Search NASA⌕ Search

SEARCH · Search NASA

Results for “neural resources”

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 37 records · Page 2
All sourcesNASA NTRSDOE OSTI
DOE OSTI2022

Scalable deep learning for watershed model calibration

Watershed models such as the Soil and Water Assessment Tool (SWAT) consist of high-dimensional physical and empirical parameters. These parameters often need to be estimated/calibrated through inverse modeling to produce reliable predictions on hydrological fluxes and states. Existing parameter estimation methods can be time consuming, inefficient, and computationally expensive for high-dimensional problems. In this paper, we present an accurate and robust method to calibrate the SWAT model (i.e., 20 parameters) using scalable deep learning (DL). We developed inverse models based on convolutional neural networks (CNN) to assimilate observed streamflow data and estimate the SWAT model parameters. Scalable hyperparameter tuning is performed using high-performance computing resources to identify the top 50 optimal neural network architectures. We used ensemble SWAT simulations to train, validate, and test the CNN models. We estimated the parameters of the SWAT model using observed streamflow data and assessed the impact of measurement errors on SWAT model calibration. We tested and validated the proposed scalable DL methodology on the American River Watershed, located in the Pacific Northwest-based Yakima River basin. Our results show that the CNN-based calibration is better than two popular parameter estimation methods (i.e., the generalized likelihood uncertainty estimation [GLUE] and the dynamically dimensioned search [DDS], which is a global optimization algorithm). For the set of parameters that are sensitive to the observations, our proposed method yields narrower ranges than the GLUE method but broader ranges than values produced using the DDS method within the sampling range even under high relative observational errors. The SWAT model calibration performance using the CNNs, GLUE, and DDS methods are compared using R 2 and a set of efficiency metrics, including Nash-Sutcliffe, logarithmic Nash-Sutcliffe, Kling-Gupta, modified Kling-Gupta, and non-parametric Kling-Gupta scores, computed on the observed and simulated watershed responses. The best CNN-based calibrated set has scores of 0.71, 0.75, 0.85, 0.85, 0.86, and 0.91. The best DDS-based calibrated set has scores of 0.62, 0.69, 0.8, 0.77, 0.79, and 0.82. The best GLUE-based calibrated set has scores of 0.56, 0.58, 0.71, 0.7, 0.71, and 0.8. The scores above show that the CNN-based calibration leads to more accurate low and high streamflow predictions than the GLUE and DDS sets. Our research demonstrates that the proposed method has high potential to improve our current practice in calibrating large-scale integrated hydrologic models.

54 ENVIRONMENTAL SCIENCES↗
DOE OSTI2022

Neural net detection of seismic features related to gas hydrates and free gas accumulations on the northern U.S. Atlantic margin

Bottom-simulating reflections (BSRs) that sometimes mark the base of the gas hydrate stability zone in marine sediments are often identified based on the reverse polarity reflections that cut across stratigraphic layering in seismic amplitude data. On the northern U.S. Atlantic margin (USAM) between Cape Hatteras and Hudson Canyon, legacy seismic data have revealed pronounced BSRs south of the deepwater extension of Hudson Canyon and more subtle ones from offshore Delaware south to Cape Hatteras, where the reflections sometimes follow stratigraphic layering. Using high-resolution seismic data acquired during the 2018 Mid-Atlantic Resource Imaging Experiment and a supervised neural net, we identify seismic features associated with gas hydrates and/or the top of gas between Hudson Canyon and Cape Hatteras. Using seismic attributes especially sensitive to the presence of gas, we train a neural network algorithm on seismic data from an area with strong BSRs and then apply the model to the rest of the data set. The results indicate that gas hydrate and/or shallow free gas are significantly more widespread on the northern part of the USAM than previously known. Seismic indicators of gas extend landward from the 2000 m isobath to the upper continental slope in sectors with (offshore Virginia) and, to a lesser extent, without (offshore New Jersey) pervasive upper slope methane seeps. Higher sand content and intermediate sediment thickness, factors related to the container size and gas charge in a petroleum systems framework, are associated with more robust gas indicators.

