracy (>90%) models can be built without the need to construct difference images, but some accuracy is lost. Because, once trained, neural networks can generate predictions at minimal computational costs, we argue that future implementations of this methodology could dramatically reduce the computational costs in the detection of transients in synoptic surveys like Rubin Observatory's Legacy Survey of Space and Time by bypassing the difference image analysis entirely.

79 ASTRONOMY AND ASTROPHYSICS↗

Predicting Geologic Behavior in Carbon Storage Projects Using Graph Neural Network

This study was invited to presented at NVIDIA's GTC conference to highlight the potential of Graph Neural Network as a novel and promising methodology for predicting pressure and saturation evolution in carbon storage projects. Carbon capture and storage (CCS) technology plays a pivotal role in mitigating greenhouse gas emissions, facilitating the transition to a low-carbon future. Effective management of subsurface reservoirs is essential to ensure the safe and efficient storage of captured carbon dioxide (CO₂). Accurate predictions of pressure and saturation over time are critical for evaluating the long-term performance and integrity of CCS projects. In recent years, Graph Neural Network (GNN) has emerged as a powerful framework for analyzing complex data in graph-structured domains. This abstract explores the application of GNN to forecast pressure and saturation evolution in carbon storage projects. Traditional numerical simulations of subsurface reservoirs have proven successful in providing pressure and saturation forecasts. However, these simulations involve massive amounts of computational effort and require extensive domain expertise for proper model calibration and validation. Graph Neural Operator offers an alternative approach that harnesses the inherent graph structure of reservoirs, where nodes represent reservoir grid cells and edges represent the geological connectivity between them.

Shih, Chung Yan↗

Improved Tomographic Binning of 3 × 2 pt Lens Samples: Neural Network Classifiers and Optimal Bin Assignments

Abstract Large imaging surveys, such as the Legacy Survey of Space and Time, rely on photometric redshifts and tomographic binning for 3 × 2 pt analyses that combine galaxy clustering and weak lensing. In this paper, we propose a method for optimizing the tomographic binning choice for the lens sample of galaxies. We divide the CosmoDC2 and Buzzard simulated galaxy catalogs into a training set and an application set, where the training set is nonrepresentative in a realistic way, and then estimate photometric redshifts for the application sets. The galaxies are sorted into redshift bins covering equal intervals of redshift or comoving distance, or with an equal number of galaxies in each bin, and we consider a generalized extension of these approaches. We find that bins of equal comoving distance produce the highest dark energy figure of merit of the initial binning choices, but that the choice of bin edges can be further optimized. We then train a neural network classifier to identify galaxies that are either highly likely to have accurate photometric redshift estimates or highly likely to be sorted into the correct redshift bin. The neural network classifier is used to remove poor redshift estimates from the sample, and the results are compared to the case when none of the sample is removed. We find that the neural network classifiers are able to improve the figure of merit by ∼13% and are able to recover ∼25% of the loss in the figure of merit that occurs when a nonrepresentative training sample is used.

79 ASTRONOMY AND ASTROPHYSICS↗

Combining protein sequences and structures with transformers and equivariant graph neural networks to predict protein function

Abstract Motivation Millions of protein sequences have been generated by numerous genome and transcriptome sequencing projects. However, experimentally determining the function of the proteins is still a time consuming, low-throughput, and expensive process, leading to a large protein sequence-function gap. Therefore, it is important to develop computational methods to accurately predict protein function to fill the gap. Even though many methods have been developed to use protein sequences as input to predict function, much fewer methods leverage protein structures in protein function prediction because there was lack of accurate protein structures for most proteins until recently. Results We developed TransFun—a method using a transformer-based protein language model and 3D-equivariant graph neural networks to distill information from both protein sequences and structures to predict protein function. It extracts feature embeddings from protein sequences using a pre-trained protein language model (ESM) via transfer learning and combines them with 3D structures of proteins predicted by AlphaFold2 through equivariant graph neural networks. Benchmarked on the CAFA3 test dataset and a new test dataset, TransFun outperforms several state-of-the-art methods, indicating that the language model and 3D-equivariant graph neural networks are effective methods to leverage protein sequences and structures to improve protein function prediction. Combining TransFun predictions and sequence similarity-based predictions can further increase prediction accuracy. Availability and implementation The source code of TransFun is available at https://github.com/jianlin-cheng/TransFun.

