Engineering Papers⌕ Search

SEARCH · Engineering Papers

Results for “regression models”

Search indexed NASA NTRS and DOE OSTI research on propulsion, heat transfer, battery materials and energy systems. Follow report and document links to the original sources.

Quote a phrase for an exact phrase match. Source license links do not imply unrestricted reuse.

At least 1,621 records · Page 90
All sourcesNASA NTRSDOE OSTI
DOE OSTI2022

Simultaneous quantification of uranium( VI ), samarium, nitric acid, and temperature with combined ensemble learning, laser fluorescence, and Raman scattering for real-time monitoring

In this work, laser-induced fluorescence spectroscopy (LIFS), Raman spectroscopy, and a stacked regression ensemble was developed for near real-time quantification of uranium(VI) (1–100 μg mL –1 ), samarium (0–200 μg mL –1 ) and nitric acid (0.1–4 M) with varying temperature (20 °C–45 °C). LIFS applications range from fundamental lab-scale studies to real-time process monitoring at industrial levels, such as nuclear reprocessing applications, provided the phenomena affecting the fluorescence spectrum are accounted for (e.g., absorption, quenching, complexation). Multiple chemometric models were examined and compared to a more traditional multivariate regression approach called partial least squares (PLS). Results obtained on synthetic samples selected using D-optimal experimental design indicated that a stacked regression method, which included ridge regression, random forest, PLS, and an eXtreme gradient boost algorithm, successfully measured uranium(VI) concentrations directly in nitric acid without measuring luminescence lifetimes or standard addition. The top model resulted in percent root-mean-square error of prediction values of 5.2, 1.9, 3.0, and 2.3% for U(VI), Sm 3+ , HNO 3 , and temperature, respectively. The approach may be useful for quantifying fluorescent fission products (e.g., Sm 3+ ) to provide information on burnup of irradiated nuclear fuel. This novel framework reinforces the applicability of LIFS for real-time applications in nuclear fuel cycle applications.

38 RADIATION CHEMISTRY, RADIOCHEMISTRY, AND NUCLEA↗
DOE OSTI2021

The Prediction Model of Risk Factors for COVID-19 Developing into Severe Illness Based on 1046 Patients with COVID-19

This study analyzed the risk factors for patients with COVID-19 developing severe illnesses and explored the value of applying the logistic model combined with ROC curve analysis to predict the risk of severe illnesses at COVID-19 patients’ admissions. The clinical data of 1046 COVID-19 patients admitted to a designated hospital in a certain city from July to September 2020 were retrospectively analyzed, the clinical characteristics of the patients were collected, and a multivariate unconditional logistic regression analysis was used to determine the risk factors for severe illnesses in COVID-19 patients during hospitalization. Based on the analysis results, a prediction model for severe conditions and the ROC curve were constructed, and the predictive value of the model was assessed. Logistic regression analysis showed that age (OR = 3.257, 95% CI 10.466–18.584), complications with chronic obstructive pulmonary disease (OR = 7.337, 95% CI 0.227–87.021), cough (OR = 5517, 95% CI 0.258–65.024), and venous thrombosis (OR = 7322, 95% CI 0.278–95.020) were risk factors for COVID-19 patients developing severe conditions during hospitalization. When complications were not taken into consideration, COVID-19 patients’ ages, number of diseases, and underlying diseases were risk factors influencing the development of severe illnesses. The ROC curve analysis results showed that the AUC that predicted the severity of COVID-19 patients at admission was 0.943, the optimal threshold was −3.24, and the specificity was 0.824, while the sensitivity was 0.827. The changes in the condition of severe COVID-19 patients are related to many factors such as age, clinical symptoms, and underlying diseases. This study has a certain value in predicting COVID-19 patients that develop from mild to severe conditions, and this prediction model is a useful tool in the quick prediction of the changes in patients’ conditions and providing early intervention for those with risk factors.