Geochemistry & Geophysics↗
DOE OSTI2025

wa-hls4ml and lui-gnn: A benchmark and GNN-based surrogate model for hls4ml resource and latency estimation

As machine learning (ML) increasingly serves as a tool for addressing real-time challenges in scientific applications, the development of advanced tooling has significantly reduced the time required to iterate on various designs. These advancements have solved major obstacles, but also exposed new challenges. For example, processes that were not previously considered bottlenecks, such as model synthesis, are now becoming limiting factors in the rapid iteration of designs. To reduce these emerging constraints, multiple efforts are being launched toward designing an ML-based surrogate model that estimates resource usage of synthesized accelerator architectures. This model would reduce the design iteration time, especially when designing within a set of given hardware constraints. This approach shows considerable potential, but as it stands, the effort is early and would benefit from coordination and standardization to assist future work as it emerges. We introduce wa-hls4ml, a benchmark for ML accelerator resource and latency estimation, and its corresponding initial dataset of more than 100,000 fully connected neural networks, all synthesized using hls4ml and targeting Xilinx FPGAs. In addition to the resource utilization and latency data provided, the dataset includes generated artifacts and log files for many of the synthesized neural networks, in order to support future research in ML-based code generation. The benchmark evaluates the performance of resource and latency predictors against several common ML model architectures, primarily originating from scientific domains, as exemplar models, as well as the average performance across a subset of the dataset. We measure the performance of a given predictor model through multiple metrics, including $R^2$ score and SMAPE on regression tasks, as well as inference time to further characterize the estimator under test. Additionally, we introduce the latency/utilization inference graph neural network (lui-gnn), a surrogate model that uses a graph neural network to represent input architectures in the form of a directed graph. This graph representation allows for a diverse set of model architectures to all be effectively handled by a surrogate model. We present the architecture and performance of the model, as evaluated by the new proposed benchmark, including SMAPE, $R^2$ score, and inference times, and find that lui-gnn generally predicts latency and utilization for the 75\% quantile within several percent of the synthesized resources on the synthetic test dataset, indicating that this approach of estimating resource and latency via a surrogate models has promise and warrants further research.

72 PHYSICS OF ELEMENTARY PARTICLES AND FIELDS↗
DOE OSTI2022

Towards Compact Neural Networks via End-to-End Training: A Bayesian Tensor Approach with Automatic Rank Determination

Post-training model compression can reduce the inference costs of deep neural networks, but uncompressed training still consumes enormous hardware resources and energy. To enable low-energy training on edge devices, it is highly desirable to directly train a compact neural network from scratch with a low memory cost. Low-rank tensor decomposition is an effective approach to reduce the memory and computing costs of large neural networks. However, directly training low-rank tensorized neural networks is a very challenging task because it is hard to determine a proper tensor rank a priori, and the tensor rank controls both model complexity and accuracy. Here, this paper presents a novel end-to-end framework for low-rank tensorized training. We first develop a Bayesian model that supports various low-rank tensor formats (e.g., CANDECOMP/PARAFAC, Tucker, tensor-train, and tensor-train matrix) and reduces neural network parameters with automatic rank determination during training. Then we develop a customized Bayesian solver to train large-scale tensorized neural networks. Our training methods shows orders-of-magnitude parameter reduction and little accuracy loss (or even better accuracy) in the experiments. On a very large deep learning recommendation system with over 4.2 ×10 9 model parameters, our method can reduce the parameter number to 1.6 ×10 5 automatically in the training process (i.e., by 2.6 ×10 4 times) while achieving almost the same accuracy. Code is available at https://github.com/colehawkins/bayesian-tensor-rank-determination.

compact neural networks↗
DOE OSTI2022

Bioplastic design using multitask deep neural networks

Abstract Non-degradable plastic waste jeopardizes our environment, yet our modern lifestyle and current technologies are impossible to sustain without plastics. Bio-synthesized and biodegradable alternatives such as polyhydroxyalkanoates (PHAs) have the potential to replace large portions of the world’s plastic supply with cradle-to-cradle materials, but their chemical complexity and diversity limit traditional resource-intensive experimentation. Here, we develop multitask deep neural network property predictors using available experimental data for a diverse set of nearly 23,000 homo- and copolymer chemistries. Using the predictors, we identify 14 PHA-based bioplastics from a search space of almost 1.4 million candidates which could serve as potential replacements for seven petroleum-based commodity plastics that account for 75% of the world’s yearly plastic production. We also discuss possible synthesis routes for the identified promising materials.

