Search

• Enhanced cellulo-xylanolytic consortium production from Hypocrea lixii GGRK4 using MOGA-ANN is reported. • The production of CMCase (9.43 fold) and xylanase (4.40 fold) higher than un-optimized process. • The improved deinking efficiency and brightness is reported for photocopier paper and newspaper. • The physical strength of the waste papers were enhanced whereas double fold property was decreased proving its reusability. • A significant fungal enzyme consortium preparation for improved waste paper deinking achieved. This study reports the cellulo-xylanolytic cocktail production from Hypocrea lixii GGRK4 using multi-objective genetic algorithm-artificial neural network tool, resulting in 8.32 ± 1.07 IU/mL, 51.53 ± 3.78 IU/mL activity of CMCase and xylanase, respectively with more than 85 % residual activity at 60 °C and pH 6.0. Interestingly, metal ions viz. K+ and Ca2+ stimulated the enzyme activity, whereas Fe2+ and Cu2+ reduced the activity. Significant amounts of hydrophobic compounds, chromophores, and phenolics were released after wastepapers deinking. The deinking efficiency of 73.60 ± 2.45 % and 38.60 ± 1.34 % was obtained for photocopier paper and newspaper, respectively, whereas brightness of 89.90 ± 2.10 % ISO and 44.90 ± 1.63 % ISO was reported for both types of waste papers. The physical strength of deinked photocopier paper and newspapers, i.e., tensile index (3.10 and 0.50 %), tearing index (7.10 and 4.83 %), and burst factor (8.61) were enhanced whereas double fold property was decreased proving wastepaper reusability. This consortium showed effective and significant enzymatic deinking efficiency for recycled wastepapers. [ABSTRACT FROM AUTHOR]

With the development of computer technology, automatic test paper generation systems have gradually become an effective tool for detecting and maintaining national machine security and protecting the rights and interests of workers. This article achieved multi-level oral scores for different types of questions through online scoring using artificial neural networks in recent years. Based on its specific situation and evaluation index requirements, an analysis module that is reasonable, efficient, and in line with the hierarchical structure and module requirements of national conditions has been designed to complete the research on automatic test paper generation technology, in order to help better manage and allocate human resources and improve production efficiency. Afterwards, this article conducted functional testing on the technical module. The test results showed that the scalability of the system was over 82%. [ABSTRACT FROM AUTHOR]

Paper mills are among the most polluting industries, responsible for many organic and inorganic compounds emissions. The fibres electro-kinetic features strongly affect the ability to retain fillers since the fillers–fibres interactions are charge induced. The control and the prediction of these parameters would represent a precious aid for process management, allowing the fillers retention enhancement, a lower environmental impact and the paper sheet properties streamlining. The work presented deals with the implementation and training of four artificial neural networks (ANNs) for the prediction of the main electrochemical and physical features of cellulose pulp and paper. First, two ANNs predict the electrochemical parameters. Following, they were applied to predict the paper sheet properties and fillers retention. The neural models implemented showed outstanding prediction performance, with R2 in the order of 0.999 and a low mean error. The results demonstrate how Artificial Neural Networks may be a valuable instrument for paper mill pollutant reduction. However, they suggest a more inclusive investigation for a better fibres behaviour representation. • Main process parameters of an industrial papermaking process were identified. • Experimental datasets were achieved during industrial production. • Artificial Neural Networks were trained for process parameters prediction. • Accurate predictions of papermaking process were obtained. [ABSTRACT FROM AUTHOR]

• Five Machine Learning (ML) models compared for wave height time series prediction. • Complex ML models do not outperform simple AR in wave height time series prediction. • Comment to related papers: signal decomposition in test set series is WRONG. Significant Wave Height (SWH) is crucial in many aspect of ocean engineering. The accurate prediction of SWH has therefore been of immense practical value. Recently, Artificial Intelligence (AI) time series prediction methods have been widely used for single-point short-term SWH time-series forecasting, resulting in many AI-based models claiming to achieve good results. However, the extent to which these complex AI models can outperform traditional methods has largely been overlooked. This study compared five different models - AutoRegressive (AR), eXtreme Gradient Boosting (XGB), Artificial Neural Network (ANN), Long Short-Term Memory (LSTM), and WaveNet - for their performance on SWH time series prediction at 16 buoy locations. Surprisingly, the results suggest that the differences of performance among different models are negligible, indicating that all these AI models have only "learned" the linear auto-regression from the data. Additionally, we noticed that many recent studies used signal decomposition method for such time series prediction, and most of them decomposed the test sets, which is WRONG. [ABSTRACT FROM AUTHOR]

An instrument-free, user-friendly, and cost-effective PAD colorimetric sensor that can quantitative detection of H 2 O 2 has been realized by using bimetallic nickel–cobalt selenides as highly active peroxidase mimic and smartphone integrated with dark cavity as detector. Integrated with an ANN model and a self-compiled easy-to-use smartphone APP, the intelligent and on-site detection of H 2 O 2 was constructed, which exhibited an ultra-wide dynamic range. [Display omitted] • An effective paper-based colorimetric sensor was prepared based on nanozyme. • The optimized Ni 0.75 Co 0.25 Se exhibited excellent peoxidase-mimetic activities. • The on-site detection of H 2 O 2 was constructed by a self-compiled smartphone APP. • The machine learning-assisted sensor can detect H 2 O 2 over a wide dynamic range. Paper-based analytical devices (PADs) integrated with smartphones have shown great potential in various fields, but they also face challenges such as single signal reading, complex data processing and significant environmental impacting. In this study, a colorimetric PAD platform has been proposed using bimetallic nickel–cobalt selenides as highly active peroxidase mimic, smartphone with 3D-printing dark-cavity as a portable detector and an artificial neural network (ANN) model as multi-signal processing tool. Notably, the optimized nickel–cobalt selenides (Ni 0.75 Co 0.25 Se with Ni to Co ratio of 3/1) exhibit excellent peoxidase-mimetic activities and are capable of catalyzing the oxidation of four chromogenic reagents in the presence of H 2 O 2. Using a smartphone with image capture function as a friendly signal readout tool, the Ni 0.75 Co 0.25 Se based four channel colorimetric sensing paper is used for multi-signal quantitative analysis of H 2 O 2 by determining the Grey, red (R), green (G) and blue (B) channel values of the captured pictures. An intelligent on-site detection method for H 2 O 2 has been constructed by combining an ANN model and a self-programmed easy-to-use smartphone APP with a dynamic range of 5 μM to 2 M. Noteworthy, machine learning-assisted smartphone sensing devices based on nanozyme and 3D printing technology provide new insights and universal strategies for visual ultrasensitive detection in a variety of fields, including environments monitoring, biomedical diagnosis and safety screening. [ABSTRACT FROM AUTHOR]