Lian, Zhichuang↗
DOE OSTI2022

Development of Short-Term Forecasting Models Using Plant Asset Data and Feature Selection

Nuclear power plants collect and store large volumes of heterogeneous data from various components and systems. With recent advances in machine learning (ML) techniques, these data can be leveraged to develop diagnostic and short-term forecasting models to better predict future equipment condition. Maintenance operations can then be planned in advance whenever degraded performance is predicted, thus resulting in fewer unplanned outages and the optimization of maintenance activities. This enables lower maintenance costs and improves the overall economics of nuclear power. This paper focuses on developing a short-term forecasting process that leverages a feature selection process to distill large volumes of heterogeneous data and predict specific equipment parameters. A variety of feature selection methods, including Shapley Additive Explanations (SHAP) and variance inflation factor (VIF), were used to select the optimal features as inputs for three ML methods: long short-term memory (LSTM) networks, support vector regression (SVR), and random forest (RF). Each combination of model and input features was used to predict a pump bearing temperature both 1 and 24 hours in advance, based on actual plant system data. The optimal inputs for the LSTM and SVR were selected using the SHAP values, while the optimal input for the RF consisted solely of the response variable itself. Each model produced similar 1-hour-ahead predictions, with root mean square errors (RMSEs) of roughly 0.006. For the 24-hour-ahead predictions, differences could be seen between LSTM, SVR, and RF, as reflected by model performances of 0.036 +- 0.014, 0.0026 +- 0, and 0.063 +- 0.004 RMSE, respectively. As big data and continuous online monitoring become more widely available, the proposed feature selection process can be used for many applications beyond the prediction of process parameters within nuclear infrastructure.

46 INSTRUMENTATION RELATED TO NUCLEAR SCIENCE AND ↗
DOE OSTI2023

Automated Credibility Assessments of User Features in Scientific Software

Scientific software (SciSoft) is complex, often containing a mixture of production capabilities co-mingled with features under active research and development. Furthermore, SciSoft is often developed over decades by non-computer scientists who may not have a strong background in or prioritize software architecture design, testing, and quality (e.g., test coverage). These conditions lead to difficulty in understanding which software components or functions implement what user-facing features and therefore those features’ software quality pedigree. This lack of understanding poses challenges in assessing readiness and credibility of user features, and often relies on a SciSoft subject matter expert’s (SME) laborious investigation and assertion. This final report of a one-year Computing and Information Sciences Lab Directed Research and Development project presents a general framework for modeling SciSoft architecture as a direct relationship between user features and the software components/functions that implement them. Our approach leverages automated labeling of the SciSoft’s regression test suite and employs machine learning algorithms to construct the architecture model. We demonstrate this framework on the Solid Mechanics component of the SIERRA multi-physics engineering analysis suite developed at Sandia National Laboratories.

97 MATHEMATICS AND COMPUTING↗
DOE OSTI2020

Active- and transfer-learning applied to microscale-macroscale coupling to simulate viscoelastic flows

Active- and transfer-learning are applied to microscale dynamics of polymer flows for the multiscale discovery of effective constitutive approximations required in viscoelastic flow simulation. The result is macroscopic rheology directly connected to a microstructural model. Micro and macroscale simulations are adaptively coupled by means of Gaussian process regression (GPR) to run the expensive microscale computations only as necessary. This multiscale method is demonstrated with flows of a polymer solution as a model system. At the microscale level dissipative particle dynamics (DPD) is employed to model the fluid as a suspension of bead-spring micro-structures subjected to steady shear flow. The results yield the non-Newtonian viscosity and the first normal stress difference at strain rates as training data used in a GPR model. DPD parameters are calibrated with respect to experimental data for a real polymer solution. Compliance with these data requires adjustment of the DPD model's cutoff radius, which then becomes a function of the second invariant of the strain rate tensor. The FENE-P model is chosen for the macroscale description using the spectral element method (SEM) to simulate channel flow and flow past a circular cylinder. The DPD results at the lowest possible shear strain rate yield an estimate of the zero-shear rate viscosity, which allows the initiation of the macroscale flow by SEM as a Newtonian fluid. The resulting strain-rate field is surveyed to determine additional shear strain rate sampling points for the DPD system. This new information allows an initial fitting of parameters of the constitutive equation followed by new SEM simulations at the macroscale. Additionally, guided by active-learning GPR to select new sampling points, this process continues until convergence is achieved. The effectiveness of this new simulation paradigm for viscoelastic flows is tested with different macroscale operating conditions. The effective closure learned in the channel simulation is then transferred directly to the flow past a circular cylinder at low Reynolds number, where the results show that only two additional DPD simulations are required to achieve a satisfactory constitutive model. With an increase of the Reynolds number, the active-learning scheme automatically detects the inaccuracy of the learned constitutive model, and initiates additional DPD simulations for the extra data needed to once again close the microscale-macroscale coupled system. This new paradigm of active- and transfer-learning for multiscale modeling is readily applicable to other microscale-macroscale coupled simulations of complex fluids and other materials. Furthermore, the coupling between microscale and macroscale solvers can be seamlessly implemented with our open source multiscale universal interface (MUI) library.

