Search NASA⌕ Search

SEARCH · Search NASA

Results for “High Performance Networking”

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 721 records · Page 40
All sourcesNASA NTRSDOE OSTI
DOE OSTI2021

Energy-efficient Mott activation neuron for full-hardware implementation of neural networks

To circumvent the von Neumann bottleneck, substantial progress has been made towards in-memory computing with synaptic devices. However, compact nanodevices implementing non-linear activation functions are required for efficient full-hardware implementation of deep neural networks. Here, in this work, we present an energy-efficient and compact Mott activation neuron based on vanadium dioxide and its successful integration with a conductive bridge random access memory (CBRAM) crossbar array in hardware. The Mott activation neuron implements the rectified linear unit function in the analogue domain. The neuron devices consume substantially less energy and occupy two orders of magnitude smaller area than those of analogue complementary metal–oxide semiconductor implementations. The LeNet-5 network with Mott activation neurons achieves 98.38% accuracy on the MNIST dataset, close to the ideal software accuracy. We perform large-scale image edge detection using the Mott activation neurons integrated with a CBRAM crossbar array. Our findings provide a solution towards large-scale, highly parallel and energy-efficient in-memory computing systems for neural networks.

electrical and electronic engineering↗
DOE OSTI2024

“Green” Fabrication of High-performance Transparent Conducting Electrodes by Blade Coating and Photonic Curing on PET for Perovskite Solar Cells

This study presents an innovative material processing approach to fabricate transparent conducting electrodes (TCEs) on polyethylene terephthalate (PET) substrates using blade coating and photonic curing. The hybrid TCEs consist of a multiscale Ag network, combining silver metal bus lines and nanowires, overcoated by an indium zinc oxide layer, and then photonically cured. Blade coating ensures film uniformity and thickness control over large areas. Photonic curing, a non-thermal processing method with significantly lower carbon emissions, enhances the conductivity and transparency of the coated layers. Our hybrid TCEs achieve an average transmittance of (81 ± 0.4)% referenced to air ((90 ± 0.4)% referenced to the PET substrate) in the visible range, an average sheet resistance of (11 ± 0.5) Ω sq−1, and an average surface roughness of (4.3 ± 0.4) nm. We benchmark these values against commercial PET/TCE substrates. Mechanical durability tests demonstrate <3% change in resistance after 2000 bending cycles at a 1 in radius. The scalable potential of the hybrid TCE fabrication method is demonstrated by high uniformity and excellent properties in 7 in × 8 in large-area samples and by performing the photonic curing process at 11 m min−1. Furthermore, halide perovskite solar cells fabricated on these hybrid TCEs achieve average and champion power conversion efficiencies of (10.5 ± 1.0) % and 12.2%, respectively, and significantly outperform devices made on commercial PET/TCEs. This work showcases our approach as a viable pathway for high-speed “green” manufacturing of high-performance TCEs on PET substrates for flexible optoelectronic devices.

Bonner, Justin C↗
DOE OSTI2020

Rapid Annotation of Photosynthetic Systems (RAPS): automated algorithm to generate genome-scale metabolic networks from algal genomes

Algae have mounting potential to serve as platform strains for engineering efforts, but tools for these species lag behind traditional model systems like E. coli and yeast. Metabolic models are one such tool that would enhance our ability to rationally engineer algae but current automated algorithms to build models do not perform well for algal genomes. We introduce a novel software pipeline, Rapid Annotation of Photosynthetic Systems (RAPS), which leverages manual curation efforts of published models to create high quality first draft metabolic networks for new species. We compared models produced by our pipeline to published models and found that these models are able to capture more genes than the published models and are more predictive of experimentally determined growth rate. We used RAPS to automate the creation of 8 first draft metabolic models of new species to enable the further study of metabolism in these species.