36 MATERIALS SCIENCE↗
DOE OSTI2021

Semantic segmentation with a sparse convolutional neural network for event reconstruction in MicroBooNE

We present the performance of a semantic segmentation network, SparseSSNet, that provides pixel-level classification of MicroBooNE data. The MicroBooNE experiment employs a liquid argon time projection chamber for the study of neutrino properties and interactions. SparseSSNet is a submanifold sparse convolutional neural network, which provides the initial machine learning based algorithm utilized in one of MicroBooNE's ν e -appearance oscillation analyses. The network is trained to categorize pixels into five classes, which are re-classified into two classes more relevant to the current analysis. The output of SparseSSNet is a key input in further analysis steps. This technique, used for the first time in liquid argon time projection chambers data and is an improvement compared to a previously used convolutional neural network, both in accuracy and computing resource utilization. Here, the accuracy achieved on the test sample is ≥ 99%. For full neutrino interaction simulations, the time for processing one image is ≈ 0.5 sec, the memory usage is at 1 GB level, which allows utilization of most typical CPU worker machine.

46 INSTRUMENTATION RELATED TO NUCLEAR SCIENCE AND ↗
DOE OSTI2025

SymbolNet: neural symbolic regression with adaptive dynamic pruning for compression

Abstract Compact symbolic expressions have been shown to be more efficient than neural network (NN) models in terms of resource consumption and inference speed when implemented on custom hardware such as field-programmable gate arrays (FPGAs), while maintaining comparable accuracy (Tsoi et al 2024 EPJ Web Conf. 295 09036). These capabilities are highly valuable in environments with stringent computational resource constraints, such as high-energy physics experiments at the CERN Large Hadron Collider. However, finding compact expressions for high-dimensional datasets remains challenging due to the inherent limitations of genetic programming (GP), the search algorithm of most symbolic regression (SR) methods. Contrary to GP, the NN approach to SR offers scalability to high-dimensional inputs and leverages gradient methods for faster equation searching. Common ways of constraining expression complexity often involve multistage pruning with fine-tuning, which can result in significant performance loss. In this work, we propose S y m b o l N e t , a NN approach to SR specifically designed as a model compression technique, aimed at enabling low-latency inference for high-dimensional inputs on custom hardware such as FPGAs. This framework allows dynamic pruning of model weights, input features, and mathematical operators in a single training process, where both training loss and expression complexity are optimized simultaneously. We introduce a sparsity regularization term for each pruning type, which can adaptively adjust its strength, leading to convergence at a target sparsity ratio. Unlike most existing SR methods that struggle with datasets containing more than O ( 10 ) inputs, we demonstrate the effectiveness of our model on the LHC jet tagging task (16 inputs), MNIST (784 inputs), and SVHN (3072 inputs).

Tsoi, Ho Fung (ORCID:0000000225502184)↗
DOE OSTI2020

Decreasing wind speed extrapolation error via domain-specific feature extraction and selection

Abstract. Model uncertainty is a significant challenge in the wind energy industry and can lead to mischaracterization of millions of dollars' worth of wind resources. Machine learning methods, notably deep artificial neural networks (ANNs), are capable of modeling turbulent and chaotic systems and offer a promising tool to produce high-accuracy wind speed forecasts and extrapolations. This paper uses data collected by profiling Doppler lidars over three field campaigns to investigate the efficacy of using ANNs for wind speed vertical extrapolation in a variety of terrains, and it quantifies the role of domain knowledge in ANN extrapolation accuracy. A series of 11 meteorological parameters (features) are used as ANN inputs, and the resulting output accuracy is compared with that of both standard log-law and power-law extrapolations. It is found that extracted nondimensional inputs, namely turbulence intensity, current wind speed, and previous wind speed, are the features that most reliably improve the ANN's accuracy, providing up to a 65 % and 52 % increase in extrapolation accuracy over log-law and power-law predictions, respectively. The volume of input data is also deemed important for achieving robust results. One test case is analyzed in depth using dimensional and nondimensional features, showing that the feature nondimensionalization drastically improves network accuracy and robustness for sparsely sampled atmospheric cases.