59 BASIC BIOLOGICAL SCIENCES↗

Top-of-the-Atmosphere Shortwave Flux Estimation from Satellite Observations: An Empirical Neural Network Approach Applied with Data from the A-Train Constellation

Estimates of top-of-the-atmosphere (TOA) radiative flux are essential for the understanding of Earth's energy budget and climate system. Clouds, aerosols, water vapor, and ozone (O3) are among the most important atmospheric agents impacting the Earth's shortwave (SW) radiation budget. There are several sensors in orbit that provide independent information related to these parameters. Having coincident information from these sensors is important for understanding their potential contributions. The A-train constellation of satellites provides a unique opportunity to analyze data from several of these sensors. In this paper, retrievals of cloud/aerosol parameters and total column ozone (TCO) from the Aura Ozone Monitoring Instrument (OMI) have been collocated with the Aqua Clouds and Earth's Radiant Energy System (CERES) estimates of total reflected TOA outgoing SW flux (SWF). We use these data to develop a variety of neural networks that estimate TOA SWF globally over ocean and land using only OMI data and other ancillary information as inputs and CERES TOA SWF as the output for training purposes. OMI-estimated TOA SWF from the trained neural networks reproduces independent CERES data with high fidelity. The global mean daily TOA SWF calculated from OMI is consistently within 1% of CERES throughout the year 2007. Application of our neural network method to other sensors that provide similar retrieved parameters, both past and future, can produce similar estimates TOA SWF. For example, the well-calibrated Total Ozone Mapping Spectrometer (TOMS) series could provide estimates of TOA SWF dating back to late 1978.

CERES data with high fidelity↗

Testing autonomous cyber-physical systems using fuzzing features from convolutional neural networks: work-in-progress

Autonomous cyber-physical systems rely on modern machine learning methods such as deep neural networks to control their interactions with the physical world. Testing of such intelligent cyber-physical systems is a challenge due to the huge state space associated with high-resolution visual sensory inputs. We demonstrate how fuzzing the input using patterns obtained from the convolutional filters of an unrelated convolutional neural network can be used to test computer vision algorithms implemented in intelligent cyber-physical systems. Our method discovers interesting counterexamples to a pedestrian detection algorithm implemented in the popular OpenCV library. Our approach also unearths counterexamples to the correct behavior of an autonomous car similar to NVIDIA's end-to-end self-driving deep neural net running on the Udacity open-source simulator.

Raj, Sunny↗

Video frame prediction of microbial growth with a recurrent neural network

The recent explosion of interest and advances in machine learning technologies has opened the door to new analytical capabilities in microbiology. Using experimental data such as images or videos, machine learning, in particular deep learning with neural networks, can be harnessed to provide insights and predictions for microbial populations. This paper presents such an application in which a Recurrent Neural Network (RNN) was used to perform prediction of microbial growth for a population of two Pseudomonas aeruginosa mutants. The RNN was trained on videos that were acquired previously using fluorescence microscopy and microfluidics. Of the 20 frames that make up each video, 10 were used as inputs to the network which outputs a prediction for the next 10 frames of the video. The accuracy of the network was evaluated by comparing the predicted frames to the original frames, as well as population curves and the number and size of individual colonies extracted from these frames. Overall, the growth predictions are found to be accurate in metrics such as image comparison, colony size, and total population. Yet, limitations exist due to the scarcity of available and comparable data in the literature, indicating a need for more studies. Both the successes and challenges of our approach are discussed.

59 BASIC BIOLOGICAL SCIENCES↗

Optical computing and neural networks; Proceedings of the Meeting, National Chiao Tung Univ., Hsinchu, Taiwan, Dec. 16, 17, 1992

The present conference discusses optical neural networks, photorefractive nonlinear optics, optical pattern recognition, digital and analog processors, and holography and its applications. Attention is given to bifurcating optical information processing, neural structures in digital halftoning, an exemplar-based optical neural net classifier for color pattern recognition, volume storage in photorefractive disks, and microlaser-based compact optical neuroprocessors. Also treated are the optical implementation of a feature-enhanced optical interpattern-associative neural network model and its optical implementation, an optical pattern binary dual-rail logic gate module, a theoretical analysis for holographic associative memories, joint transform correlators, image addition and subtraction via the Talbot effect, and optical wavelet-matched filters. (No individual items are abstracted in this volume)

Hsu, Ken-Yuh↗

Predicting Ecologically Important Vegetation Variables from Remotely Sensed Optical/Radar Data Using Neural Networks

