Machine Learning Regression Techniques Research Articles

Bio-oil production from biogenic wastes, the hydrothermal conversion step.

Background: Food wastes are an abundant resource that can be effectively valorised by hydrothermal liquefaction to produce bio-fuels. The objective of the European project WASTE2ROAD is to demonstrate the complete value chain from waste collection to engine tests. The principle of hydrothermal liquefaction is well known but there are still many factors that make the science very empirical. Most experiments in the literature are performed on batch reactors. Comparison of results from batch reactors with experiments with continuous reactors are rare in the literature. Methods: Various food wastes were transformed by hydrothermal liquefaction. The resources used and the products from the experiments have been extensively analysed. Two different experimental reactors have been used, a batch reactor and a continuous reactor. This paper presents a dataset of fully documented experiments performed in this project, on food wastes with different compositions, conditions and solvents. The data set is extended with data from the literature. The data was analysed using machine learning analysis and regression techniques. Results: This paper presents experimental results on various food wastes as well as modelling and analysis with machine learning algorithms. The experimental results were used to attempt to establish a link between batch and continuous experiments. The molecular weight of bio-oil from continuous experiments appear higher than that of batch experiments. This may be due to the configuration of our reactor. Conclusions: This paper shows how the use of regression models help with understanding the results, and the importance of process variables and resource composition. A novel data analysis technique gives an insight on the accuracy that can be obtained from these models.

Kullback–Leibler Importance Estimation Procedure to Improve Gas Quantification in an Electronic Nose

An electronic nose sensor array can classify and quantify different types of gases; however, the sensor can alter its measurement capability over time. The main problem presented during the measurements of the sensors is related to the variation of the data acquired for long periods due to changes in the chemosensory response, thus affecting the correct functioning of the implemented measuring system. This research presents an approach to improve gas quantification through the implementation of machine learning regression techniques in an array of nose-type electronic sensors. The implemented methodology uses a domain adaptation approach with the Kullback–Leibler importance estimation procedure (KLIEP) to improve the performance of the gas quantification electronic nose array. This approach is validated using a three-year dataset measured by a 16-electronic-nose-sensor array. The R2 regression error obtained for each of the gases fits the resulting dataset’s measured values with good precision.

Searching glycolate oxidase inhibitors based on QSAR, molecular docking, and molecular dynamic simulation approaches

Primary hyperoxaluria type 1 (PHT1) treatment is mainly focused on inhibiting the enzyme glycolate oxidase, which plays a pivotal role in the production of glyoxylate, which undergoes oxidation to produce oxalate. When the renal secretion capacity exceeds, calcium oxalate forms stones that accumulate in the kidneys. In this respect, detailed QSAR analysis, molecular docking, and dynamics simulations of a series of inhibitors containing glycolic, glyoxylic, and salicylic acid groups have been performed employing different regression machine learning techniques. Three robust models with less than 9 descriptors—based on a tenfold cross (Q2CV) and external (Q2EXT) validation—were found i.e., MLR1 (Q2CV = 0.893, Q2EXT = 0.897), RF1 (Q2CV = 0.889, Q2EXT = 0.907), and IBK1 (Q2CV = 0.891, Q2EXT = 0.907). An ensemble model was built by averaging the predicted pIC50 of the three models, obtaining a Q2EXT = 0.933. Physicochemical properties such as charge, electronegativity, hardness, softness, van der Waals volume, and polarizability were considered as attributes to build the models. To get more insight into the potential biological activity of the compouds studied herein, docking and dynamic analysis were carried out, finding the hydrophobic and polar residues show important interactions with the ligands. A screening of the DrugBank database V.5.1.7 was performed, leading to the proposal of seven commercial drugs within the applicability domain of the models, that can be suggested as possible PHT1 treatment.

Improving Water Quality Index Prediction Using Regression Learning Models.

Rivers are the main sources of freshwater supply for the world population. However, many economic activities contribute to river water pollution. River water quality can be monitored using various parameters, such as the pH level, dissolved oxygen, total suspended solids, and the chemical properties. Analyzing the trend and pattern of these parameters enables the prediction of the water quality so that proactive measures can be made by relevant authorities to prevent water pollution and predict the effectiveness of water restoration measures. Machine learning regression algorithms can be applied for this purpose. Here, eight machine learning regression techniques, including decision tree regression, linear regression, ridge, Lasso, support vector regression, random forest regression, extra tree regression, and the artificial neural network, are applied for the purpose of water quality index prediction. Historical data from Indian rivers are adopted for this study. The data refer to six water parameters. Twelve other features are then derived from the original six parameters. The performances of the models using different algorithms and sets of features are compared. The derived water quality rating scale features are identified to contribute toward the development of better regression models, while the linear regression and ridge offer the best performance. The best mean square error achieved is 0 and the correlation coefficient is 1.