Abstract: Advanced neuro-fuzzy modeling, namely an adaptive network-based fuzzy inference system (ANFIS), was employed to develop models for the prediction of suspended solids (SS) and chemical oxygen demand (COD) removal of a full-scale wastewater treatment plant treating process wastewaters from a paper mill. In order to improve the network performance, fuzzy subtractive clustering was used to identify model''s architecture and optimize fuzzy rule, meanwhile principal component analysis (PCA) was applied to reduce the input variable dimensionality. Input variables were reduced from six to four for COD and SS models, by considering PCA results and linear correlation matrices among input and output variables. The results indicate that reasonable forecasting and control performances have been achieved through the developed system. The minimum mean absolute percentage errors of 1.003% and 0.5161% for CODeff and SSeff could be achieved using ANFIS. The maximum correlation coefficient values for CODeff and SSeff were 0.9912 and 0.9882, respectively. The minimum mean square errors of 1.2883 and 0.0342, and the minimum RMSEs of 1.135 and 0.1849 for CODeff and SSeff could also be achieved. [Copyright &y& Elsevier]

Electronic nose systems when deployed in network mesh can effectively provide a low budget and onsite solution for the industrial obnoxious gaseous measurement. For accurate and identical prediction capability by all the electronic nose systems, a reliable calibration transfer model needs to be implemented in order to overcome the inherent sensor array variability. In this work, robust regression (RR) is used for calibration transfer between two electronic nose systems using a Box–Behnken (BB) design. Out of the two electronic nose systems, one was trained using industrial gas samples by four artificial neural network models, for the measurement of obnoxious odours emitted from pulp and paper industries. The emissions constitute mainly of hydrogen sulphide (H 2 S), methyl mercaptan (MM), dimethyl sulphide (DMS) and dimethyl disulphide (DMDS) in different proportions. A Box–Behnken design consisting of 27 experiment sets based on synthetic gas combinations of H 2 S, MM, DMS and DMDS, were conducted for calibration transfer between two identical electronic nose systems. Identical sensors on both the systems were mapped and the prediction models developed using ANN were then transferred to the second system using BB–RR methodology. The results showed successful transmission of prediction models developed for one system to other system, with the mean absolute error between the actual and predicted concentration of analytes in mg L −1 after calibration transfer (on second system) being 0.076, 0.1801, 0.0329, 0.427 for DMS, DMDS, MM, H 2 S respectively. [ABSTRACT FROM AUTHOR]

Abstract: In this study, the determinability of time related colour changes in prints made using ink that dries on matte coated paper with the offset printing technique and infrared method, has been investigated with artifical neural networks after analysis of the experimental and numerical methods. In the experimental part a ECI 2002 CMYK test scale is printed on a 135g/m2 matte coated. The prints obtained were measured with a spectrophotometer at certain intervals and by converting the reflection data obtained into the CIELAB colour description spectrum, time related changes in printed colours have been determined. In the empirical study however, colour changes in prints made on matte coated paper using inks that dry with the infrared method are modelled numerically. The average deviation has been found to be 0.4%. As a result, artificial neural networks can be used with the aim to digitally calculate time related colour changes in the offset printing system. [Copyright &y& Elsevier]

Abstract: Neural networks can provide effective predictive models for complex processes that are poorly described by first principle models, such as wastewater biological treatment systems. In this paper multilayer perceptron (MLP) and functional-link neural networks (FLN) are developed to predict inlet and outlet biochemical oxygen demand (BOD) of an aerated lagoon operated by International Paper of Brazil. In Part I, predictive models for both inlet and outlet BOD for the aerated lagoon were developed using linear multivariate regression techniques. For the current case study, MLP networks are the best choice for the prediction models. When only a relatively small number of samples is available, substantial improvement in inlet and outlet BOD prediction is shown for both FLN and MLP modeling using a reduced input variable set that was generated using partial least squares (PLS). Thus, this paper provides a novel approach for developing PLS–FLN model structures. [Copyright &y& Elsevier]

• Adsorptive uptake of fluoride using novel paper industrial waste (PWA) was studied. • Fluoride removal process was optimized via RSM, ANN and GA. • PWA was characterized through SEM, EDX and FTIR. • Maximum removal of 99% was achieved by PWA. • 96.7% of fluoride removal achieved after four cycles of use of PWA. The current investigation emphasizes on the fluoride adsorption by chemically activated industrial paper waste (PWA) from aqueous medium. Sorption process was modelled and optimized using the techniques viz. response surface methodology (RSM), artificial neural network (ANN), and genetic algorithm (GA). Parametric effects of pH, adsorbate concentration, duration of contact, temperature, and PWA dose on percentage removal were assessed. Structural morphologies of PWA were analysed using BET, SEM, EDX and FTIR. Maximum removal of more than 99% was observed at adsorbate concentration of 12 mg/L, pH 5, temperature 25 °C and PWA dose 0.3 g/L and duration of contact of 18 h. The adsorption process conformed Langmuir isotherm model, and the process is thermodynamically feasible and spontaneous in nature as obtained from thermodynamics study. The PWA is reusable and capable of adsorbing 96.7% of total fluoride even after four cycles of use. Therefore, PWA might be a promising adsorbent for fluoride decontamination of water and can be used in sorption-based water treatment plant under optimised parametric condition to substantiate its large-scale applicability. [ABSTRACT FROM AUTHOR]

Abstract: A new approach of paper defect recognition is proposed. To implement this method, computer is used to acquire paper image and the subtraction algorithm is used to determine if the image contains defect at first. For the image with paper defect, wavelet packet decomposition is applied to extracting the feature vector of the defect. A BP neural network is designed for recognizing the defect type according to the feature vector. This approach provides a unified detection algorithm for different types of paper defects. The experiment data prove this method is valid and can be applied in modern paper defect inspection system. [Copyright &y& Elsevier]