A number of satellite sensor systems will collect large data sets of the Earth's surface during NASA's Earth Observing System (EOS) era. Efforts are being made to develop efficient algorithms that can incorporate a wide variety of spectral data and ancillary data in order to extract vegetation variables required for global and regional studies of ecosystem processes, biosphere-atmosphere interactions, and carbon dynamics. These variables are, for the most part, continuous (e.g. biomass, leaf area index, fraction of vegetation cover, vegetation height, vegetation age, spectral albedo, absorbed photosynthetic active radiation, photosynthetic efficiency, etc.) and estimates may be made using remotely sensed data (e.g. nadir and directional optical wavelengths, multifrequency radar backscatter) and any other readily available ancillary data (e.g., topography, sun angle, ground data, etc.). Using these types of data, neural networks can: 1) provide accurate initial models for extracting vegetation variables when an adequate amount of data is available; 2) provide a performance standard for evaluating existing physically-based models; 3) invert multivariate, physically based models; 4) in a variable selection process, identify those independent variables which best infer the vegetation variable(s) of interest; and 5) incorporate new data sources that would be difficult or impossible to use with conventional techniques. In addition, neural networks employ a more powerful and adaptive nonlinear equation form as compared to traditional linear, index transformations, and simple nonlinear analyses. These neural networks attributes are discussed in the context of the authors' investigations of extracting vegetation variables of ecological interest.

Kimes, Daniel S.↗

Study of Single Top Quark Production Using Bayesian Neural Networks With D0 Detector at the Tevatron

Top quark, the heaviest and most intriguing among the six known quarks, can be created via two independent production mechanisms in {\pp} collisions. The primary mode, strong {\ttbar} pair production from a $gtt$ vertex, was used by the {\d0} and CDF collaborations to establish the existence of the top quark in March 1995. The second mode is the electroweak production of a single top quark or antiquark, which has been observed recently in March 2009. Since single top quarks are produced at hadron colliders through a $Wtb$ vertex, thereby provide a direct probe of the nature of $Wtb$ coupling and of the Cabibbo-Kobayashi-Maskawa matrix element, $V_{tb}$. So this mechanism provides a sensitive probe for several, standard model and beyond standard model, parameters such as anomalous $Wtb$ couplings. In this thesis, we measure the cross section of the electroweak produced top quark in three different production modes, $s+t$, $s$ and $t$-channels using a technique based on the Bayesian neural networks. This technique is applied for analysis of the 5.4 $fb^{-1}$ of data collected by the {\d0} detector. From a comparison of the Bayesian neural networks discriminants between data and the signal-background model using Bayesian statistics, the cross sections of the top quark produced through the electroweak mechanism have been measured as: \[\sigma(p\bar{p}→tb+X,tqb+X) = 3.11^{+0.77}_{-0.71}\;\rm pb\] \[\sigma(p\bar{p}→tb+X) = 0.72^{+0.44}_{-0.43}\;\rm pb\] \[\sigma(p\bar{p}→tqb+X) = 2.92^{+0.87}_{-0.73}\;\rm pb\] % The $s+t$-channel has a gaussian significance of $4.7\sigma$, the $s$-channel $0.9\sigma$ and the $t$-channel~$4.7\sigma$. The results are consistent with the standard model predictions within one standard deviation. By combining these results with the results for two other analyses (using different MVA techniques) improved results \[\sigma(p\bar{p}→tb+X,tqb+X) = 3.43^{+0.73}_{-0.74}\;\rm pb\] \[\sigma(p\bar{p}→tb+X) = 0.68^{+0.38}_{-0.35}\;\rm pb\] \[\sigma(p\bar{p}→tqb+X) = 2.86^{+0.69}_{-0.63}\;\rm pb\] % were obtained with a significance of $5.4\sigma$, $1.8\sigma$ and $5.0\sigma$ respectively for $s+t$, $s$ and $t$-channels. Using this measured cross section and constraining $0 \leq |V_{tb}|^2 \leq 1$, the lower limit has been calculated to be $|V_{tb}| > 0.79$ with 95\% confidence level (C.L.). Another measurement of $tqb$ production cross section is done using the same dataset and discriminant but without any assumption on the $tb$ production rate. From this measurement, we obtain a cross section of $2.90^{+0.59}_{-0.59}$~pb for $t$-channel and corresponding significance of $5.5\sigma$. In addition to the above mentioned work, a search is made for the anomalous $Wtb$ couplings in single top quark production. Within the Standard Model, the $Wtb$ vertex is purely left-handed, and its amplitude is given by the $V_{tb}$, related to weak interaction between a top and a $b$-quark. In a more general way, additional anomalous couplings such as right-handed vectorial couplings and left and right-handed tensorial couplings can also be considered. An analysis based on the Bayesian neural networks method is used to separate the signal from expected backgrounds. We find no evidence for anomalous couplings and set 95\% C.L. limits on these couplings as $|V_{tb} \cdot f_{L_T}|^2<0.06$, $|V_{tb} \cdot f_{R_V}|^2<0.93$ and $|V_{tb} \cdot f_{R_T}|^2<0.13$. This result represents the most stringent direct constraints on anomalous $Wtb$ interactions. This work has been done in collaboration with \d0 experiment but the analyses and results presented in this thesis are my contribution.

72 PHYSICS OF ELEMENTARY PARTICLES AND FIELDS↗