Bio-oil production from biogenic wastes, the hydrothermal conversion step

Background: Food wastes are an abundant resource that can be effectively valorised by hydrothermal liquefaction to produce bio-fuels. The objective of the European project WASTE2ROAD is to demonstrate the complete value chain from waste collection to engine tests. The principle of hydrothermal liquefaction is well known but there are still many factors that make the science very empirical. Most experiments in the literature are performed on batch reactors. Comparison of results from batch reactors with experiments with continuous reactors are rare in the literature. Methods: Various food wastes were transformed by hydrothermal liquefaction. The resources used and the products from the experiments have been extensively analysed. Two different experimental reactors have been used, a batch reactor and a continuous reactor. This paper presents a dataset of fully documented experiments performed in this project, on food wastes with different compositions, conditions and solvents. The data set is extended with data from the literature. The data was analysed using machine learning analysis and regression techniques. Results: This paper presents experimental results on various food wastes as well as modelling. Aqueous phase recycling allows the re-use of some of the solubilised organics, but this paper shows that after some recycles, the yield is principally in the char yield and not so much in the oil yield. The experimental results were further used to attempt to establish a link between batch and continuous experiments. The molecular weight of bio-oil from continuous experiments appear higher than that of batch experiments. This may be due to the configuration of our reactor. Conclusions: This paper shows how the use of regression models help with understanding the results, and the importance of process variables and resource composition. A novel data analysis technique gives an insight on the accuracy that can be obtained from these models.

An investigation of machine learning techniques to estimate minimum horizontal stress magnitude from borehole breakout

Borehole breakout is a widely utilised phenomenon in horizontal stress orientation determination, and breakout geometrical parameters, such as width and depth, have been used to estimate both horizontal stress magnitudes. However, the accuracy of minimum horizontal stress estimation from borehole breakout remains relatively low in comparison to maximum horizontal stress estimation. This paper aims to compare and improve the minimum horizontal stress estimation via a number of machine learning (ML) regression techniques, including parametric and non-parametric models, which have rarely been explored. ML models were trained based on 79 laboratory data from published literature and validated against 23 field data. A systematic bias was observed in the prediction for the validation dataset whenever the horizontal stress value exceeded the maximum value in the training data. Nevertheless, the pattern was captured, and the removal of systematic bias showed that the artificial neural network is capable of predicting the minimum horizontal stress with an average error rate of 10.16% and a root mean square error of 3.87 MPa when compared to actual values obtained through conventional in-situ measurement techniques. This is a meaningful improvement considering the importance of in-situ stress knowledge for underground operations and the availability of borehole breakout data.

Prediction of Joint Acceleration of 2 DOF Robot Manipulator Using Supervised Learning

Robo-analyzer (RA) is open-source software that uses a 3D representation of a robot manipulator to carry out various analytical studies. It was created primarily to assist instructors and students in getting started with robotics teaching and learning utilizing framework-based skeleton models or computer aided design (CAD) software designs of serial robots i.e., articulated robot. The RA software is used in this work to simulate and examine a two-degree of freedom (DOF) robot with two link and two revolute joints respectively. The joint length is kept constant at 0.2m, and the joint velocity is varied from 0 to 180 degrees per second. The two-link manipulator is permitted to carry out forward kinematics after generating and establishing the input parameters for the simulation of the 2DOF model, which results in simulating the joint acceleration values, and that is the primary prerequisite for the machine learning (ML) process. The model tends to deduce the relationship between the input and output parameters in this study, which further aids in the deduction of a linear relationship between the two parameters, especially input and output parameter i.e., link length coordinates, joint velocity, and joint acceleration. The experimentation was then carried out on the basis of RA data to apply linear regression machine learning technique (LRMLT), which will assist in the prediction of an output, namely joint acceleration. The model tends to pave way for future research which can be carried out for joint vibration which is solely based on the basis of the acceleration present at joint.