09 BIOMASS FUELS↗
DOE OSTI2020

Multimodal Data Representation with Deep Learning for Extracting Cancer Characteristics from Clinical Text

This paper presents a multimodal data representation to improve the performance of deep learning models for extracting cancer key characteristics from unstructured text in pathology reports. Specifically, in addition to using the text as the input to deep learning models, we use concept unique identifiers (CUIs) as another source of information to the models. We analyze the performance of different text and CUI data representations, including word embeddings and bag of embeddings (BOE), with a convolutional neural network (CNN) and a fully connected multilayer perceptron neural network (MLP-NN). The high level document embeddings from text and CUI inputs are combined by concatenating them and then applying a classifier. The model is used for extracting cancer subsite and histology from pathology reports. These two classification tasks have a large number of labels, i.e. 317 for subsite and 556 for histology, with extreme class imbalance. We compare the performance of the developed DL models across the two tasks based on micro- and macro-F1 scores. The evaluation shows that a multi-channel DL model that utilizes text represented by word embeddings and CUIs represented by BOE outperforms other DL models. Also, this approach significantly improves the model performance on low prevalence classes.

Alawad, Mohammed↗
DOE OSTI2021

Lithium Oxide Superionic Conductors Inspired by Garnet and NASICON Structures

Abstract The key component in lithium solid‐state batteries (SSBs) is the solid electrolyte composed of lithium superionic conductors (SICs). Lithium oxide SICs offer improved electrochemical and chemical stability compared with sulfides, and their recent advancements have largely been achieved using materials in the garnet‐ and NASICON (sodium superionic conductor)‐ structured families. In this work, using the ion‐conduction mechanisms in garnet and NASICON as inspiration, a common pattern of an “activated diffusion network” and three structural features that are beneficial for superionic conduction: a 3D percolation Li diffusion network, short distances between occupied Li sites, and the “homogeneity” of the transport path are identified. A high‐throughput computational screening is performed to search for new lithium oxide SICs that share these features. From this search, seven candidates are proposed exhibiting high room‐temperature ionic conductivity evaluated using ab initio molecular dynamics simulations. Their structural frameworks including spinel, oxy‐argyrodite, sodalite, and LiM(SeO 3 ) 2 present new opportunities for enriching the structural families of lithium oxide SICs.

36 MATERIALS SCIENCE↗
DOE OSTI2018

Optimal Control of Connected and Automated Vehicles at Roundabouts: An Investigation in a Mixed-Traffic Environment

Connectivity and automation in vehicles provide the most intriguing opportunity for enabling users to better monitor transportation network conditions and make better operating decisions to improve safety and reduce pollution, energy consumption, and travel delays. This paper investigates the implications of optimally coordinating vehicles that are wirelessly connected to each other in roundabouts to achieve a smooth traffic flow. We apply an optimization framework and an analytical solution that allows optimal coordination of vehicles for merging in such traffic scenario. The effectiveness of the proposed approach is validated through simulation and it is shown that fully coordinated vehicles reduce total travel time by 51% and fuel consumption by 35%. Furthermore, we study the influence of vehicle coordination in a mixed-traffic environment and compare the network performance under different market penetration rates of connected and automated vehicles (CAVs). For this particular study with near-capacity demand, high penetration of CAVs is necessary to ensure an efficient and effective control, due to extremely unstable traffic.

Zhao, Liuhui↗
DOE OSTI2022

Machine-learning from Pseudomonas putida KT2440 transcriptomes reveals its transcriptional regulatory network

Bacterial gene expression is orchestrated by numerous transcription factors (TFs). Elucidating how gene expression is regulated is fundamental to understanding bacterial physiology and engineering it for practical use. In this study, a machine-learning approach was applied to uncover the genome-scale transcriptional regulatory network (TRN) in Pseudomonas putida KT2440, an important organism for bioproduction. We performed independent component analysis of a compendium of 321 high-quality gene expression profiles, which were previously published or newly generated in this study. We identified 84 groups of independently modulated genes (iModulons) that explain 75.7% of the total variance in the compendium. With these iModulons, we (i) expand our understanding of the regulatory functions of 39 iModulon associated TFs (e.g., HexR, Zur) by systematic comparison with 1993 previously reported TF-gene interactions; (ii) outline transcriptional changes after the transition from the exponential growth to stationary phases; (iii) capture group of genes required for utilizing diverse carbon sources and increased stationary response with slower growth rates; (iv) unveil multiple evolutionary strategies of transcriptome reallocation to achieve fast growth rates; and (v) define an osmotic stimulon, which includes the Type VI secretion system, as coordination of multiple iModulon activity changes. Taken together, this study provides the first quantitative genome-scale TRN for P. putida KT2440 and a basis for a comprehensive understanding of its complex transcriptome changes in a variety of physiological states.