Abstract: Aiming at the problem that paper smoothness cannot be measured on line, a new method for soft measurement of paper smoothness is put forward on the basis of studying paper quantization distribution. A soft measurement system is established, by which, paper quantization noise is processed by time–frequency analysis and non-linear transformation, so that online paper smoothness can be obtained through a Radial Basis Function Neural Networks (RBFNNs). The results show that this soft computing method is a feasible way with high measuring accuracy. [Copyright &y& Elsevier]

• Robust functional form contains true parameters far more often than popular models. • Matches/outperforms widely used regression and Neural Network models. • Finds appropriate balance between Model Fit, Inference and Prediction (MIPs). • Introduces new large-sample DGP test; can use to improve A.I. models. • For MIS field finds Popularity Parameter to be important for predicting citations. This article introduces a versatile functional form for Generalized Linear Models (GLMs) through a simple, yet effective, transformation of the current framework. The models are applied through a new hierarchical bayesian estimation procedure for logistic regression to highly-cited papers in the Management Information Systems (MIS) field. The results are uniformly better, in regards to model fit and inference for in-sample and out-of-sample data, for simulation studies and real-world data applications, requiring very little time to convergence to true population parameters. In simulation studies, I show that the method contains the true parameters nearly three times as often as widely used existing GLMs, and does so while having confidence intervals that are 54.50% smaller, while requiring around two-thirds the number of MCMC iterations as existing bayesian methods. In Scientometric applications the methodology is shown to be highly robust with predictive/classification accuracy, either equaling or exceeding existing methods for identifying highly-cited articles including Artificial Neural Networks (ANN). Thus, the method is shown to be robust to the amount of asymmetry (or symmetry) of the probability of success (or failure) and robust to unbalanced samples and varying Data Generating Processes. Further, the methodology is equivalent to current methods if the data support them and is therefore complementary to existing methods, without loss of interpretability of model parameters. For the MIS field it finds that Popularity Parameter (PP) of an article Keywords can predict whether a paper will be highly-cited (top 25% of highly-cited articles) between two to three years after publication and beyond. Furthermore, given the small number of iterations needed for convergence, the methodology can also be used as a baseline method in Big Data (BD) settings for both Artificial Intelligence (AI) and Machine Learning (ML) contexts as well. [ABSTRACT FROM AUTHOR]

• This study improved the citation prediction accuracy by applying the BP neural network. • The BP neural network significantly outperformed the other six baselines (XGBoost, RF, LR, SVR, KNN, and RNN). • Five essential features were determined for citation prediction. Predicting the citation counts of academic papers is of considerable significance to scientific evaluation. This study used a four-layer Back Propagation (BP) neural network model to predict the five-year citations of 49,834 papers in the library, information and documentation field indexed by the CSSCI database and published from 2000 to 2013. We extracted six paper features, two journal features, nine author features, eight reference features, and five early citation features to make the prediction. The empirical experiments showed that the performance of the BP neural network is significantly better than those of the six baseline models. In terms of the prediction effect, the accuracy of the model at predicting infrequently cited papers was higher than that for frequently cited ones. We determined that five essential features have significant effects on the prediction performance of the model, i.e., 'citations in the first two years', 'first-cited age', 'paper length', 'month of publication', and 'self-citations of journals', and the other features contribute only slightly to the prediction. [ABSTRACT FROM AUTHOR]

This paper presents a fuzzy neural network predictive control scheme for studying the coagulation process of wastewater treatment in a paper mill. An adaptive fuzzy neural network is employed to model the nonlinear relationships between the removal rate of pollutants and the chemical dosages, in order to adapt the system to a variety of operating conditions and acquire a more flexible learning ability. The system includes a fuzzy neural network emulator of the reaction process, a fuzzy neural network controller, and an optimization procedure based on a performance function that is used to identify desired control inputs. The gradient descent algorithm method is used to realize the optimization procedure. The results indicate that reasonable forecasting and control performances have been achieved through the developed system. [Copyright &y& Elsevier]

Abstract: Using advanced machine learning techniques as an alternative to conventional double-entry volume equations, a regression model of the inside-bark volume (dependent variable) for standing Eucalyptus globulus trunks (or main stems) has been built as a function of the following three independent variables: age, height and outside-bark diameter at breast height (DBH). The experimental observed data (age, height, outside-bark DBH and inside-bark volume) for 142 trees (E. globulus) were measured and a nonlinear model was built using a data-mining methodology based on support vector machines (SVM) and multilayer perceptron networks (MLP) for regression problems. Coefficients of determination and Furnival’s indices indicate the superiority of the SVM with a radial kernel over the allometric regression models and the MLP. [Copyright &y& Elsevier]

Lower-dimensional representation of scientific text has attracted much attention among researchers due to its impact on many data mining and recommendation tasks. This paper studies two main research streams in scientific literature representation. First, both local and distributed representation viewpoints are reviewed and their advantages and disadvantages in lower dimensional representation are discussed. The paper then proposes a novel hybrid distributed technique for text representation. Using scientific articles as the major source of textual information, both the article's content and citation network are used to build a distributed and universal lower dimensional representation. The superiority of the new technique to the traditional methods is then justified in predicting the existence of links in large citation graphs. [ABSTRACT FROM AUTHOR]

A new adsorbent was prepared from municipal wastes (a mixture of Corn Stover, Paper Waste, and Yard Waste) by cationization with 3 ̶ Chloro ̶ 2 ̶ Hydroxypropyl Trimethylammonium Chloride. The FTIR spectrum confirmed the quaternary ammonium group's presence on the adsorbent surface (1450 cm−1). The maximum adsorption capacity (148 mg/g) was higher than the earlier reported values. Liu isotherm described well the adsorption process, with a high R2 adj value (0.997). The pseudo-first-order equation fits well for kinetic data, and thermodynamic experiments demonstrated the endothermic nature of the adsorption. The deep neural network (DNN) is applied to simulate the adsorption process, which outperformed the classical machine learning and shallow neural network models. The DNN model predicted accurately the adsorption process with the lowest deviation from the actual values with Mean Absolute Error (MAE = 3.2), Root Mean Squared Error (RMSE = 4.89), and the highest performance accuracy of R2 (0.96) as compared to various classical ML algorithms such as Linear Regressions (MAE = 12.53, RMSE = 18.01, R2 = 0.42), Random Forest (MAE = 5.81, RMSE = 10.05, R2 = 0.82), and Extra Trees (MAE = 4.35, RMSE = 8.22, R2 = 0.88). The utilized DNN model can be used for predicting the removal efficiency of dyes for various combinations of input parameters without going through laboratory experiments. ga1 • Matrix of municipal waste was converted into an absorbent by surface cationization. • The adsorbent had higher adsorption capacity for anionic dye Indigo Carmine. • Liu isotherm described well the adsorption process. • Classical machine learning models didn't predict accurately. • Deep Neural Network was able to model the adsorption process with 96% accuracy. [ABSTRACT FROM AUTHOR]