71 CLASSICAL AND QUANTUM MECHANICS, GENERAL PHYSIC↗
DOE OSTI2024

A multifidelity Bayesian optimization method for inertial confinement fusion design

Due to their cost, experiments for inertial confinement fusion (ICF) heavily rely on numerical simulations to guide design. As simulation technology progresses, so too can the fidelity of models used to plan for new experiments. However, these high-fidelity models are by themselves insufficient for optimal experimental design, because their computational cost remains too high to efficiently and effectively explore the numerous parameters required to describe a typical experiment. Therefore, traditionally, ICF design has relied on low-fidelity modeling to initially identify potentially interesting design regions, which are then subsequently explored via selected high-fidelity modeling. In this paper, we demonstrate that this two-step approach can be insufficient: even for simple design problems, a two-step optimization strategy can lead high-fidelity searching toward incorrect regions and consequently waste computational resources on parameter regimes far away from the true optimal solution. We reveal that a primary cause of this behavior in ICF design problems is the presence of low-fidelity optima in different regions of the parameter space far away from high-fidelity optima. To address this issue, we propose an iterative multifidelity Bayesian optimization method based on Gaussian Process Regression that leverages both low- and high-fidelity models simultaneously. We demonstrate, using both two- and eight-dimensional ICF test problems, that our algorithm can effectively utilize both low-fidelity and high-fidelity models to refine the designs. This approach proves to be more efficient than relying solely on high-fidelity modeling for optimization.

70 PLASMA PHYSICS AND FUSION TECHNOLOGY↗
DOE OSTI2023

Zero-truncated Poisson regression for sparse multiway count data corrupted by false zeros

Abstract We propose a novel statistical inference methodology for multiway count data that is corrupted by false zeros that are indistinguishable from true zero counts. Our approach consists of zero-truncating the Poisson distribution to neglect all zero values. This simple truncated approach dispenses with the need to distinguish between true and false zero counts and reduces the amount of data to be processed. Inference is accomplished via tensor completion that imposes low-rank tensor structure on the Poisson parameter space. Our main result shows that an $N$-way rank-$R$ parametric tensor $\boldsymbol{\mathscr{M}}\in (0,\infty )^{I\times \cdots \times I}$ generating Poisson observations can be accurately estimated by zero-truncated Poisson regression from approximately $IR^2\log _2^2(I)$ non-zero counts under the nonnegative canonical polyadic decomposition. Our result also quantifies the error made by zero-truncating the Poisson distribution when the parameter is uniformly bounded from below. Therefore, under a low-rank multiparameter model, we propose an implementable approach guaranteed to achieve accurate regression in under-determined scenarios with substantial corruption by false zeros. Several numerical experiments are presented to explore the theoretical results.

97 MATHEMATICS AND COMPUTING↗
DOE OSTI2021

Importance of Engineered and Learned Molecular Representations in Predicting Organic Reactivity, Selectivity, and Chemical Properties