17 WIND ENERGY↗
DOE OSTI2022

Meta-optic accelerators for object classifiers

Rapid advances in deep learning have led to paradigm shifts in a number of fields, from medical image analysis to autonomous systems. These advances, however, have resulted in digital neural networks with large computational requirements, resulting in high energy consumption and limitations in real-time decision-making when computation resources are limited. Here, we demonstrate a meta-optic–based neural network accelerator that can off-load computationally expensive convolution operations into high-speed and low-power optics. In this architecture, metasurfaces enable both spatial multiplexing and additional information channels, such as polarization, in object classification. End-to-end design is used to co-optimize the optical and digital systems, resulting in a robust classifier that achieves 93.1% accurate classification of handwriting digits and 93.8% accuracy in classifying both the digit and its polarization state. This approach could enable compact, high-speed, and low-power image and information processing systems for a wide range of applications in machine vision and artificial intelligence.

42 ENGINEERING↗
DOE OSTI2021

Bayesian tensorized neural networks with automatic rank selection

Tensor decomposition is an effective approach to compress over-parameterized neural networks and to enable their deployment on resource-constrained hardware platforms. However, directly applying tensor compression in the training process is a challenging task due to the difficulty of choosing a proper tensor rank. In order to address this challenge, this paper proposes a low-rank Bayesian tensorized neural network. Our Bayesian method performs automatic model compression via an adaptive tensor rank determination. We also present approaches for posterior density calculation and maximum a posteriori (MAP) estimation for the end-to-end training of our tensorized neural network. Here, we provide experimental validation on a two-layer fully connected neural network, a 6-layer CNN and a 110-layer residual neural network where our work produces 7.4x to 137x more compact neural networks directly from the training while achieving high prediction accuracy.

Low-rank tensor↗
DOE OSTI2024

Voltage Calculations in Secondary Distribution Networks via Physics-Inspired Neural Network Using Smart Meter Data

The increasing penetration of distributed energy resources (DERs) leads to voltage issues across distribution networks, necessitating voltage calculations by utilities. Electric model-free voltage calculation offers an enticing solution. However, most researches mainly focus on primary distribution networks ignoring secondary distribution networks and commonly overlook extreme voltage case calculations, which require the model’s extrapolation abilities. Here, in addressing the gaps, this paper presents a customized physics-inspired neural network (PINN) model, the structure of which is inspired by the derived coupled power flow model of primary-secondary distribution networks. To ensure precision and rapid convergence, a crafted training framework for the PINN model is proposed. The PINN’s “structure-mimetic” design enables superior extrapolation for unseen scenarios and enhances physical information awareness. We demonstrate this through two applications: hosting capacity analysis and customer-transformer connectivity. The effectiveness and advantages of the proposed PINN model are validated on two public testing systems and one utility distribution feeder model.

Distribution network↗
DOE OSTI2022

Deep learning to estimate permeability using geophysical data

Time-lapse electrical resistivity tomography (ERT) is a popular geophysical method to estimate three-dimensional (3D) permeability fields from electrical potential difference measurements. Traditional inversion and data assimilation methods are used to ingest this ERT data into hydrogeophysical models to estimate permeability. Due to ill-posedness and the curse of dimensionality, existing inversion strategies provide poor estimates and low resolution of the 3D permeability field. Recent advances in deep learning provide us with powerful algorithms to overcome this challenge. This paper presents a deep learning (DL) framework to estimate the 3D subsurface permeability from time-lapse ERT data. To test the feasibility of the proposed framework, we train DL-enabled inverse models on simulation data. Each measurement in both synthetic and field data is standardized by removing the mean and scaling the time-series to unit variance. This pre-processing step is necessary to bring simulation data closer to field observations. Subsurface process models based on hydrogeophysics are used to generate this synthetic data. Training performed on limited simulation data resulted in the DL model over-fitting. An advanced data augmentation based on mixup is implemented to generate additional training samples to overcome this issue. This mixup technique creates weakly labeled (low-fidelity) samples from strongly labeled (high-fidelity) data. The weakly labeled training data is then used to develop DL-enabled inverse models and reduce over-fitting. As both time-lapse ERT (1133048 features/realization) and 3D permeability (585453 features/realization) data samples are from a high-dimensional space, principal component analysis (PCA) is employed to reduce dimensionality. Encoded ERT and encoded permeability are generated using the trained PCA estimators. A deep neural network is then trained to map the encoded ERT to encoded permeability. This mixup training and unsupervised learning allowed us to build a fast and reasonably accurate DL-based inverse model under limited simulation data. Results show that proposed weak supervised learning can capture salient spatial features in the 3D permeability field. Quantitatively, the average mean squared error (in terms of the natural log) on the strongly labeled training, validation, and test datasets is less than 0.5. The R 2 -score (global metric) is greater than 0.75, and the percent error in each cell (local metric) is less than 10%. Finally, an added benefit in terms of computational cost is that the proposed DL-based inverse model is at least O(10 4 ) times faster than running a forward model once it is trained. Data generation, DL model training, and hyperparameter tuning to identify optimal neural network architectures utilized high-performance computing resources while the DL inference is performed on a standard laptop. Approximately, O(10 5 ) processor hours are used for generating data and DL tuning and training. We acknowledge that the data generation and DL model development are expensive. But once a DL model is trained, it can be re-used for inversion rapidly for the given system, with set physics and domain. Note that traditional inversion may require multiple forward model simulations (e.g., in the order of 10 to 1000), which are very expensive. This computational savings ≈ O(10 5 ) – O(10 7 )) makes the proposed DL-based inverse model attractive for subsurface imaging and real-time ERT monitoring applications due to fast and yet reasonably accurate estimations of permeability field.