This paper first formulates a novel long-term beam power allocation (BPA) problem to tackle the harmful co-linear interference issue in the geostationary earth orbit (GEO) and low earth orbit (LEO) co-existing satellite system. This BPA problem intends to optimize the long-term weighted sum rate of the LEO system while ensuring that GEO user's received interference from the LEO satellite system is lower than a pre-fixed threshold. To solve it in a real-time manner, a deep reinforcement learning (DRL) framework based on the proximal policy optimization (PPO) algorithm is proposed, named as drlBPA. In addition, for the existing most relevant baseline, the fractional optimization (FO)-based BPA scheme, on the one hand, this paper improves it via a greedy strategy to fully exploit time resource. On the other hand, to further reduce the computational complexity stemming from its iterative solving procedure, a deep neural network approximation scheme is also developed. Simulation results demonstrate that (i) The trained DRL model of the proposed drlBPA scheme has good convergence and generality performance. (ii) Compared with the three FO-based benchmarks, the drlBPA scheme not only achieves the highest throughput of the LEO system within a significantly reduced computation time, but also yields the best system stability. • Maximizes the LEO system throughput while ensuring the GEO system service quality. • Builds a PPO algorithm based deep reinforcement learning framework. • Constructs two benchmark schemes: improved FO (IFO) and the DNN-accelerated IFO. • Achieves the highest LEO system throughput with the lowest computational complexity. • Yields the best system stability compared with the benchmarks. [ABSTRACT FROM AUTHOR]

High precision and low complexity control algorithm plays an important role in the developing of the end-effector instrumentation of different robot manipulators. In order to reduce the kinetic energy and the high-speed drift phenomenon of the repetitive motion tracking task, the robot manipulator needs to calibrate its configuration. In this paper, we formulate the configuration calibration of the robot manipulator for the repetitive motion task as a future quadratic programming optimization problem constrained with equality constraints, which is also regarded as a fundamental problem in artificial intelligence and modern control engineering. Zhang neural network, which is a canonical method, can be adopted to deal with the continuous form of the future optimization problem, named as temporally dependent quadratic programming problem with equality constraints. In order to overcome the issue of temporally dependent inverse computing, a novel Zhang neural network model and its uncertain disturbance tolerant model, which are termed as filtered reciprocal-kind Zhang neural network model and uncertain disturbance tolerant filtered reciprocal-kind Zhang neural network model, respectively, are proposed by integrating the energy-type cost function and Zhang neural network design formula for solving the temporally dependent quadratic programming problem with equality constraints in this paper. Based on the Euler discrete formula and the models, the discrete filtered reciprocal-kind Zhang neural network and the discrete uncertain disturbance tolerant filtered reciprocal-kind Zhang neural network algorithms are proposed for solving the future quadratic programming problem with equality constraints and the robot manipulator configuration calibration problem of repetitive motion. The convergence properties of the reciprocal-kind Zhang neural network model and its corresponding uncertain disturbance tolerant model are obtained by Lyapunov stability theory of nonlinear system and its corresponding perturbed system, while the convergence property of the filtered reciprocal-kind Zhang neural network model is analyzed by the limit thinking. For the repetitive motion task, three experiments for solving the configuration calibration problem of PUMA560, Kinova Jaco2, and Franka Emika Panda robot manipulators are performed to illustrate the effectiveness, robustness and superiority of our proposed discrete filtered reciprocal-kind Zhang neural network algorithms. • A novel configuration calibration scheme of the robot manipulators is formulated. • An inverse-free Zhang neural network algorithm is proposed. • The proposed algorithm is robustness under uncertain disturbance environment. [ABSTRACT FROM AUTHOR]

Diabetic retinopathy (DR) poses a significant challenge in diabetes management, with its progression often asymptomatic until advanced stages. This underscores the urgent need for cost-effective and reliable screening methods. Consequently, the integration of artificial intelligence (AI) tools presents a promising avenue to address this need effectively. We provide an overview of the current state of the art results and techniques in DR screening using AI, while also identifying gaps in research for future exploration. By synthesizing existing database and pinpointing areas requiring further investigation, this paper seeks to guide the direction of future research in the field of automatic diabetic retinopathy screening. There has been a continuous rise in the number of articles detailing deep learning (DL) methods designed for the automatic screening of diabetic retinopathy especially by the year 2021. Researchers utilized various databases, with a primary focus on the IDRiD dataset. This dataset consists of color fundus images captured at an ophthalmological clinic situated in India. It comprises 516 images that depict various stages of DR and diabetic macular edema. Each of the chosen papers concentrates on various DR signs. Nevertheless, a significant portion primarily focused on detecting exudates, which remains insufficient to assess the overall presence of this disease. Various AI methods have been employed to identify DR signs. Among the chosen papers, 4.7 % utilized detection methods, 46.5 % employed classification techniques, 41.9 % relied on segmentation, and 7 % opted for a combination of classification and segmentation. Metrics calculated from 80 % of the articles employing preprocessing techniques demonstrated the significant benefits of this approach in enhancing results quality. In addition, multiple DL techniques, starting by classification, detection then segmentation. Researchers used mostly YOLO for detection, ViT for classification, and U-Net for segmentation. Another perspective on the evolving landscape of AI models for diabetic retinopathy screening lies in the increasing adoption of Convolutional Neural Networks for classification tasks and U-Net architectures for segmentation purposes; however, there is a growing realization within the research community that these techniques, while powerful individually, can be even more effective when integrated. This integration holds promise for not only diagnosing DR, but also accurately classifying its different stages, thereby enabling more tailored treatment strategies. Despite this potential, the development of AI models for DR screening is fraught with challenges. Chief among these is the difficulty in obtaining the high-quality, labeled data necessary for training models to perform effectively. This scarcity of data poses significant barriers to achieving robust performance and can hinder progress in developing accurate screening systems. Moreover, managing the complexity of these models, particularly deep neural networks, presents its own set of challenges. Additionally, interpreting the outputs of these models and ensuring their reliability in real-world clinical settings remain ongoing concerns. Furthermore, the iterative process of training and adapting these models to specific datasets can be time-consuming and resource-intensive. These challenges underscore the multifaceted nature of developing effective AI models for DR screening. Addressing these obstacles requires concerted efforts from researchers, clinicians, and technologists to develop new approaches and overcome existing limitations. By doing so, a full potential of AI may transform DR screening and improve patient outcomes. [ABSTRACT FROM AUTHOR]