Machine-readable chemical structure representations are foundational in all attempts to harness machine learning for the prediction of reactivities, selectivities, and chemical properties directly from molecular structure. The featurization of discrete chemical structures into a continuous vector space is a critical phase undertaken before model selection, and the development of new ways to quantitatively encode molecules is an active area of research. Here, we highlight the application and suitability of different representations, from expert-guided “engineered” descriptors to automatically “learned” features, in different prediction tasks relevant to organic and organometallic chemistry, where differing amounts of training data are available. These tasks include statistical models of stereo- and enantioselectivity, thermochemistry, and kinetics developed using experimental and quantum chemical data. The use of expert-guided molecular descriptors provides an opportunity to incorporate chemical knowledge, domain expertise, and physical constraints into statistical modeling. In applications to stereoselective organic and organometallic catalysis, where data sets may be relatively small and 3D-geometries and conformations play an important role, mechanistically informed features can be used successfully to obtain predictive statistical models that are also chemically interpretable. We provide an overview of several recent applications of this approach to obtain quantitative models for reactivity and selectivity, where topological descriptors, quantum mechanical calculations of electronic and steric properties, along with conformational ensembles, all feature as essential ingredients of the molecular representations used. Alternatively, more flexible, general-purpose molecular representations such as attributed molecular graphs can be used with machine learning approaches to learn the complex relationship between a structure and prediction target. This approach has the potential to out-perform more traditional representation methods such as “hand-crafted” molecular descriptors, particularly as data set sizes grow. One area where this is particularly relevant is in the use of large sets of quantum mechanical data to train quantitative structure–property relationships. A general approach toward curating useful data sets and training highly accurate graph neural network models is discussed in the context of organic bond dissociation enthalpies, where this strategy outperforms regression using precomputed descriptors. Finally, we describe how graph neural network predictions can be incorporated into mechanistically informed statistical models of chemical reactivity and selectivity. Once trained, this approach avoids the expensive computational overhead associated with quantum mechanical calculations, while maintaining chemical interpretability. We illustrate examples for which fast predictions of bond dissociation enthalpy and of the identities of radicals formed through cleavage of a molecule’s weakest bond are used in simple physical models of site-selectivity and reactivity.

37 INORGANIC, ORGANIC, PHYSICAL, AND ANALYTICAL CH↗
DOE OSTI2022

Fast increased fidelity samplers for approximate Bayesian Gaussian process regression

Gaussian processes (GPs) are common components in Bayesian non-parametric models having a rich methodological literature and strong theoretical grounding. The use of exact GPs in Bayesian models is limited to problems containing several thousand observations due to their prohibitive computational demands. We develop a posterior sampling algorithm using H-matrix approximations that scales at O(n log 2 n). We show that this approximation’s Kullback-Leibler divergence to the true posterior can be made arbitrarily small. Though multidimensional GPs could be used with our algorithm, d-dimensional surfaces are modeled as tensor products of univariate GPs to minimize the cost of matrix construction and maximize computational efficiency. We illustrate the performance of this fast increased fidelity approximate GP, FIFA-GP, using both simulated and non-synthetic data sets

97 MATHEMATICS AND COMPUTING↗
NASA NTRS1981

The use of frequency methods in rotorcraft system identification

A new approach to model structure determination is examined. Flight data from the Rotor Systems Research Aircraft (RSRA) are transformed into the frequency domain and truncated to provide band limiting. The stepwise regression technique is then used to identify a quasistatic state-space model from the transformed data. The data processing requirements for both time domain and frequency domain identification are discussed and the results of the two techniques are compared.

Duval, R. W.↗
NASA NTRS2012

Greenhouse Gas Concentration Data Recovery Algorithm for a Low Cost, Laser Heterodyne Radiometer

The goal of a coordinated effort between groups at GWU and NASA GSFC is the development of a low-cost, global, surface instrument network that continuously monitors three key carbon cycle gases in the atmospheric column: carbon dioxide (CO2), methane (CH4), carbon monoxide (CO), as well as oxygen (O2) for atmospheric pressure profiles. The network will implement a low-cost, miniaturized, laser heterodyne radiometer (mini-LHR) that has recently been developed at NASA Goddard Space Flight Center. This mini-LHR is designed to operate in tandem with the passive aerosol sensor currently used in AERONET (a well established network of more than 450 ground aerosol monitoring instruments worldwide), and could be rapidly deployed into this established global network. Laser heterodyne radiometry is a well-established technique for detecting weak signals that was adapted from radio receiver technology. Here, a weak light signal, that has undergone absorption by atmospheric components, is mixed with light from a distributed feedback (DFB) telecommunications laser on a single-mode optical fiber. The RF component of the signal is detected on a fast photoreceiver. Scanning the laser through an absorption feature in the infrared, results in a scanned heterodyne signal io the RF. Deconvolution of this signal through the retrieval algorithm allows for the extraction of altitude contributions to the column signal. The retrieval algorithm is based on a spectral simulation program, SpecSyn, developed at GWU for high-resolution infrared spectroscopies. Variations io pressure, temperature, composition, and refractive index through the atmosphere; that are all functions of latitude, longitude, time of day, altitude, etc.; are modeled using algorithms developed in the MODTRAN program developed in part by the US Air Force Research Laboratory. In these calculations the atmosphere is modeled as a series of spherically symmetric shells with boundaries specified at defined altitudes. Temperature, pressure, and species mixing ratios are defined at these boundaries. Between the boundaries, temperature is assumed to vary linearly with altitude while pressure (and thus gas density) vary exponentially. The observed spectrum at the LHR instrument will be the integration of the contributions along this light path. For any absorption measurement the signal at a particular spectral frequency is a linear combination of spectral line contributions from several species. For each species that might absorb in a spectral region, we have pre-calculated its contribution as a function of temperature and pressure. The integrated path absorption spectrum can then by calculated using the initial sun angle (from location, date, and time) and assumptions about pressure and temperature profiles from an atmospheric model. The modeled spectrum is iterated to match the experimental observation using standard multilinear regression techniques. In addition to the layer concentrations, the numerical technique also provides uncertainty estimates for these quantities as well as dependencies on assumptions inherent in the atmospheric models.