Machine learning based iterative learning control for non‐repetitive time‐varying systems

AbstractThe repetitive tracking task for time‐varying systems (TVSs) with non‐repetitive time‐varying parameters at each trial, which is also called non‐repetitive TVSs, is realized in this article using iterative learning control (ILC). A machine learning (ML) based nominal model update mechanism, which utilizes the linear regression technique to update the nominal model at each ILC trial using only the current trial information, is proposed for non‐repetitive TVSs in order to enhance the ILC performance. Given that the ML mechanism forces the model uncertainties to remain within the ILC robust tolerance, an ILC update law is proposed to deal with non‐repetitive TVSs. How to tune parameters inside ML and ILC algorithms to achieve the desired aggregate performance is also provided. The robustness and reliability of the proposed method are verified by real experiments. Real data comparison with current state‐of‐the‐art methods demonstrates its superior control performance in terms of controlling precision. This article broadens ILC applications from time‐invariant systems to non‐repetitive TVSs, adopts ML regression technique to estimate non‐repetitive time‐varying parameters between two ILC trials and proposes a detailed parameter tuning mechanism to achieve desired performance, which are the main contributions.

Spatiotemporal neural network for estimating surface NO2 concentrations over north China and their human health impact

Atmospheric nitrogen dioxide (NO2) is an important reactive gas pollutant harmful to human health. The spatiotemporal coverage provided by traditional NO2 monitoring methods is insufficient, especially in the suburban and rural areas of north China, which have a high population density and experience severe air pollution. In this study, we implemented a spatiotemporal neural network (STNN) model to estimate surface NO2 from multiple sources of information, which included satellite and in situ measurements as well as meteorological and geographical data. The STNN predicted NO2 with high accuracy, with a coefficient of determination (R2) of 0.89 and a root mean squared error of 5.8 μg/m3 for sample-based 10-fold cross-validation. Based on the surface NO2 concentration determined by the STNN, we analyzed the spatial distribution and temporal trends of NO2 pollution in north China. We found substantial drops in surface NO2 concentrations ranging between 9.1% and 33.2% for large cities during the 2020 COVID-19 lockdown when compared to those in 2019. Moreover, we estimated the all-cause deaths attributed to NO2 exposure at a high spatial resolution of about 1 km, with totals of 6082, 4200, and 18,210 for Beijing, Tianjin, and Hebei Provinces in 2020, respectively. We observed remarkable regional differences in the health impacts due to NO2 among urban, suburban, and rural areas. Generally, the STNN model could incorporate spatiotemporal neighboring information and infer surface NO2 concentration with full coverage and high accuracy. Compared with machine learning regression techniques, STNN can effectively avoid model overfitting and simultaneously consider both spatial and temporal correlations of input variables using deep convolutional networks with residual blocks. The use of the proposed STNN model, as well as the surface NO2 dataset, can benefit air quality monitoring, forecasting, and health burden assessments.

A stacked regression ensemble approach for the quantitative determination of biomass feedstock compositions using near infrared spectroscopy

Rapid, robust, and accurate biomass compositional analyses are required in the bioenergy industry to accurately determine the chemical composition of biomass feedstocks. A stacked regression ensemble approach using near infrared spectroscopic method was developed for the quantitative determination of glucan, xylan, lignin, ash, and extract in biomass feedstocks. A comprehensive comparison of the performance of various machine learning techniques including support vector regression (linear and radial), least absolute shrinkage and selection operator (LASSO), ridge regression, elastic net, partial least squares, random forests, recursive partitioning and regression trees, gradient boosting, and gaussian process regression was assessed in the training set data (n = 188). The predictive performance of the aforementioned machine learning approaches was then compared with stacked regression, an ensemble learning algorithm which collates the performance of the abovementioned machine learning regression techniques. Results show that the stacked regression primarily outperformed other machine learning techniques (Root mean square error of prediction (RMSEP)average=1.660%wt,R2=0.907) across all five constituents in the validation set data (n = 81). Further results also show that the RMSEP of the stacked ensemble technique is significantly different than that of the partial least squares (PLS) approach in predicting glucan, ash, lignin, and extract components in biomass samples. The stacked ensemble learning approach offers an alternative method for a more accurate prediction of biomass compositions than the traditional PLS technique.