58 GEOSCIENCES↗
DOE OSTI2023

ML-based Real-Time Control at the Edge: An Approach Using hls4ml

This study focuses on implementing a real-time control system for a particle accelerator facility that performs high energy physics experiments. A critical operating parameter in this facility is beam loss, which is the fraction of particles deviating from the accelerated proton beam into a cascade of secondary particles. Accelerators employ a large number of sensors to monitor beam loss. The data from these sensors is monitored by human operators who predict the relative contribution of different sub-systems to the beam loss. Using this information, they engage control interventions. In this paper, we present a controller to track this phenomenon in real-time using edge-Machine Learning (ML) and support control with low latency and high accuracy. We implemented this system on an Intel Arria 10 SoC. Optimizations at the algorithm, high-level synthesis, and interface levels to improve latency and resource usage are presented. Our design implements a neural network, which can predict the main source of beam loss (between two possible causes) at speeds up to 575 frames per second (fps) (average latency of 1.74 ms). The practical deployed system is required to operate at 320 fps, with a 3ms latency requirement, which has been met by our design successfully.

Shi, R.↗
DOE OSTI2022

DRAS: Deep Reinforcement Learning for Cluster Scheduling in High Performance Computing

Cluster schedulers are crucial in high-performance computing (HPC). They determine when and which user jobs should be allocated to available system resources. Existing cluster scheduling heuristics are developed by human experts based on their experience with specific HPC systems and workloads. However, the increasing complexity of computing systems and the highly dynamic nature of application workloads have placed tremendous burden on manually designed and tuned scheduling heuristics. More aggressive optimization and automation are needed for cluster scheduling in HPC. In this work, we present an automated HPC scheduling agent named DRAS (Deep Reinforcement Agent for Scheduling) by leveraging deep reinforcement learning. DRAS is built on a hierarchical neural network incorporating special HPC scheduling features such as resource reservation and backfilling. An efficient training strategy is presented to enable DRAS to rapidly learn the target environment. Once being provided a specific scheduling objective given by the system manager, DRAS automatically learns to improve its policy through interaction with the scheduling environment and dynamically adjusts its policy as workload changes. We implement DRAS into a HPC scheduling platform called CQGym. CQGym provides a common platform allowing users to flexibly evaluate DRAS and other scheduling methods such as heuristic and optimization methods. Furthermore, the experiments using CQGym with different production workloads demonstrate that DRAS outperforms the existing heuristic and optimization approaches by up to 50%.

97 MATHEMATICS AND COMPUTING↗
DOE OSTI2025

Toward accelerating rare-earth metal extraction using equivariant neural networks