09 BIOMASS FUELS↗
DOE OSTI2026

Machine learning for seismic low-frequency extrapolation

The cycle-skipping problem that plagues full waveform inversion (FWI) can be at least partially mitigated if low frequencies (which encode the kinematics of wave propagation in seismic data) are recorded. However, seismic sources and receivers are band-limited, so seismic data does not generally include signals down to 0 Hz. To improve our ability to solve the seismic inverse problem, one can synthesize this missing low-frequency (LF) content from the recorded high-frequency (HF) data using machine learning (ML) models. Deep learning models such as convolutional neural networks (CNNs) demonstrate impressive ability to perform low frequency extrapolation. However, such models require powerful hardware (GPU machines) and careful training. We assess the extrapolation capabilities of three different ML models that do not require GPU machines, namely, random forest, Gaussian process regression and gradient boosting, on both synthetic and real data. Experimental results on two synthetic data sets (generated from a low velocity lens embedded in a homogeneous medium, and the Marmousi model) demonstrate that FWI applied to the extrapolated data consistently improves inversion accuracy relative to FWI applied to the original data sets that do not contain low frequencies. Application of low-frequency extrapolation to real data from the Northwest Shelf of Australia demonstrates that tree-based ML models such as gradient boosting can outperform CNNs in terms of both accuracy and computational cost on non-GPU architectures.

58 GEOSCIENCES↗
DOE OSTI2019

Combining Tensor Decompositions and Graph Analytics to Provide Cyber Situational Awareness at HPC Scale

This paper describes MADHAT (Multidimensional Anomaly Detection fusing HPC, Analytics, and Tensors), an integrated workflow that demonstrates the applicability of HPC resources to the problem of maintaining cyber situational awareness. MADHAT combines two high-performance packages: ENSIGN for large-scale sparse tensor decompositions and HAGGLE for graph analytics. Tensor decompositions isolate coherent patterns of network behavior in ways that common clustering methods based on distance metrics cannot. Parallelized graph analysis then uses directed queries on a representation that combines the elements of identified patterns with other available information (such as additional log fields, domain knowledge, network topology, whitelists and blacklists, prior feedback, and published alerts) to confirm or reject a threat hypothesis, collect context, and raise alerts. MADHAT was developed using the collaborative HPC Architecture for Cyber Situational Awareness (HACSAW) research environment and evaluated on structured network sensor logs collected from Defense Research and Engineering Network (DREN) sites using HPC resources at the U.S. Army Engineer Research and Development Center DoD Supercomputing Resource Center (ERDC DSRC). To date, MADHAT has analyzed logs with over 650 million entries.

Cyber security, Tensor decomposition, graph analyt↗
DOE OSTI2024

“Understanding Robustness Lottery”: A Geometric Visual Comparative Analysis of Neural Network Pruning Approaches

Deep learning approaches have provided state-of-the-art performance in many applications by relying on large and overparameterized neural networks. However, such networks are very brittle and are difficult to deploy on resource-limited platforms. Model pruning, i.e., reducing the size of the network, is a widely adopted strategy that can lead to a more robust and compact model. Many heuristics exist for model pruning, but our understanding of the pruning process remains limited due to the black-box nature of a neural network model. Empirical studies show that some heuristics improve performance whereas others can make models more brittle. Here, this work aims to shed light on how different pruning methods alter the network’s internal feature representation and the corresponding impact on model performance. To facilitate a comprehensive comparison and characterization of the high-dimensional model feature space, we introduce a visual geometric analysis of feature representations. We evaluated a set of critical geometric concepts decomposed from the commonly adopted classification loss and used them to design a visualization system to compare and highlight the impact of pruning on model performance and feature representation. The proposed tool provides an environment for an in-depth comparison of pruning methods and a comprehensive understanding of how the model responds to common data corruption. By leveraging the proposed visualization, machine learning researchers can reveal the similarities between pruning methods and redundancy in robustness evaluation benchmarks, obtain geometric insights about the differences between pruned models that achieve superior robustness performance, and identify samples that are robust or fragile to model pruning and common data corruption.