Calibration of RGB sensor for estimation of real-time correlated color temperature using machine learning regression techniques

Correlated Color Temperature (CCT) is an important parameter to determine the quality of lighting in an indoor space. We have presented here the calibration of RGB sensor for estimation of real-time CCT values using Support Vector Machine Regression, General Regression Neural Network, and Gaussian Process Regression techniques. Further, comparative performance assessment have been done on the evaluating parameters: Percentage Absolute Error, R-Squared Error, Mean Absolute Error, Mean Absolute Percentage Error, and Root Mean Squared Error with respect to a calibrated meter. The RGB values have been acquired from a sensor interfaced with a microcontroller and the CCT data are simultaneously observed from a calibrated chroma meter at the client terminal. The machine learning regression techniques have been applied at the server terminal to find out the CCT values and a comparative performance analysis have been done to find the best possible prediction model that can be compared to a standard meter based on the performance indices to decide on their accuracy.

Fast and Accurate Quantum Mechanical Modeling of Large Molecular Systems Using Small Basis Set Hartree-Fock Methods Corrected with Atom-Centered Potentials.

There has been significant interest in developing fast and accurate quantum mechanical methods for modeling large molecular systems. In this work, by utilizing a machine learning regression technique, we have developed new low-cost quantum mechanical approaches to model large molecular systems. The developed approaches rely on using one-electron Gaussian-type functions called atom-centered potentials (ACPs) to correct for the basis set incompleteness and the lack of correlation effects in the underlying minimal or small basis set Hartree-Fock (HF) methods. In particular, ACPs are proposed for ten elements common in organic and bioorganic chemistry (H, B, C, N, O, F, Si, P, S, and Cl) and four different base methods: two minimal basis sets (MINIs and MINIX) plus a double-ζ basis set (6-31G*) in combination with dispersion-corrected HF (HF-D3/MINIs, HF-D3/MINIX, HF-D3/6-31G*) and the HF-3c method. The new ACPs are trained on a very large set (73 832 data points) of noncovalent properties (interaction and conformational energies) and validated additionally on a set of 32 048 data points. All reference data are of complete basis set coupled-cluster quality, mostly CCSD(T)/CBS. The proposed ACP-corrected methods are shown to give errors in the tenths of a kcal/mol range for noncovalent interaction energies and up to 2 kcal/mol for molecular conformational energies. More importantly, the average errors are similar in the training and validation sets, confirming the robustness and applicability of these methods outside the boundaries of the training set. In addition, the performance of the new ACP-corrected methods is similar to complete basis set density functional theory (DFT) but at a cost that is orders of magnitude lower, and the proposed ACPs can be used in any computational chemistry program that supports effective-core potentials without modification. It is also shown that ACPs improve the description of covalent and noncovalent bond geometries of the underlying methods and that the improvement brought about by the application of the ACPs is directly related to the number of atoms to which they are applied, allowing the treatment of systems containing some atoms for which ACPs are not available. Overall, the ACP-corrected methods proposed in this work constitute an alternative accurate, economical, and reliable quantum mechanical approach to describe the geometries, interaction energies, and conformational energies of systems with hundreds to thousands of atoms.

Estimation of Above-Ground Biomass of Winter Wheat Based on Consumer-Grade Multi-Spectral UAV

One of the problems of optical remote sensing of crop above-ground biomass (AGB) is that vegetation indices (VIs) often saturate from the middle to late growth stages. This study focuses on combining VIs acquired by a consumer-grade multiple-spectral UAV and machine learning regression techniques to (i) determine the optimal time window for AGB estimation of winter wheat and to (ii) determine the optimal combination of multi-spectral VIs and regression algorithms. UAV-based multi-spectral data and manually measured AGB of winter wheat, under five nitrogen rates, were obtained from the jointing stage until 25 days after flowering in the growing season 2020/2021. Forty-four multi-spectral VIs were used in the linear regression (LR), partial least squares regression (PLSR), and random forest (RF) models in this study. Results of LR models showed that the heading stage was the most suitable stage for AGB prediction, with R2 values varying from 0.48 to 0.93. Three PLSR models based on different datasets performed differently in estimating AGB in the training dataset (R2 = 0.74~0.92, RMSE = 0.95~2.87 t/ha, MAE = 0.75~2.18 t/ha, and RPD = 2.00~3.67) and validation dataset (R2 = 0.50~0.75, RMSE = 1.56~2.57 t/ha, MAE = 1.44~2.05 t/ha, RPD = 1.45~1.89). Compared with PLSR models, the performance of the RF models was more stable in the prediction of AGB in the training dataset (R2 = 0.95~0.97, RMSE = 0.58~1.08 t/ha, MAE = 0.46~0.89 t/ha, and RPD = 3.95~6.35) and validation dataset (R2 = 0.83~0.93, RMSE = 0.93~2.34 t/ha, MAE = 0.72~2.01 t/ha, RPD = 1.36~3.79). Monitoring AGB prior to flowering was found to be more effective than post-flowering. Moreover, this study demonstrates that it is feasible to estimate AGB for multiple growth stages of winter wheat by combining the optimal VIs and PLSR and RF models, which overcomes the saturation problem of using individual VI-based linear regression models.

