Trained Neural Network Model Research Articles

Structural damage detection based on transmissibility functions with unsupervised domain adaptation

Deep learning (DL) has been used for structural damage detection by training neural network models with a large amount of data. These trained models perform well when the test samples are from the data with an identical distribution of the training data. The distributions are always different due to numerical modelling errors and operational environmental varieties, and the data for damage scenarios is difficult to be obtained in practice. A novel method based on the joint maximum discrepancy and adversarial discriminative domain adaptation (JMDAD) for structural damage detection without labelled measurement data has been developed to address the above issues. To reduce the influence of external excitations in practice, the transmissibility function of measured structural responses is used for structural damage detection. The proposed network includes a feature generator, two classifiers and one discriminator. Firstly, the feature generator and domain discriminator are to extract features and merge their distributions at the domain level to overcome the issue of insufficient data from the target real structure. Secondly, the generator and two classifiers are optimised to align their distributions at the class level using the classification discrepancy between the two classifiers. As a result, the damage-sensitive features are extracted and aligned to eliminate modelling errors and operational environmental varieties between the source and target structures. Three case studies have been conducted in this paper, e.g., one case between two building structures with different storeys, another case between the numerical and experimental structures subjected to random and earthquake excitations and the application of Canton Tower. The results show that the proposed method is much robust to the measurement noise and operational environment and the structural damage is identified accurately without labelled measurement data from the target structure in operational environments.

IMPLEMENTASI MACHINE LEARNING ALGORITMA NEURAL NETWORK DALAM MEMPREDIKSI HARGA SAHAM MENGGUNAKAN RAPIDMINER

The research is aimed at implementing the Neural Network algorithm in predicting stock prices using RapidMiner. Historical stock price data is analyzed to train neural network models in identifying patterns and trends that can be used to predict future stock prices. The research phases include data collection, data preprocessing, Neural Network model development, model training, and model performance evaluation. Data used include daily closing prices, trade volumes, and other technical indicators. The data is processed through several stages of preprocesing, including normalization and filling of missing values, before being used to train Neural network models. The model was built with several layers and the number of neurons adjusted based on early experiments to find the optimal configuration. Models are trained using datasets that are divided into training sets and testing sets. The predicted result is compared to the actual value to evaluate the performance of the model. The results showed that the Neural Network model was able to predict stock prices with a fairly high accuracy rate, with an RMSE of 0.989 indicating a low prediction error rate. The research concludes that the Neural Network algorithm implemented in RapidMiner is effective in predicting stock prices and has significant potential to be used in investment decision-making. The use of RapidMiner as a data processing tool has proven to be efficient and facilitating the development and evaluation of predictive models. Keywords: Neural Network, Stock Price Prediction, RapidMiner, Machine Learning, Investment

Automated detection of Bornean white-bearded gibbon (Hylobates albibarbis) vocalizations using an open-source framework for deep learning.

Passive acoustic monitoring is a promising tool for monitoring at-risk populations of vocal species, yet, extracting relevant information from large acoustic datasets can be time-consuming, creating a bottleneck at the point of analysis. To address this, an open-source framework for deep learning in bioacoustics to automatically detect Bornean white-bearded gibbon (Hylobates albibarbis) "great call" vocalizations in a long-term acoustic dataset from a rainforest location in Borneo is adapted. The steps involved in developing this solution are described, including collecting audio recordings, developing training and testing datasets, training neural network models, and evaluating model performance. The best model performed at a satisfactory level (F score = 0.87), identifying 98% of the highest-quality calls from 90 h of manually annotated audio recordings and greatly reduced analysis times when compared to a human observer. No significant difference was found in the temporal distribution of great call detections between the manual annotations and the model's output. Future work should seek to apply this model to long-term acoustic datasets to understand spatiotemporal variations in H. albibarbis' calling activity. Overall, a roadmap is presented for applying deep learning to identify the vocalizations of species of interest, which can be adapted for monitoring other endangered vocalizing species.