Miller, J. Houston↗
DOE OSTI2022

Hyperspectral Reflectance for Measuring Canopy‐Level Nutrients and Photosynthesis in a Salt Marsh

Abstract Salt marsh ecosystems are underrepresented in process‐based models due to their unique location across the terrestrial–aquatic interface. Particularly, the role of leaf nutrients on canopy photosynthesis (F A ) remains unclear, despite their relevance for regulating vegetation growth. We combined multiyear information of canopy‐level nutrients and eddy covariance measurements with canopy surface hyperspectral remote sensing (CSHRS) to quantify the spatial and temporal variability of F A in a temperate salt marsh. We found that F A showed a positive relationship with canopy‐level N at the ecosystem scale and for areas dominated by Spartina cynosuroides , but not for areas dominated by short S. alterniflora . F A showed a positive relationship with canopy‐level P, K, and Na, but a negative relationship with Fe, for areas associated with S. cynosuroides , S. alterniflora , and at the ecosystem scale. We used partial least squares regression (PLSR) with CSHRS and found statistically significant data–model agreements to predict canopy‐level nutrients and F A . The red‐edge electromagnetic region and ∼770 nm showed the highest contribution of variance in PLSR models for canopy‐level nutrients and F A , but we propose that underlying sediment biogeochemistry can complicate interpretation of reflectance measurements. Our findings highlight the relevance of spatial variability in salt marshes vegetation and the promising application of CSHRS for linking information of canopy‐level nutrients with F A . We call for further development of canopy surface hyperspectral methods and analyses across salt marshes to improve our understanding of how these ecosystems will respond to global environmental change.

Environmental Sciences & Ecology↗
DOE OSTI2026

Machine learning inversion of interatomic force constants from single-crystal inelastic neutron scattering

Atomic vibrations govern many macroscopic properties of materials, but experiments to comprehensively probe them remain challenging. Inelastic neutron scattering (INS) is a powerful technique to map phonon dispersions in crystals, especially when leveraging modern time-of-flight (ToF) spectrometers with large detectors. However, efficiently and robustly extracting interatomic force constants (FCs) parameterizing phonon dynamics from experimental spectra remains a bottleneck due to the complexity and high dimensionality of ToF INS datasets. Here, we present a machine learning approach for the direct inversion of FCs from single-crystal INS measurements. The framework leverages synthetic training data generated using universal machine-learned force fields and an efficient physics-based forward model. We benchmark two neural architectures–one emphasizing structured latent representation learning and the other direct, supervised spectral regression–across simulated datasets for two materials under idealized and noisy conditions. The latent-representation model is subsequently applied to experimental single-crystal INS data on germanium. The model is shown to reproduce FCs derived from both first-principles simulations and from iterative optimization, and furthermore achieves reliable inference even from sparse, single-orientation measurements representing short data acquisitions. Analysis of the learned latent space reveals semantically continuous and physically interpretable encodings that support strong cross-domain generalization. By bridging theoretical and experimental domains, we establish a path toward rapid inversion of experimental spectra and data-driven interpretation of temperature-dependent lattice dynamics.