Li, Zhimin [Univ. of Utah, Salt Lake City, UT (Uni↗
DOE OSTI2019

NoC-enabled Software/Hardware Co-Design Framework for Accelerating k-mer Counting

Counting k-mers (substrings of fixed length k) in DNA and protein sequences generate non-uniform and irregular memory access patterns. Processing-in-Memory (PIM) architectures have potential to significantly reduce the overheads associated with such frequent and irregular memory accesses. However, existing k-mer counting algorithms are not designed to exploit the advantages of PIM architectures. Furthermore, owing to thermal constraints, the allowable power budget is limited in conventional PIM designs. Moreover, k-mer counting generates long-range, non-uniform traffic patterns which needs to be handled by an efficient Network-on-Chip (NoC). In this paper, we present an NoC-enabled hardware/software co-design framework to implement high-performance k-mer counting. The proposed architecture enables more computational power, efficient communication between cores/memory – all without creating a thermal bottleneck. The software component overcomes the I/O bottleneck, aids in NoC design and exposes more in-memory opportunities to exploit in PIM. Experimental results show that the proposed architecture outperforms a state-of-the-art software implementation using Hybrid Memory Cube (HMC), by up to 7.14X, while allowing significantly higher power budget.

Joardar, Biresh K.↗
DOE OSTI2018

Synchrotron Radiation Beam Diagnostics for the Integrable Optics Test Accelerator

The Integrable Optics Test Accelerator (IOTA) is a research electron and proton storage ring currently being commissioned at Fermilab's Accelerator Science and Technology (FAST) facility. An extensive beam physics research program is planned, including tests of novel techniques for improving beam intensity, stability, and emittance. A key part of IOTA beam diagnostics suite are synchrotron light beam monitors, mounted onto each dipole. In this paper, we present the hardware and software design of this system. Mechanical layout and actuator control electronics are described. High throughput image acquisition and analysis architecture is outlined, and its preliminary performance is explored. Integration of the system within accelerator control network and possible user applications, such as camera auto-focusing, are discussed.

43 PARTICLE ACCELERATORS↗
DOE OSTI2018

Fibre Channel Development Final Report CRADA No. TC-0340-92-D

The Fibre Channel (FC) Development project was designed to enhance U.S. industrial competitiveness by accelerating the standardization and subsequent industrial acceptance of the American National Standards Institute’s (AIW) Fibre Channel Standard (now X3T1 1). It was also intended to give the principal U.S. participants a one- to two-year lead in developing heterogeneous high-speed computer networks, dktibuted computing through clustering, scientific visualization techniques, and high speed channels. The project was proposed to benefit U.S. industry as these high performance computing methods became generally available. In 1992-93, during the initial project phases, Fibre Channel PC) was emerging as a new ANSI standard. It was envisioned as becoming the mechanism for data transfer between the next generation of computers, and between these computers and their peripherals. Using this channel standard, U.S. computer vendors were expecting to be able to provide high-speed interoperability between heterogeneous computers and their storage devices, with data transfer rates greater than a GBit/sec. Our project was composed of key principal players in the FC standardization effort from both U.S. industry and the Department of Energy’s (DOES) Lawrence Liverrnore National Laboratory (LLNL). The industial participants were developing individual FC interfaces to operate with their platform products. LLNL’s role was to be a neutral testing site and a facilitator for helping to resolve conflicting interpretations of the embryonic Fibre Channel specification. LLNL anticipated that conflicts in the interpretation of such a standard would arise and require neutral resolution and that such resolutions would best be achieved in cooperation with the ANSI FC Working Group. Project plans included transferring to U.S. industry any resilient FC testing capabilities at the project conclusion.

42 ENGINEERING↗
DOE OSTI2018

Fibre Channel Development Final Report CRADA No. TC-0340-92-C

The Fibre Channel (FC) Development project was designed to enhance U.S. industrial competitiveness by accelerating the standardization and subsequent industrial acceptance of the American National Standards Institute’s (ANSI) Fibre Channel Standard (now X3T1 1). It was also intended to give the principal U.S. participants a one-to two-year lead in developing heterogeneous high-speed computer networks, distributed computing through clustering, scientific visualization techniques, and high speed channels. The project was proposed to benefit U.S. industry as these high performance computing methods became generally available. In 1992-93, during the initial project phases, Fibre Channel (FC) was emerging as a new ANSI standard. It was envisioned as becoming the mechanism for data transfer between the next generation of computers, and between these computers and their peripherals. Using this channel standard, U.S. computer vendors were expecting to be able to provide high-speed interoperatillity between heterogeneous computers and their storage devices, with data transfer rates greater than a GBit/sec. Our project was composed of key principal players in the FC standardization effort from both U.S. industry and the Department of Energy’s@OEs) Lawrence Liverrnore National Laboratory (LLNL). The industrial participants were developing individual FC interfaces to operate with their platform products. LLNL’s.role was to be a neutral testing site and a facilitator for helping to resolve conflicting interpretations of the embryonic Fibre Channel specification. LLNL anticipated that conflicts in the interpretation of such a standard would arise and require neutral resolution and that such resolutions would best be achieved in cooperation with the ANSI FC Working Group. Project plans included transferring to U.S. industry any resilient FC testing capabilities at the project conclusion.