Feed-forward State of Charge estimation of LiFePO[formula omitted] batteries using time-series machine learning prediction with autoregressive models

This paper presents a Recurrent Neural Network (RNN) model for the State of Charge (SoC) estimation for an Electric Vehicle’s Lithium-Ion battery. A new Autoregressive modification is proposed to improve upon earlier published Machine Learning models to enhance the prediction accuracy and make results less sensitive to varying usage conditions. By building upon earlier published findings, assessing several models’ ability to estimate charge using a history window of sensory data, this work proposes, implements, and reports its methods of improving accuracy by introducing the State of Charge as part of sensory inputs in a Feed-Forward manner. The method is based on the Autoregressive implementation of neural network model training, where successively calculated charge values are used as input in the subsequent predictions, but the potential inaccuracies with those predictions that might otherwise cause significant and rapid divergence are accounted for within a modified recursive training procedure. As a result, with a slight increase in training time, the accuracy of the output prediction for three different realistic charge–discharge driving profiles doubled compared to the two best previously published models. Overall, the test results have demonstrated the usability of the current model in actual driving scenarios, making such models a viable replacement to estimation approaches used in electric vehicle battery management systems.

Integrated deep learning model for automatic detection and classification of stenosis in coronary angiography

Coronary artery disease poses a significant threat to human health. In clinical settings, coronary angiography remains the gold standard for diagnosing coronary heart disease. A crucial aspect of this diagnosis involves detecting arterial narrowings. Categorizing these narrowings can provide insight into whether patients should receive vascular revascularization treatment. The majority of current deep learning methods for analyzing coronary angiography are mostly confined to the theoretical research domain, with limited studies offering direct practical support to clinical practitioners. This paper proposes an integrated deep-learning model for the localization and classification of narrowings in coronary angiography images. The experimentation employed 1606 coronary angiography images obtained from 132 patients, resulting in an accuracy of 88.9 %, a recall rate of 85.4 %, an F1 score of 0.871, and a MAP value of 0.875 for vascular stenosis detection. Furthermore, we developed the "Hemadostenosis" web platform (http://bioinfor.imu.edu.cn/hemadostenosis) using Django, a highly mature HTTP framework. Users are able to submit coronary angiography image data for assessment via a visual interface. Subsequently, the system sends the images to a trained convolutional neural network model to localize and categorize the narrowings. Finally, the visualized outcomes are displayed to users and are downloadable. Our proposed approach pioneers the recognition and categorization of arterial narrowings in vascular angiography, offering practical support to clinical practitioners in their learning and diagnostic processes.

CRSFL: Cluster-based Resource-aware Split Federated Learning for Continuous Authentication

In the ever-changing world of technology, continuous authentication and comprehensive access management are essential during user interactions with a device. Split Learning (SL) and Federated Learning (FL) have recently emerged as promising technologies for training a decentralized Machine Learning (ML) model. With the increasing use of smartphones and Internet of Things (IoT) devices, these distributed technologies enable users with limited resources to complete neural network model training with server assistance and collaboratively combine knowledge between different nodes. In this study, we propose combining these technologies to address the continuous authentication challenge while protecting user privacy and limiting device resource usage. However, the model’s training is slowed due to SL sequential training and resource differences between IoT devices with different specifications. Therefore, we use a cluster-based approach to group devices with similar capabilities to mitigate the impact of slow devices while filtering out the devices incapable of training the model. In addition, we address the efficiency and robustness of training ML models by using SL and FL techniques to train the clients simultaneously while analyzing the overhead burden of the process. Following clustering, we select the best set of clients to participate in training through a Genetic Algorithm (GA) optimized on a carefully designed list of objectives. The performance of our proposed framework is compared to baseline methods, and the advantages are demonstrated using a real-life UMDAA-02-FD face detection dataset. The results show that CRSFL, our proposed approach, maintains high accuracy and reduces the overhead burden in continuous authentication scenarios while preserving user privacy.

Research on thermal runaway early warning algorithm of lithium battery based on improved particle swarm algorithm optimized BP neural network

Abstract In response to the issues of false alarms and missed alarms in traditional fire alarm systems for energy storage station fire prevention and control, this study proposes a fire alarm approach employing improved PSO-BP neural networks. Firstly, we set adaptive inertial weights and asymmetric learning factors on the PSO algorithm that are automatically optimized as fitness changes to improve the optimization accuracy and convergence speed of the algorithm. Secondly, the thresholds or weights are mapped to random particles for training to obtain the optimal value. Finally, utilizing the Pyrosim software, a model of the battery cabinet was established, and the thermal runaway data samples were input into the neural network model for training. The test consequence indicates that the improved PSO-BP model achieved a 5% and 3.75% rise in precision relative to the neural network and PSO-BP model, respectively, while the frequency of iterations declined by 55% relative to the PSO-BP.