• The indicators were collected based on three levels of situation awareness (SA). • The drivers' SA assessment system was built using grey correlation analysis. • The constructed backpropagation neural network model can effectively predict SA. Based on combining the relevant studies on situation awareness (SA), this paper integrated multiple indicators, including eye movement, electroencephalogram (EEG), and driving behavior, to evaluate SA. SA is typically divided into three stages: perception, understanding and prediction. This paper used eye movement indicators to represent perception, EEG indicators to represent understanding, and driving behavior indicators to represent prediction. After identifying indicators for evaluating SA, a driving simulation experiment was designed to collect data on the indicators. 41 subjects were recruited to participate in the investigation, and the experimenter collected data from each subject in a total of 9 groups. After removing 4 groups of invalid data, 365 groups of valid data were finally obtained. The grey correlation analysis was used to optimize the SA indicators, and 10 SA evaluation indicators were finally determined. There were the average fixation duration, the nearest neighbor index, pupil area, the percentage power spectral density values of the 3 rhythmic waves (θ, α, β), rhythmic wave energy combination parameters (α / θ), mean speed, SD of speed and acceleration. Taking the optimized 10 indicators as input and the SA scores as output, a backpropagation neural network model with a topological structure of 10-8-1 was constructed. 75% of the data were randomly selected for model training, and the final network training's mean square error was 0.0025. Using the remaining 25% of data for verification, the average absolute error and average relative error of the predicted results are 0.248 and 0.046, respectively. This showed that the model was effective, and it was feasible to evaluate the SA by using the data of eye movement, EEG and driving behavior parameters. [ABSTRACT FROM AUTHOR]

[Display omitted] • Large datasets of U and UP are constructed with 1526 and 743 fuel samples. • The effect of C, H, and O atomic ratios on HHV is investigated by ternary plots. • HHV of any type of waste and fuel is estimated by artificial neural networks. • Hyperparameter optimization is carried out to obtain best performing ANN models. • Ultimate analysis data is observed to be sufficient for HHV estimation of fuels. Higher heating value (HHV) is one of the most important parameters in determining the quality of the fuels. In this study, comparatively large datasets of ultimate and proximate analysis are constructed to be used in HHV estimation of several classes of fuels, including char & fossil fuels, agricultural wastes, manure (chicken, cow, horse, sheep, llama, and pig), sludge (like paper, paper-mil, sewage, and pulp), micro/macro-algae's, wastes (RDF and MSW), treated woods, untreated woods, and others (non-fossil pyrolysis oils) between the HHV range of 4.22–55.55 MJ/kg. The relationships of carbon, hydrogen, and oxygen atomic ratios for fuel classes are illustrated by using ternary plots, and the effects of elemental composition on HHV was analyzed with the extensive dataset. Then, the ultimate (U) and ultimate & proximate (UP) datasets were utilized separately to estimate the HHV by using artificial neural networks (ANN). Hyperparameter optimization was carried out and the best performing ANNs were determined for each dataset, which yielded R2 values of 0.9719 and 0.9715, respectively. The results indicated that while ANNs trained by both datasets perform remarkably well, utilization of U dataset is sufficient for HHV estimation. Finally, the best performing ANN models for both U and UP datasets are given in a directly utilizable format enabling the accurate estimation of HHV of any fuel for optimization of fuel processing and waste management operations. [ABSTRACT FROM AUTHOR]

The unstoppable trend of globalization has made the integration of voice recognition with science and technology increasingly close, and has called for the emergence of more powerful oral voice recognition systems. At present, most speech recognition research uses Hidden Markov Model-Gaussian Mixture Modell (HMM-GMM) and onnectionist Temporal Classification (CTC) end-to-end method, which has obvious limitations. Therefore, based on neural network, this paper introduces deep neural network posterior probability algorithm into speech recognition system, and constructs the spoken speech recognition system based on deep neural network posterior probability algorithm. The design principle of the constructed speech system is discussed and the performance analysis and effectiveness evaluation are carried out. The experimental results show that the performance of the neural network-based speech recognition system constructed in this paper has made a breakthrough compared with the traditional technology, and the Character Error Rate (CER) is only 4.6%, and the End-to-End Latency and Equal Error Rate (EER) have reached a low level. This study points out the future development direction of speech recognition and provides the impetus to promote its further development [ABSTRACT FROM AUTHOR]

Sintering performance index can reflect the quality of sintered ore, and the level of sintered ore performance directly affects the stability of blast furnace production. Accurate prediction of sinter performance metrics is essential to optimise sinter quality, improve production efficiency and reduce energy consumption. Based on the sintering big data, this paper proposes the Optuna-DFNN model by combining the fuzzy neural network algorithm, the deep neural network algorithm and the Optuna framework to address the problems of difficulty in tuning the parameter, lack of self-learning ability, and poor generalisation of the model of the traditional method when faced with the input of multiple parameters. The paper predicts the four prediction indexes of yield rate, return fines ratio, drum index and RDI +3.15 respectively. Among them, the relative error of the prediction of the yield rate is within 1.3 %, the relative error of the prediction of the return fines ratio is within 2.8 %, the relative error of the prediction of the drum index is within 0.32 %, and the relative error of the prediction of the RDI +3.15 is within 1.4 %. The results indicate that the Optuna-DFNN model has better performance in predicting sintering performance metrics. Based on the Optuna-DFNN sintered ore performance prediction model, the quality of sintered ore can be accurately predicted, the sintering process parameters can be adjusted in a timely manner, the sintering process can be optimized, and the utilization coefficient of sintering can be improved. Provide a reference for the optimisation of production processes, with a positive impact on cost reduction, environmental protection and increased sustainability of production. [ABSTRACT FROM AUTHOR]