42 ENGINEERING↗
DOE OSTI2014

Synthesis and Properties of Nanostructured Exotic Superconductors (Final Report)

Superconductors are materials that conduct electricity with no resistance, i.e. no energy dissipation. In future electrically activated nanodevices, the use of superconducting interconnects is highly desirable, since it circumvents the damaging heat produced by energy dissipation in normal nano-conductors whose high resistance is inversely proportional to its cross-section area. Furthermore, the wave functions of the charge carriers - Cooper pairs will be strongly modified in a superconductor when they are squeezed into a small volume comparable to the coherence length or penetration depth. Thus, superconducting nanostructures provide unique platforms to investigate and discover novel superconducting phenomena in confined geometries. The objective of this DOE funded project was the development of novel approaches to synthesizing a new class of superconducting nanostructures such as nanowires and nanoribbons that are stable in atmosphere, enabling the exploration of superconducting properties and potential applications of individual nanostructures. We also pursued the modification of properties of the fabricated nanostructures by introducing artificial defects through nanoengineering. We developed a conversion approach comprised of two-steps to synthesize NbSe 2 and NbN nanowires/ribbons that were not achievable using other routes: first, we synthesized one-dimensional (1D) nanostructures of NbSe 3 and second, we converted these into desired 1D superconducting nanostructures of NbSe 2 and NbN by adjusting the composition of NbSe 3 nanostructure precursors while maintaining their shapes. Novel phenomena such as resistance oscillations with increasing magnetic fields were observed in both NbSe 2 and NbN nanowires/ribbons. By replacing the catalytic microparticles used in conventional whisker growth with the same kind of nanoparticles that are available now commercially, we improved whisker technologies for nanofabrication of high-T c superconductor nanowires/ribbons: we succeeded in the growth of Bi 2 Sr 2 CaCu 2 O 8+x (BSCCO-2212) and YBa 2 Cu 3 O 7-x (YBCO) nanowires and nanoribbons by annealing glassy precursors with alumina nanoparticles as growth sites. The high-quality whiskers enabled us to observe the ‘matching effect’ that represents itself as dips in the field dependence of the magnetoresistance and the influence of the periodic defects on the melting behavior of vortex matter. We also observed N-shaped V-I characteristics similar to those observed in semiconductor and plasma devices for the first time. Each nanofabrication technique has its advantages and drawbacks. For example, electron-beam lithography (EBL), in combination thin film deposition, has been widely used to fabricate nanowires (nanobridge) with well-defined cross-sections. However, electron-beam resist usually contaminates the sample up to a few hundred nanometers and hence degrades the quality of the entire nanowires if their widths are only a few hundred nanometers. We found that e-beam resist can degrade the superconducting critical temperature of a Nb nanowire from 9 K down to below 4 K if its width is less than 200 nm. We investigated the suitability of the various nanofabrication methods for achieving high-quality samples with desired properties. For example, we developed a novel way to achieve high quality Nb nanostrips with transverse dimensions of 150 nm or smaller. In the new process a superconducting bridge a few micrometers wide is fabricated with EBL. Subsequently, focused-ion-beam (FIB) is used to mill nano- 2 grooves to isolate an uncontaminated area of the microbridge to form a nanostrip. The high-quality superconducting Nb nanostrips with well-defined rectangular cross-sections fabricated using this new technique have enabled experimental confirmation of the theoretically predicted confinement-induced anisotropy effect on the magnetoresistance and critical field. We also found that FIB milling is more suitable than EBL in producing a pinning potential array with desired barrier between them into a MoGe film, which enabled us to realize the predicted vortex ices for the first time. By utilizing a filtration membrane as the substrate, we developed a new method to achieve networks of superconducting MoGe nanowires with transverse dimensions smaller than 10 nm. We observed superconducting confinement effect such as thermally activated phase slips and Little-Parks effect in these new samples. This research significantly advanced our knowledge on confined superconductors, in both their synthesis and physical properties. Though the new fabrication methods developed in this project aimed to achieve samples for studying new phenomena in confined superconductors, they can be utilized to obtain nanostructures of other functional materials. For example, we utilized the method developed to produce superconducting nanowire networks for the fabrication of networks of ultrasmall palladium (Pd) nanowires, which take advantage of single Pd nanowires while eliminating their nanofabrication obstacles, resulting in high-performance hydrogen sensors that are much more cost effective than those based on single Pd nanowires fabricated with lithography methods.