Assessment of reinforced concrete corrosion degree based on the quantum particle swarm optimised-generative adversarial network

Reinforced concrete corrosion inspection methods based on deep learning have been widely used in the engineering field to monitor the service status of reinforced concrete. However, in engineering practice, it is difficult to obtain a large amount of reinforced concrete corrosion data of different types, which greatly hinders the improvement of the accuracy of neural network models in predicting corrosion conditions. The classic generative adversarial network (GAN) model gives poor model quality for datasets with small amounts of data and high concentration. This paper proposes an improved generative adversarial network approach to optimise reinforced concrete corrosion data. Firstly, a quantum particle swarm optimisation (QPSO) algorithm is used to improve the generative adversarial network. Then, existing corrosion characteristic data is used to train the improved generative adversarial network until the ideal equilibrium state is reached. Next, the feature data generated by the generator are fused with the original data and the fused data are input into several common machine learning models for training. Experimental results show that compared with other conventional results obtained by directly inputting corrosion data into a neural network model for training, the improved method makes full use of multi-source signal data and achieves better classification performance.

Thickness Characterization of Steel Plate Coating Materials with Terahertz Time-Domain Reflection Spectroscopy Based on BP Neural Network.

Accurate monitoring of steel plate coating thickness is crucial in construction quality control and durability assessments. To address this challenge, this study introduces a terahertz time-domain reflection spectroscopy based on a BP neural network model to achieve a quantitative visualization characterization of coating thickness. The BP neural network eliminates the inherent dependence of terahertz reflection spectroscopy on the refractive index value in thickness calculation. This trained BP neural network model effectively establishes a functional relationship between signal feature parameters and the corresponding thickness values. Additionally, the proposed model can innovatively measure different coating materials' refractive indexes, revealing the corresponding values for the black paint, white paint, epoxy resin, and rubber as 2.212, 1.967, 1.924, and 2.185, respectively. The experimental results demonstrate the trained BP neural network model possesses remarkable accuracy in predicting coating thickness within the scanning area, achieving a precision level exceeding 96%. This method enables the visualization of coating thickness and the extraction of thickness characterization values. Furthermore, using the thickness imaging results as a reference, the method can accurately identify the thickness abnormalities across the scanning area, locating the position and size of potential defects such as internal scratches and foreign object defects. This innovative approach offers a superior means of monitoring and assessing the thickness distribution quality of the steel plate coating layer materials.

A robust synthetic data generation framework for machine learning in high-resolution transmission electron microscopy (HRTEM)

Machine learning techniques are attractive options for developing highly-accurate analysis tools for nanomaterials characterization, including high-resolution transmission electron microscopy (HRTEM). However, successfully implementing such machine learning tools can be difficult due to the challenges in procuring sufficiently large, high-quality training datasets from experiments. In this work, we introduce Construction Zone, a Python package for rapid generation of complex nanoscale atomic structures which enables fast, systematic sampling of realistic nanomaterial structures and can be used as a random structure generator for large, diverse synthetic datasets. Using Construction Zone, we develop an end-to-end machine learning workflow for training neural network models to analyze experimental atomic resolution HRTEM images on the task of nanoparticle image segmentation purely with simulated databases. Further, we study the data curation process to understand how various aspects of the curated simulated data—including simulation fidelity, the distribution of atomic structures, and the distribution of imaging conditions—affect model performance across three benchmark experimental HRTEM image datasets. Using our workflow, we are able to achieve state-of-the-art segmentation performance on these experimental benchmarks and, further, we discuss robust strategies for consistently achieving high performance with machine learning in experimental settings using purely synthetic data. Construction Zone and its documentation are available at https://github.com/lerandc/construction_zone.

Parameterized physics-informed neural networks (P-PINNs) solution of uniform flow over an arbitrarily spinning spherical particle