In order to accurately estimate the period of X-ray pulsars to improve navigation performance, this paper proposes a pulsar period estimation method based on photon energy using reverse Hilbert coding and double Convolutional Neural Networks (CNNs) discrimination. This method uses photon energy information to perform epoch folding. It converts one-dimensional contour information into two-dimensional image signals through reverse Hilbert curve and recognizes contour image features through CNN models to determine the optimal period. According to the simulation results, the period estimation accuracy of this method is 0.5350 ns, which is 65.71% higher than the χ 2 epoch folding method. The contour phase accuracy is 0.0004767, which is 41.93% higher than the χ 2 epoch folding method. At the same time, the observation time and noise interference that may affect the accuracy of period estimation are also quantitatively analyzed. In addition, in order to simulate real navigation scenarios, this paper also conducts dynamic period estimation experiments on PSR B0531+21 pulsar in a specific period of time. The method proposed in this paper has the advantages of high precision, fast calculation speed, strong anti-interference ability and accurate dynamic period estimation, which can significantly improve the navigation performance of X-ray pulsars. • The proposed period estimation method uses photon energy information for contour weighted folding. • The proposed method uses reverse Hilbert curve to convert one-dimensional contour information into two-dimensional image information. • Convolutional Neural Network models are used to optimize pulsar period estimation. [Display omitted] [ABSTRACT FROM AUTHOR]

• A model to explore the multiphysics characteristic of semiconductors is presented. • The model determines the thermal and mechanical properties of complex patterns. • Results show a strong correlation between thermal and mechanical properties. • Dielectric bands asymmetrically affect thermal and mechanical behavior. • Machine learning can predict one property from another. As semiconductor performance improves through advanced package designs, it becomes important to consider both thermal and mechanical properties. Better understanding their relationship enhances system design and optimization. This study has introduced a numerical method that takes into account both thermal and mechanical fluxes to construct thermal and mechanical property maps for practical application in semiconductor engineering. Furthermore, this paper investigated the relationship between the two properties using the constructed property maps in commercially available semiconductor packages. Owing to the complexity of packaging design patterns, a coupled isoparametric mapping and machine learning (ML) method was introduced to investigate their effects. The model then determines and investigates the anisotropic equivalent thermal and mechanical properties of the commercialized semiconductor packages with complex pattern designs, according to the package materials, volume fractions of each material, and design patterns. This approach pursues more sophistication and practicality compared to simplified composite structure property evaluation models. Additionally, all processes in the algorithm are automated based on ML methods, making them practically applicable in the semiconductor industry. The study shows that there is a strong correlation between the thermal and mechanical characteristics in the complex package patterns. This relationship was able to form a fairly clear category based on the shape of the dielectric band because the dielectric band had significantly different effects on the thermal and mechanical fluxes. The study discussed the physical implications of the similarities and differences between the two behaviors and validated the results of the equivalent properties through finite element (FE) analysis. Finally, the study cross-validates the relationship by showing that information from one flux behavior can be used to predict the other based on the ML method. Based on the results, this paper can provide a better understanding of thermal and mechanical fluxes in semiconductor package patterns for package design and optimization. [ABSTRACT FROM AUTHOR]

During the COVID-19 pandemic, the healthcare sector faced unprecedented challenges in effectively managing hospital resources. A crucial aspect of resource planning and allocation is the ability to predict the expected length of a patient's stay. Detecting whether a patient requires extended hospitalization or a shorter stay becomes vital for efficient hospital resource allocation and utilization. This paper aims to build a deep learning-based analytical model named LoSNet that predicts the length of stay for each patient at the time of admission to the Hospital. The early prediction of the length of stay requirement would aid healthcare professionals in optimizing the utility of hospital beds and other resources. In this direction, this paper compares the prediction ability of various machine-learning models including Random Forest, Decision Tree, Logistic Regression, and Naïve Bayes with a customized deep neural network model. The dataset used for this analysis includes ground truth on 3,18,438 patients' length of stay categorized into eleven classes such as 0-10 days being one class, 11-20 days being another class, and so on to more than 100 days. The methodology employed in this study involves data collection, data transformation, and training LoSNet, a deep neural network with no attention mechanisms. The results indicate impressive performance over other models and a random classifier, with a cross-entropy loss of 1.531 and an accuracy of 0.408 in predicting hospitalization durations in an 11-class classification setup, highlighting the framework's effectiveness in healthcare management. [ABSTRACT FROM AUTHOR]

Machine learning performs well in many problems. However, the tendency to generate predictions that violate theoretical knowledge makes it difficult to apply to practical processing. To resolve this situation, this paper combines domain knowledge with a data-driven model, proposes a theory-guided machine learning framework based on a parameter transfer strategy, and applies it to the width prediction of plates after multiple passes of hot rolling. The framework applies a swarm optimization algorithm to the original theoretical model and generates numerous highly-physical consistent samples. The established deep neural network (DNN) model is trained with simulated data, and the parameters are fine-tuned using a parameter transfer strategy combined with actual data to ensure excellent adaptation to the actual environment based on adequate learning of theoretical knowledge. In tests, the proposed model had the best overall prediction performance in this paper. Meanwhile, the developed model is consistent with the existing perception of rolling theory. This allows for the quick and reliable application of machine learning models in production. [Display omitted] • A novel theory-guided machine learning framework for width prediction of rolled plates. • Pre-training of the model using the data generated by the optimized theoretical model. • The proposed model has the best accuracy in both sufficient and less actual samples. • The variable influence of the proposed model is consistent with theoretical perceptions. [ABSTRACT FROM AUTHOR]