Regression Analysis for COVID-19 Infections and Deaths Based on Food Access and Health Issues.

COVID-19, or SARS-CoV-2, is considered as one of the greatest pandemics in our modern time. It affected people’s health, education, employment, the economy, tourism, and transportation systems. It will take a long time to recover from these effects and return people’s lives back to normal. The main objective of this study is to investigate the various factors in health and food access, and their spatial correlation and statistical association with COVID-19 spread. The minor aim is to explore regression models on examining COVID-19 spread with these variables. To address these objectives, we are studying the interrelation of various socio-economic factors that would help all humans to better prepare for the next pandemic. One of these critical factors is food access and food distribution as it could be high-risk population density places that are spreading the virus infections. More variables, such as income and people density, would influence the pandemic spread. In this study, we produced the spatial extent of COVID-19 cases with food outlets by using the spatial analysis method of geographic information systems. The methodology consisted of clustering techniques and overlaying the spatial extent mapping of the clusters of food outlets and the infected cases. Post-mapping, we analyzed these clusters’ proximity for any spatial variability, correlations between them, and their causal relationships. The quantitative analyses of the health issues and food access areas against COVID-19 infections and deaths were performed using machine learning regression techniques to understand the multi-variate factors. The results indicate a correlation between the dependent variables and independent variables with a Pearson correlation R2-score = 0.44% for COVID-19 cases and R2 = 60% for COVID-19 deaths. The regression model with an R2-score of 0.60 would be useful to show the goodness of fit for COVID-19 deaths and the health issues and food access factors.

Compact Quad-Band CP Series-Fed Circular Slit Microstrip Array Antenna Using Machine Learning

This article presents the design and characterization of a dual-band, compact, and dual-circularly polarized (CP) traveling wave series-fed circular slit microstrip array (TWSCSMA) antenna operating in quad-band CP at 2.5/3.5/3.8 GHz WiMAX and 5.2 GHz WLAN applications. The radiating structure consists of an annular ring traveling wave series-fed circular slit array elements, rotates in sequence with seven circular quasi-lumped resonators contributes to two-impedance bandwidth (IBW) and quad CP bands. The four split-ring resonators with double-annular rings are introduced on the defected ground plane to improve the impedance matching, IBW, and axial ratio bandwidth (ARBW) effectively. The measured results show that the proposed dual-CP TWSCSMA antenna yields two <i>S</i>₁₁ &#x2264; &#x2013;10 dB IBW of 33.85% and 50.73% and four &#x003C; 3 dB ARBW of 16.26%, 12.52%, 8.22%, and 13.82% for port 1. Furthermore, modern machine learning regression techniques are used to improve the accuracy of the model performance requirements for the optimal design of the proposed antenna. The TWSCSMA antenna demonstrates a compact, low profile, wide impedance bandwidth, and minimum axial ratio which is making it a good candidate for WiMAX and WLAN applications.

IoT based monitoring of air quality and traffic using regression analysis

Dynamic traffic management (DTM) systems are used to reduce the negative externalities of traffic congestion, such as air pollution in urban areas. They require traffic and environmental monitoring infrastructures. In this paper we present a prototype of a low-cost Internet of Things (IoT) system for monitoring traffic flow and the Air Quality Index (AQI). The computation of the traffic flows is based on processing video in the compressed domain. Only using motion vectors as input, traffic flow is computed in real-time over an embedded architecture. An estimation of the AQI is supported by machine learning regression techniques, using different feature data obtained from the IoT device. These automatic learning techniques overcome the need for complex calibration and other limitations of embedded devices in making the needed measurements of the pollutant gases for the computation of the actual AQI. The experimentation with the data obtained from different cities representing different scenarios with a variety of climate and traffic conditions, allows validating the proposed architecture. As regressors, Linear Regression (LR), Gaussian Process Regression (GPR) and Random Forest (RF) are compared using the performance metrics R2, MSE, MAE and MRE resulting in a relevant improvement of the AQI estimations of our proposal.