Neural network-based approaches have emerged as alternatives to conventional computational fluid dynamics (CFD) in solving multiphase flow problems. However, most of these approaches are based on data-driven machine learning methods that require training datasets prepared beforehand through CFD simulations or experiments and cannot be used independently. This study presents an unsupervised parameterized physics-informed neural networks (P-PINNs) approach for obtaining the full flow field information over a multi-dimensional parametric space. This approach is then applied to solve the three-dimensional flow around an arbitrarily rotating sphere subjected to a cross-flow. This solution is obtained for a continuous range of three parameters: the Reynolds number based on cross-flow (Re∈[1,400]), non-dimensional streamwise and spanwise angular velocity components (Ωx∗,Ωy∗∈[−3,3]). The predictions from the P-PINNs model are compared against particle-resolved direct numerical simulation (PR-DNS) results, demonstrating that the neural network model predicts all flow variables (velocity, pressure, and their spatial derivatives) accurately with R2≳0.9. The force and torque on the particle obtained from the P-PINNs model are also compared well against the corresponding PR-DNS results with R2>0.94. The key observation is that the computational cost of the unsupervised parameterized neural network model training and deployment is substantially lower than performing numerous conventional CFD simulations to systematically vary the parameters. Furthermore, once trained, the resultant neural network model is substantially more compact than the huge database of discretized numerical solutions. The P-PINNs model is then used to reveal several interesting flow features and phenomena about the rotating sphere, including flow symmetry, secondary drag and lift, critical Reynolds number, flow bifurcation, and flow separation. This study presents P-PINNs as an efficient alternative for solving parameterized flow problems, demonstrating its practical usage in flow physics discovery with the example of flow past an arbitrarily rotating sphere.

The Perspective of Using Neural Networks and Machine Learning Algorithms for Modelling and Forecasting the Quality Parameters of Coking Coal—A Case Study

The quality of coking coal is vital in steelmaking, impacting final product quality and process efficiency. Conventional forecasting methods often rely on empirical models and expert judgment, which may lack accuracy and scalability. Previous research has explored various methods for forecasting coking coal quality parameters, yet these conventional methods frequently fall short in terms of accuracy and adaptability to different mining conditions. Existing forecasting techniques for coking coal quality are limited in their precision and scalability, necessitating the development of more accurate and efficient methods. This study aims to enhance the accuracy and efficiency of forecasting coking coal quality parameters by employing neural networks and artificial intelligence algorithms, specifically in the context of Knurow and Szczyglowice mines. The research involves gathering historical data on various coking coal quality parameters, including a proximate and ultimate analysis, to train and test neural network models using the Group Method of Data Handling (GMDH). Real-world data from Knurow and Szczyglowice mines’ coal production facilities form the basis of this case study. The integration of neural networks and artificial intelligence techniques significantly improves the accuracy of predicting key quality parameters such as ash content, sulfur content, volatile matter, and calorific value. This study also examines the impact of these quality indicators on operational costs and highlights the importance of final indicators like the Coke Reactivity Index (CRI) and Coke Strength after Reaction (CSR) in expanding industrial reserve concepts. Model performance is evaluated using metrics such as mean absolute error (MAE), root mean square error (RMSE), and coefficient of determination (R2). The findings demonstrate the effectiveness of these advanced techniques in enhancing predictive modeling in the mining industry, optimizing production processes, and improving overall operational efficiency. Additionally, this research offers insights into the practical implementation of advanced analytics tools for predictive maintenance and decision-making support within the mining sector.

Quantitative assessment of the contribution of risk factors to the formation of nutrition-dependent diseases based on neural network modelling in schoolchildren

Introduction. The increased interest on the part of the state in the problem of healthy nutrition makes it urgent to develop methodological approaches to quantify the likelihood of the occurrence of nutrition-related diseases and assess health risks. Materials and methods. The source of information was data obtained from a sociological study conducted by the Federal Service for Supervision in Protection of the Rights of Consumer and Man Wellbeing (Rospotrebnadzor). For the analysis, there were used eleven thousand five hundred forty three questionnaires, characterizing the nutritional factors of schoolchildren in the Russian Federation in the regions. Associated relationships between the studied factors and morbidity were obtained through neural network modelling. The study of the contributions of factors to the formation of diseases was carried out based on the analysis of a simulation matrix containing 300 thousand possible nutrition scenarios. Results. In the course of mathematical modelling, there were delivered 12 neural network models that describe the dependence of the additional probability of developing nutrition-dependent diseases on nutritional factors, characterized by a high proportion of correct predictions (more than 70%). The contributions of factors to changes in the probability of developing obesity were analyzed as a model with a high degree of reliability of parameters. The factors most influencing the development of obesity have been established to include daily consumption of foods with a high glycemic index, high-calorie confectionery in the form of cakes and pastries, sweet juices, consumption of products from a vending machine, excluding the purchase of salads and vegetable dishes. This type of diet increases the likelihood of developing obesity from the original 0.033 to 0.98 for the average schoolchild. Limitations. The study is limited to a set of factors measured as a result of a sociological survey and used in training neural network models. Conclusion. The neural network models obtained as a result of the study and the information materials created on their basis made it possible to develop tools that make it possible to quickly create arbitrary nutrition scenarios for schoolchildren and calculate the additional probability of the formation of nutrition-related diseases.