75 CONDENSED MATTER PHYSICS, SUPERCONDUCTIVITY AND↗
DOE OSTI2025

Seasonal Reconfiguration of Electrical Distribution Systems to Mitigate the Impact of Electric Vehicle Charging

Power grids face challenges in their infrastructure related to the integration of electric vehicles (EV). In particular, EV charging stations may induce instability in key system parameters such as substantial voltage drops, active power losses, and transformer overload due to high demand during charging periods. This article presents a seasonal reconfiguration strategy based on the differential evolution (DE) algorithm, aimed at enhancing system performance under highly variable and stochastic load profiles, particularly those driven by EV charging. The DEA algorithm is hybridized with the find-union (FU) algorithm to efficiently ensure network radiality throughout the optimization process. The proposed methodology is validated on a hybrid distribution system composed of the IEEE 33-bus network, a modified IEEE 13-bus system, and a specific 13-bus microgrid. Results have demonstrated that seasonal reconfiguration significantly reduces active power losses and mitigates transformer loading during critical demand hours, thereby quantifiably increasing the system’s performance. As an integral component of the proposed approach, an analysis of CO2 emissions associated with energy losses is included, allowing a contextualized assessment of the environmental benefits of seasonal reconfiguration in various geographical areas.

24 POWER TRANSMISSION AND DISTRIBUTION↗
DOE OSTI2022

Non-Blind Deblurring for Fluorescence: A Deformable Latent Space Approach with Kernel Parameterization

We report N\non-blind deblurring (NBD) is a modeling method of the image deblurring problem in computer vision, where the blurring kernel is known or can be externally estimated. In this paper, we attempt to solve a parametric NBD problem, inspired by the simultaneous acquisition of ptychography and fluorescent imaging (FI). Ptychography is an imaging method that favors larger probes, i.e. convolutional kernels, while FI relies on a small probe for high resolution. Also, the kernel can be solved during ptychographic reconstruction. With Ptycho-FI using the same larger kernel, we can perform NBD on the blurred fluorescent images to achieve high-resolution FI, and thus speed up the experiments. To this end, we design a deep latent space deformation network that is directly parameterized by the kernel. The network consists of three components: encoder, deformer, and decoder, where the deformer is specifically meant to rectify the latent space representations of blurred images to a standard latent space, regardless of the kernel. The deformation network is trained with a two-stage training scheme. We conduct extensive experiments to confirm that our parametric model can adapt to drastically different blurring kernels and perform robust deblurring.

77 NANOSCIENCE AND NANOTECHNOLOGY↗
DOE OSTI2023

Semi-supervised graph neural networks for pileup noise removal

Abstract The high instantaneous luminosity of the CERN Large Hadron Collider leads to multiple proton–proton interactions in the same or nearby bunch crossings (pileup). Advanced pileup mitigation algorithms are designed to remove this noise from pileup particles and improve the performance of crucial physics observables. This study implements a semi-supervised graph neural network for particle-level pileup noise removal, by identifying individual particles produced from pileup. The graph neural network is firstly trained on charged particles with known labels, which can be obtained from detector measurements on data or simulation, and then inferred on neutral particles for which such labels are missing. This semi-supervised approach does not depend on the neutral particle pileup label information from simulation, and thus allows us to perform training directly on experimental data. The performance of this approach is found to be consistently better than widely-used domain algorithms and comparable to the fully-supervised training using simulation truth information. The study serves as the first attempt at applying semi-supervised learning techniques to pileup mitigation, and opens up a new direction of fully data-driven machine learning pileup mitigation studies.

43 PARTICLE ACCELERATORS↗