Risk Factor Prediction by Naive Bayes Classifier, Logistic Regression Models, Various Classification and Regression Machine Learning Techniques

Risk Factor Prediction by Naive Bayes Classifier, Logistic Regression Models, Various Classification and Regression Machine Learning Techniques

Geospatial and Machine Learning Regression Techniques for Analyzing Food Access Impact on Health Issues in Sustainable Communities

Food access is a major key component in food security, as it is every individual’s right to proper access to a nutritious and affordable food supply. Low access to healthy food sources influences people’s diet and activity habits. Guilford County in North Carolina has a high ranking in low food security and a high rate of health issues such as high blood pressure, high cholesterol, and obesity. Therefore, the primary objective of this study was to investigate the geospatial correlation between health issues and food access areas. The secondary objective was to quantitatively compare food access areas and heath issues’ descriptive statistics. The tertiary objective was to compare several machine learning techniques and find the best model that fit health issues against various food access variables with the highest performance accuracy. In this study, we adopted a food-access perspective to show that communities that have residents who have equitable access to healthy food options are typically less vulnerable to health-related disasters. We propose a methodology to help policymakers lower the number of health issues in Guilford County by analyzing such issues via correlation with respect to food access. Specifically, we conducted a geographic information system mapping methodology to examine how access to healthy food options influenced health and mortality outcomes in one of the largest counties in the state of North Carolina. We created geospatial maps representing food deserts—areas with scarce access to nutritious food; food swamps—areas with more availability of unhealthy food options compared to healthy food options; and food oases—areas with a relatively higher availability of healthy food options than unhealthy options. Our results presented a positive correlation coefficient of R2 = 0.819 among obesity and the independent variables of transportation access, and population. The correlation coefficient matrix analysis helped to identify a strong negative correlation between obesity and median income. Overall, this study offers valuable insights that can help health authorities develop preemptive preparedness for healthcare disasters.

Fintech Predictive Modeling and Performance of Investment Firms in Kenya

Predictive analytics is concerned with the prediction of future trends and outcomes. The approaches used to conduct predictive analytics can be classified into machine learning techniques and regression techniques. This study dteremined the influence of fintech predictive modeling on performance of investment firms in Kenya. The study population was 57 investment firms. The study employed mixed method research design by incorporating descriptive and explanatory research designs. Data was collected using questionnaires and an in-depth interview guide. Coefficient of fintech predictive modeling has a positive and significant effect on performance of investment firms. The study concluded that fintech predictive modeling allows investment firms to forecast business growth and customer behaviour chnages. It is important for an investment firm to be able to understand business growth by accurately forecasting future growth and survival. Moreover, it is of vital necessity to understand changes in customer buying/consumption behavior so as to develop products and services that suit their needs and preferences. As a result, predictive modeling is required to project future business growth and changes in customer consumption pattern.

S-frame discrepancy correction models for data-informed Reynolds stress closure

Despite their well-known limitations, Reynolds averaged Navier-Stokes (RANS) models remain the most commonly employed tool for modeling turbulent flows in engineering practice. RANS models are predicated on the solution of the RANS equations, but the RANS equations involve an unclosed term, the Reynolds stress tensor, which must be modeled. The Reynolds stress tensor is often modeled as an algebraic function of mean flow field variables and turbulence variables. This, however, introduces a discrepancy between the Reynolds stress tensor predicted by the Reynolds stress model and the exact Reynolds stress tensor. This discrepancy can result in inaccurate mean flow field predictions for complex flows of industrial relevance. In this paper, we introduce a data-informed approach for arriving at Reynolds stress models with improved predictive performance. Our approach relies on learning the components of the Reynolds stress discrepancy tensor associated with a given Reynolds stress model in the mean strain-rate tensor eigenframe. These components are typically smooth and hence simple to learn using state-of-the-art machine learning strategies and regression techniques. Our approach automatically yields Reynolds stress models that are symmetric, and it yields Reynolds stress models that are both Galilean and frame invariant provided the inputs are themselves Galilean and frame invariant. To arrive at computable models of the discrepancy tensor, we employ feed-forward neural networks and an input space spanning the integrity basis of the mean strain-rate tensor, the mean rotation-rate tensor, the mean pressure gradient, and the turbulent kinetic energy gradient, and we introduce a framework for dimensional reduction of the input space to further reduce computational cost. Numerical results illustrate the effectiveness of the proposed approach for data-informed Reynolds stress closure for a suite of turbulent flow problems of increasing complexity.