Enhancing Robotic Grasping Performance through Data-Driven Analysis

In automation, reliability in robotic grasping in dynamic environment is still a problem encountered. Further, there is the need to consider deep learning methods, as traditional approaches are not easily flexible in dealing with different objects and situations. In this work, we aim to analyze how well deep neural network models perform in predicting grasp strength based on data collected from the Smart Grasping Sandbox simulation trials. Hence, the proposed approach for analyzing the joint positions, velocities and efforts led to the design of a deep neural network for improving robot grasp performance. The question here could is if deep learning could help better predict grasp stability. To implement the method, we underwent rigorous data pre-processing such as outlier detection, and feature normalization and employed a structured neural network model for training. Our model got a training accuracy of nearly 99% and the test accuracy of nearly 96% demonstrating significant promise. These results were also better than CNN and LSTM where their accuracy rate was 94.12% and 91.81%, respectively. High deep learning performance was proven in robotic grasping, which can contribute to the creation of more flexible and sophisticated robotic platforms. This work also provides the foundation for subsequent research in robotics and automation, with emphasis on the role of data-driven methods in robotic grasping tasks.

Neural network architecture search model for thermal radiation in dense particulate systems

In the CFD-DEM simulation of the many dense particulate systems, particle-scale thermal radiation is an important heat transfer mode under high temperatures. In this work, neural network architecture search model with different layer connection matrix is applied to the view factor regression of the thermal radiation in dense particulate systems. Deep neural networks with 3346 feasible architectures are evaluated by the HyperBand algorithm to find the local optimal solution. Neural architecture search model trained by the big data of the view factor gives a good prediction of the macroscopic radiative properties and it is in general agreement with the empirical correlations and experimental data. The particle–wall radiation decreases strongly with the distance and the maximum interaction depth is about 2.0 times the sphere diameter. The trained deep neural network model provides an efficient data-driven closure to discuss the thermal radiation of the particle–particle and particle–wall interactions in particle bed.

A study of pipeline parallelism in deep neural networks

The current popularity in the application of artificial intelligence to solve complex problems is growing. The appearance of chats based on artificial intelligence or natural language processing has generated the creation of increasingly large and sophisticated neural network models, which are the basis of current developments in artificial intelligence. These neural networks can be composed of billions of parameters and their training is not feasible without the application of approaches based on parallelism. This paper focuses on studying pipeline parallelism, which is one of the most important types of parallelism used to train neural network models in deep learning. In this study we offer a look at the most important concepts related to the topic and we present a detailed analysis of 3 pipeline parallelism libraries: Torchgpipe, FairScale, and DeepSpeed. We analyze important aspects of these libraries such as their implementation and features. In addition, we evaluated them experimentally, carrying out parallel trainings and taking into account aspects such as the number of stages in the training pipeline and the type of balance.

A dataset of drilling site object detection in underground coal mines

Drilling in underground coal mine is an important measure for dealing with gas, water and hidden geological disasters, which can significantly enhance the effectiveness of disaster prevention and control in coal mining operations. In order to monitor the drilling process in real time and improve drilling efficiency, it is necessary to carry out object detection to identify and locate key targets at the drilling site. Compared with traditional object detection method, deep learning-based object detection method can improve accuracy, timeliness and stability of object detection, but it requires high-quality datasets to perform well. At present, research on object detection in underground coal mine drilling sites mainly relies on small-scale private datasets, which are insufficient for providing necessary or reliable data for deep neural network model training. In this study, we constructed a dataset of drilling site object detection using photos taken by intrinsic safety law enforcement recorders. This dataset is developed through several steps, including data cleaning, data labeling, and expert sampling verification. The mainstream YOLO series object detection model is used for data quality assessment. This dataset comprises 70,948 images from drilling sites under different environmental conditions, covering five categories of objects: gripper, chuck, coal miner, mine safety helmet, and drill pipe. It provides annotated files in PASCAL VOC format. This dataset can provide strong data support for object detection research in underground coal mine drilling sites, and plays an important role in promoting intelligent underground coal mine monitoring and early warning.

Neural Net-Enhanced Competitive Swarm Optimizer for Large-Scale Multiobjective Optimization.