The separation of rare-earth metals, vital for numerous advanced technologies, is hampered by their similar chemical properties, making ligand discovery a significant challenge. Traditional experimental and quantum chemistry approaches for identifying effective ligands are often resource-intensive. We introduce a machine learning protocol based on an equivariant neural network, Allegro, for the rapid and accurate prediction of binding energies in rare-earth complexes. Key to this work is our newly curated dataset of rare-earth metal complexes—made publicly available to foster further research—systematically generated using the Architector program. This dataset distinctively features functionalized derivatives of proven rare-earth-chelating scaffolds, hydroxypyridinone (HOPO), catecholamide (CAM), and their thio-analogues, selected for their established efficacy in binding these elements. Trained on this valuable resource, our Allegro models demonstrate excellent performance, particularly when trained to directly predict DFT-level binding energies, yielding highly accurate results that closely correlate with theoretical calculations on a diverse test set. Furthermore, this strategy exhibited strong out-of-sample generalization, accurately predicting binding energies for an isomeric HOPO-derivative ligand not seen during training. By substantially reducing computational demands, this machine learning framework, alongside the provided dataset, represent powerful tools to accelerate the high-throughput screening and rational design of novel ligands for efficient rare-earth metal separation.

Gupta, Ankur K. [Lawrence Berkeley National Labora↗
DOE OSTI2026

wa-hls4ml: A Benchmark and Surrogate Models for hls4ml Resource and Latency Estimation

As machine learning (ML) is increasingly implemented in hardware to address real-time challenges in scientific applications, the development of advanced toolchains has significantly reduced the time required to iterate on various designs. These advancements have solved major obstacles, but also exposed new challenges. For example, processes that were not previously considered bottlenecks, such as hardware synthesis, are becoming limiting factors in the rapid iteration of designs. To mitigate these emerging constraints, multiple efforts have been undertaken to develop an ML-based surrogate model that estimates resource usage of ML accelerator architectures. We introduce wa-hls4ml, a benchmark for ML accelerator resource and latency estimation, and its corresponding initial dataset of over 680,000 fully connected and convolutional neural networks, all synthesized using hls4ml and targeting Xilinx FPGAs. The benchmark evaluates the performance of resource and latency predictors against several common ML model architectures, primarily originating from scientific domains, as exemplar models, and the average performance across a subset of the dataset. Additionally, we introduce GNN- and transformer-based surrogate models that predict latency and resources for ML accelerators. We present the architecture and performance of the models and find that the models generally predict latency and resources for the 75% percentile within several percent of the synthesized resources on the synthetic test dataset.

Hawks, Benjamin [Fermilab] (ORCID:0000000157000288↗
DOE OSTI2025

wa-hls4ml: A Benchmark and Surrogate Models for hls4ml Resource and Latency Estimation

As machine learning (ML) is increasingly implemented in hardware to address real-time challenges in scientific applications, the development of advanced toolchains has significantly reduced the time required to iterate on various designs. These advancements have solved major obstacles, but also exposed new challenges. For example, processes that were not previously considered bottlenecks, such as hardware synthesis, are becoming limiting factors in the rapid iteration of designs. To mitigate these emerging constraints, multipleefforts have been undertaken to develop an ML-based surrogate model that estimates resource usage of synthesized ML accelerator architectures. We introduce wa-hls4ml, a benchmark for ML accelerator resource and latency estimation, and its corresponding initial dataset of over 680 000 fully connected and convolutional neural networks, all synthesized using hls4ml and targeting Xilinx FPGAs. The benchmark evaluates the performance of resource and latency predictors against several common ML model architectures, primarily originating from scientific domains, as exemplar models, and the average performance across a subset of the dataset. Additionally, we introduce GNN- and transformer-based surrogate models that predict latency and resources for ML accelerators. We present the architecture and performance of the models and find that the models generally predict latency and resources for the 75% percentile within several percent of the synthesized resources on the synthetic test dataset.

Hawks, Benjamin G. [Fermilab]↗
DOE OSTI2023

Ultra-low latency recurrent neural network inference on FPGAs for physics applications with hls4ml

Abstract Recurrent neural networks have been shown to be effective architectures for many tasks in high energy physics, and thus have been widely adopted. Their use in low-latency environments has, however, been limited as a result of the difficulties of implementing recurrent architectures on field-programmable gate arrays (FPGAs). In this paper we present an implementation of two types of recurrent neural network layers—long short-term memory and gated recurrent unit—within the hls4ml framework. We demonstrate that our implementation is capable of producing effective designs for both small and large models, and can be customized to meet specific design requirements for inference latencies and FPGA resources. We show the performance and synthesized designs for multiple neural networks, many of which are trained specifically for jet identification tasks at the CERN Large Hadron Collider.

97 MATHEMATICS AND COMPUTING↗