Autonomous spacecraft relative navigation via monocular images became a hot topic in the past few years and, recently, received a further push thanks to the constantly growing field of artificial neural networks and the publication of several spaceborne image datasets. Despite the proliferation of spacecraft relative-state initialization algorithms developed, most architectures adopt computationally expensive solutions relying on convolutional neural networks (CNNs) that provide accurate output at the cost of a high computational burden that seems unfeasible for current spaceborne hardware. The paper addresses this issue by proposing a novel pose initialization algorithm based on lightweight CNNs. Inspired by previous state-of-the-art algorithms, the developed architecture leverages a fast and accurate target detection CNN followed by a line segment detection CNN capable of running with low inference time on mobile devices. The line segments and their junctions are grouped into complex geometrical groups, reducing the solution search space, and subsequently, they are adopted to extract the final pose estimate. As a main outcome, the analyses demonstrate that the lightweight architecture developed scores high accuracy in the pose estimation task, with a mean estimation error of less than 10 cm in translation and 2.5°in rotation. The baseline algorithm scores a mean SLAB error of 0.04552 with a standard deviation of 0.22972 in the test dataset. Detailed analyses demonstrate that the uncertainties on the overall pose score are driven mainly by errors in the relative attitude, which gives the highest contribution to the pose error metric adopted. The analyses on the error distributions point out that the uncertainties on the estimated relative position are higher in the camera boresight axis direction. Concerning the relative attitude, the algorithm proposed has higher uncertainties in estimating directions of the target x and y axes due to ambiguities related to the target geometry. Notably, the target detection CNN trained in this work outperforms the previous top scores in the benchmark dataset. The performances of the proposed algorithm have been investigated further by analyzing the effects on the accuracy due to the relative distance and the presence of background in the images. Lastly, the paper delves into the possibility of adopting a sub-portion of the 2D-to-3D match matrix made by the most complex perceptual groups identified that positively affects the overall run-time, pointing out the performances in terms of accuracy of the estimates and providing a comparison of both the baseline and the reduced match matrix versions against state-of-the-art algorithms concerning relative position and attitude errors and solution availability, highlighting the high accuracy and solution availability of the proposed architectures. • New relative pose initialization algorithm based on the detection of line segments. • Detailed literature review and description of available spaceborne image datasets. • Improvement of state of the art in spacecraft detection using lightweight CNN. • Assessment of accuracy of the proposed architecture and error sources analyses. • Proven feasibility of the approach via comparison with other architectures. [ABSTRACT FROM AUTHOR]

This paper presents a novel machine learning (ML) based approach to predict machining cycle (part program running) times for complex 5-axis machining. Typical 5-axis machining toolpaths consist of short-segmented discrete linear cutter location lines (CL-lines) with simultaneously varying tool center point (TCP) and orientation vectors (ORI). As the 5-axis machine tool NC (numerical control) system tries to interpolate such part programs smoothly, the TCP motion decelerates and accelerates repeatedly causing the actual feedrate to fluctuate. The actual observed (resultant) feedrate can be approximately 30 % lower than the user commanded one. Furthermore, if the toolpath requires the 5-axis machine tool to travel closer to its singular point, or if the workpiece placement on the table is not optimal, actual feedrate is even lowered further. This paper presents two ML-based approaches to accurately predict 5-axis machining toolpaths. The first strategy uses a bidirectional long short-term memory (Bi-LSTM) network to model the machine tool behavior and generates a direct final cycle time estimation for any given part program. This approach only uses the toolpath geometry to predict the cycle time, and for training it only needs the "final" cycle times of similar part programs making it practical on today's shop floors. The second strategy requires individual CL-line processing times for training. In return, it can provide highly accurate cycle time predictions. Both strategies can capture the interpolation dynamics of 5-axis machine tool NC systems accurately. Simulation studies and experimental validations are conducted on modern 5-axis machine tools with varying workpiece placements and interpolation parameters. Proposed approaches have shown to predict cycle times with 90–95 % accuracy on real-life complex 5-axis machining toolpaths. [ABSTRACT FROM AUTHOR]

With the rapidly evolving geo-political landscape and unceasing advancements in technology, sustainable energy security is a very important topic. Research indicates that electrochemical energy systems are quite promising to solve many of energy conversion, storage, and conservation challenges while offering high efficiencies and low pollution. The paper provides an overview of electrochemical energy devices and the various optimization techniques used to evaluate them. The optimization techniques include linear regression, factorial design, the Taguchi method, artificial neural networks, filters, and a combination of such methods to improve these systems. To support the growing interest in research, the bulk of this study focuses on a review of the most promising and highly researched electrochemical energy devices, such as fuel cells, batteries, and supercapacitors. The paper also provides modest commentary on hydrogen production technologies, electrochemical reactors, and membrane separation technologies, amongst other technologies. Building on a previous paper by the authors of this paper on artificial intelligence in hybrid renewable energy systems with fuel cells, this work provides a comparative review of optimization techniques for supercapacitors by highlighting key findings based on model accuracy. A summary of the advantages and disadvantages of the different major optimization techniques is presented. The paper concludes that a combination of optimization techniques is used to overcome the drawbacks of individual techniques, with adaptive filters being the most widely studied. This paper presents studies on the Design of Experiments (DoE) with the goal of building a better understanding of the relationships that exist between different operating variables in various electrochemical devices. • The electrochemical systems can be optimized for better performance with high accuracy. • The optimization techniques used are either model-based or data-driven. • Linear regression and artificial neural networks are the most common techniques studied. • Hybrid models could be applied to enhance systems' accuracy and minimize errors. [ABSTRACT FROM AUTHOR]

Large deep learning models are impressive, but they struggle when real-time data is not available. Few-shot class-incremental learning (FSCIL) poses a significant challenge for deep neural networks to learn new tasks from just a few labeled samples without forgetting the previously learned ones. This setup can easily leads to catastrophic forgetting and overfitting problems, severely affecting model performance. Studying FSCIL helps overcome deep learning model limitations on data volume and acquisition time, while improving practicality and adaptability of machine learning models. This paper provides a comprehensive survey on FSCIL. Unlike previous surveys, we aim to synthesize few-shot learning and incremental learning, focusing on introducing FSCIL from two perspectives, while reviewing over 30 theoretical research studies and more than 20 applied research studies. From the theoretical perspective, we provide a novel categorization approach that divides the field into five subcategories, including traditional machine learning methods, meta learning-based methods, feature and feature space-based methods, replay-based methods, and dynamic network structure-based methods. We also evaluate the performance of recent theoretical research on benchmark datasets of FSCIL. From the application perspective, FSCIL has achieved impressive achievements in various fields of computer vision such as image classification, object detection, and image segmentation, as well as in natural language processing and graph. We summarize the important applications. Finally, we point out potential future research directions, including applications, problem setups, and theory development. Overall, this paper offers a comprehensive analysis of the latest advances in FSCIL from a methodological, performance, and application perspective. • In-depth survey on FSCIL methods, applications, performance. • Categorizing FSCIL into five subcategories for clear analysis. • Evaluating FSCIL research on benchmarks for strengths, weaknesses. • FSCIL applications in computer vision, NLP, and graph. [ABSTRACT FROM AUTHOR]