The competitive swarm optimizer (CSO) classifies swarm particles into loser and winner particles and then uses the winner particles to efficiently guide the search of the loser particles. This approach has very promising performance in solving large-scale multiobjective optimization problems (LMOPs). However, most studies of CSOs ignore the evolution of the winner particles, although their quality is very important for the final optimization performance. Aiming to fill this research gap, this article proposes a new neural net-enhanced CSO for solving LMOPs, called NN-CSO, which not only guides the loser particles via the original CSO strategy, but also applies our trained neural network (NN) model to evolve winner particles. First, the swarm particles are classified into winner and loser particles by the pairwise competition. Then, the loser particles and winner particles are, respectively, treated as the input and desired output to train the NN model, which tries to learn promising evolutionary dynamics by driving the loser particles toward the winners. Finally, when model training is complete, the winner particles are evolved by the well-trained NN model, while the loser particles are still guided by the winner particles to maintain the search pattern of CSOs. To evaluate the performance of our designed NN-CSO, several LMOPs with up to ten objectives and 1000 decision variables are adopted, and the experimental results show that our designed NN model can significantly improve the performance of CSOs and shows some advantages over several state-of-the-art large-scale multiobjective evolutionary algorithms as well as over model-based evolutionary algorithms.

Hot rolled steel surface defect detection and classification using an automatic ensemble approach

This study introduces an ensemble-based Deep Neural Network (DNN) model for detecting defects on steel surfaces. The method suggested in this study classifies steel surface conditions into six possible fault categories, namely, crazing, inclusion, rolled in, pitted surface, scratches, and patches. The images undergo preprocessing and extraction of features in spatial and frequency domains using image segmentation techniques such as grey level difference method (GLDM), fast Fourier Transform (FFT), grey level co-occurrence matrix (GLCM), texture analysis and discrete wavelet transform (DWT). The ensembling of image features into a fused feature pool is carried out after the preprocessing of input images that are provided as input to a light-weight neural network model for training and testing. The performance of the model is comprehensively evaluated via an ablation study both before and after ensembling. In addition, the model capability is effectively analyzed using receiver operating characteristics (ROC) curve, confusion matrix from which classification accuracy of the model could be obtained and other parameters including precision and f1-score. It was observed that the proposed deep learning network presents phenomenally high accuracy of 99.72% for detection and classification of steel surface faults. This result was found to be superior when compared with the performance of the same neural network over each feature type individually. This study also compares the classification results of the model built based on the ensembled feature set with the results of various other classification approaches available in literature. The ensemble-based model could potentially be integrated into existing inspection systems for real-time, efficient and robust condition monitoring of steel surfaces.

Improving GOCI ocean color data under high solar-zenith angle over open oceans using neural networks

ABSTRACT With hourly measurements available during daytime between local times of 09:00–16:00, ocean color data derived from the Geostationary Ocean Color Imager (GOCI) onboard the Korean Communication, Ocean, and Meteorological (COMS) satellite have been useful for research and surveillance of diurnal processes in the western Pacific Ocean region. However, in early morning and late afternoon measurements, there are significant errors in GOCI-derived ocean color products as the solar-zenith angle (θ 0 ) goes beyond 70°, especially in autumn and winter seasons. In this study, we employ a neural network (NN) model to make corrections on the GOCI-measured normalize water-leaving radiance spectra, nL w (λ), with high θ 0 (>70°) in open oceans. Results show that NN-corrected nL w (λ) are consistent with the previous-hour nL w (λ) and make the diurnal variations in the region much more stable and reasonable. Specifically, the GOCI-measured nL w (λ) with θ 0 ≤65° in earlier hours of the day (including some nL w (λ) diurnal variation) are considered accurate and used as the ground truth to train NN models for nL w (λ) correction with high θ 0 . Further analysis of the relationship of ratios (between the NN-corrected and original nL w (λ)) with θ 0 shows that the nL w (λ) ratios increase as θ 0 increase, which indicates that there are more significant corrections with larger θ 0 (>70°). The performance evaluation of the NN models is based on the comparison of NN-corrected nL w (λ) with the original previous hour nL w (λ) data. The ratios of NN-corrected nL w (λ) to the original previous hour nL w (λ) are 0.968–1.045 for the short blue/blue and green bands, and the performance of nL w (λ) correction at 14:00 and 15:00 is slightly better than that at 16:00, due to significantly large θ 0 at late afternoon hours. The NN-corrected nL w (λ) data are also used to derive chlorophyll-a (Chl-a) concentration, showing significantly improved Chl-a in GOCI’s late afternoon measurements with θ 0 >70°.