42 ENGINEERING↗
NASA NTRS1984

Prediction Model for the Life of Nickel-cadmium Batteries in Geosynchronous Orbit Satellites

A mathematical model is described which predicts the service life of nickel-cadmium batteries designed for geosynchronous orbit satellites. Regression analysis technique is used to analyze orbital data on second generation trickle charged batteries. The model gives average cell voltage as a function of design parameters, operating parameters and time. The voltage model has the properties of providing a good fit to the data, good predictive capability, and agreement with known battery performance characteristics. Average cell voltage can be predicted to within 0.02 volts for up to 8 years. This modeling shows that these batteries will operate reliably for 10 years. Third-generation batteries are expected to operate even longer.

Engleman, J. H.↗
DOE OSTI2021

Fast and Non-Destructive Determination of Water Content in Ionic Liquids at Varying Temperatures by Raman Spectroscopy and Multivariate Regression Analysis

Imidazolium acetate ionic liquids (ILs) have been utilized as promising solvents in many applications that involve varying water content and temperature. These experimental variables affect the anion-cation intermolecular interactions, which in turn influence the performance of the ILs in these applications. Here, this paper shows Raman spectroscopy can be used as an operando method to measure water content in IL solvents when simultaneous temperature changes may occur. The Raman spectra of 1-alkyl-3-methylimidazolium acetate ILs (alkyl chain length n = 2, 4, 6, 8) with varying water content (from 0.028 to 0.899 water mole fraction) and temperature (from 78.1 K to 423.1 K) were measured. Increasing the water content or decreasing the temperature of the tested ILs weakens the anion-cation intermolecular interactions. The water content of these ILs can be quantified even in conditions when the temperature is changing using Raman spectroscopy combined with multivariate regression analysis, including principal component regression (PCR), partial-least-squares regression (PLSR), and artificial neural networks (ANNs). The ANN model combined with partial-least-squares (PLS) achieves the highest prediction accuracy of water content in ILs at varying temperatures (RMSECV = 0.017, R 2 CV = 99.1%, RMSEP = 0.019, R 2 P = 98.8%, RPD = 8.93). Raman spectroscopy provides a potential fast non-destructive operando method to monitor the water content of ILs even in applications when the temperature may be simultaneously altered; this information can lead to the optimized use of these ILs in many applications.

37 INORGANIC, ORGANIC, PHYSICAL, AND ANALYTICAL CH↗
NASA NTRS1993

Flap-lag damping in hover and forward flight with a three-dimensional wake

Prediction of lag damping is difficult owing to the delicate balance of drag, induced drag and Coriolis forces in the in-plane direction. Moreover, induced drag is sensitive to dynamic wake, both shed and trailing components, and thus its prediction requires adequate unsteady-wake representation. Accordingly, rigid-blade flap-lag equations are coupled with a three-dimensional finite-state wake model; three isolated rotor configurations with three, four and five blades are treated over a range of thrust levels, Lock numbers, lag frequencies and advance ratios. The investigation includes convergence characteristics of damping with respect to the number of radial shape functions and harmonics of the wake model for multiblade modes of low frequency (less than 1/rev.) to high frequency (greater than 1/rev.). Predicted flap and lag damping levels are then compared with similar predictions with (1) rigid wake (no unsteady induced flow, (2) Loewy lift defficiency and (3) dynamic inflow. The coverage also includes correlations with the measured lag regressive-mode damping in hover and forward flight and comparisons with similar correlations with dynamic wake model are consistently higher than the predictions with the dynamic inflow model; even for the low frequency lag regressive mode, the number of wake harmonics should at least be equal to twice the number of blades.

Manjunath, A. R.↗
DOE OSTI2025

Sparse-Data Deep Learning Strategies for Radiographic Non-Destructive Testing

Radiography is an imaging technique used in a variety of applications, such as medical diagnosis, airport security, and nondestructive testing. We present a deep learning system for extracting information from radiographic images. We perform various prediction tasks using our system, including material classification and regression on the dimensions of a given object that is being radiographed. Our system is designed to address the sparse-data issue for radiographic nondestructive testing applications. It uses a radiographic simulation tool for synthetic data augmentation, and it uses transfer learning with a pre-trained convolutional neural network model. Using this system, our preliminary results indicate that the object geometry regression task saw an improvement of 70% in the R-squared value when using a multi-regime model. In addition, we increase the performance of the object material classification tasks by utilizing data from different imaging systems. In particular, using neutron imaging improved the material classification accuracy by 20% when compared to x-ray imaging.

convolutional neural networks↗