Deep neural networks have achieved outstanding performance in computer vision tasks. Convolutional Neural Networks (CNNs) typically operate in the spatial domain with raw images, but in practice, images are usually stored and transmitted in their compressed representation where JPEG is one of the most widely used encoder. Also, these networks are computationally intensive and slow. This paper proposes performing the learning and inference processes in the compressed domain in order to reduce the computational complexity and improve the speed of popular CNNs. For this purpose, a novel graph-based frequency channel selection method is proposed to identify and select the most important frequency channels. The computational complexity is reduced by retaining the important frequency components and discarding the insignificant ones as well as eliminating the unnecessary layers of the network. Experimental results show that the modified ResNet-50 operating in the compressed domain is up to 70% faster than the spatial-based traditional ResNet-50 while resulting in similar classification accuracy. Moreover, this paper proposes a preprocessing step with partial encoding to improve the resilience to distortions caused by low-quality encoded images. Finally, we show that training a network with highly compressed data can achieve a good classification accuracy with up to 93% reduction in the storage requirements of the training data. [ABSTRACT FROM AUTHOR]

This paper analyzes global robust exponential stability of interval bidirectional associative memory (BAM) neural networks with multiple time-varying delays, proposes a direct method based on system solutions, and gives sufficient conditions under which interval BAM neural networks have a unique and globally robustly exponentially stable equilibrium point. This method not only avoids the difficult to set up any Lyapunov–Krasovskii functional, but also derives simpler global robust exponential stability criteria. Compared with the data from other literature, the robust exponential stability criteria obtained in this paper have been presented to have more merits theoretically and numerically. • A new method based on system solutions is proposed to investigate robust GES criteria. • The obtained robust GES criteria are more effective than the existing ones. • The method is applicable to several neural network models after a small change. [ABSTRACT FROM AUTHOR]

With the prevalence of hardware accelerators as an integral part of the modern systems on chip (SoCs), the ability to model accelerators quickly and accurately within the system in which it operates is critical. This paper presents gem5-SALAMv2 as a novel system architecture for LLVM-based modeling and simulation of custom hardware accelerators integrated into the gem5 framework. It overcomes the inherent limitations of state-of-the-art trace-based pre-register-transfer level (RTL) simulators by offering a truly "execute-in-execute" LLVM-based model. It enables scalable modeling of multiple dynamically interacting accelerators with full-system simulation support. To create long-term sustainable expansion compatible with the gem5 system framework, gem5-SALAM offers a general-purpose and modular communication interface and memory hierarchy integrated into the gem5 ecosystem, streamlining designing and modeling accelerators for new and emerging applications. gem5-SALAMv2 expands upon the framework established in gem5-SALAMv1 with improved LLVM-based elaboration and simulation, improved and more extensible system integration, and new automations to simplify rapid prototyping and design space exploration. 1 1 Conference Paper Extension: This work extends the work presented in gem5-SALAM: A System Architecture for LLVM-based Accelerator Modeling from MICRO 2020 (Rogers et al., 2020). This work expands on the aforementioned work by revamping the gem5-SALAM internals to provide more robust and extensible simulations, introducing new automation tools for expanding and simplifying design space exploration, and demonstrates the new capabilities of gem5-SALAMv2 by exploring multiple configurations of simple neural network architectures. Validation on the MachSuite (Reagen et al., 2014) benchmarks presents a timing estimation error of less than 1% against the Vivado High-Level Synthesis (HLS) tool. Results also show less than a 4% area and power estimation error against Synopsys Design Compiler. Additionally, system validation against implementations on an Ultrascale+ ZCU102 shows an average end-to-end timing error of less than 2%. Lastly, we demonstrate the upgraded capabilities of gem5-SALAMv2 by exploring accelerator platforms for two deep neural networks, LeNet5 and MobileNetv2. In these explorations, we demonstrate how gem5-SALAMv2 can simulate such systems and guide architectural optimizations for these types of accelerator-rich architectures. 2 2 The most up-to-date version of gem5-SALAMv2 is available at https://github.com/TeCSAR-UNCC/gem5-SALAM.. [ABSTRACT FROM AUTHOR]

With the acceleration of globalization, machine translation (MT) plays an increasingly prominent role in cross-language communication. However, how to evaluate the quality of machine translation, especially considering the nuances in different semantic contexts, remains a challenge. This paper proposes an automatic scoring model for machine translation quality based on deep transfer learning, which aims to accurately perceive and evaluate the quality of translated texts in different semantic contexts. Firstly, a pre-trained deep neural network model is used to extract the semantic feature representation of the sentences in the source language and the target language, so as to capture the semantic information of the sentences. Then, deep transfer learning is used to map the semantic features of source language and target language into the shared feature space. By sharing feature space, an effective relationship is established between the semantic representations of two languages, so as to achieve cross-language quality evaluation. The experimental results show that the model has made significant progress in evaluating the quality of machine translation. Compared with traditional methods, this model can evaluate translation quality more accurately and has better generalization ability. The automatic scoring model of machine translation quality based on deep transfer learning not only improves the accuracy and efficiency of machine translation quality assessment, but also provides new ideas and methods for future research. [ABSTRACT FROM AUTHOR]

Highly distorted closed surfaces pose significant challenges for machining trajectory planning due to their intricate surface constraints and closed structures. Despite these challenges, components with such features are prevalent in industries like aerospace. This paper presents a machine learning-driven multi-objective optimization method for electrical discharge machining (EDM) trajectory planning of highly distorted closed surfaces. The method transforms the structural design of forming electrodes and trajectory planning into a multi-objective decision problem. And a discrete point trajectory planning method, guided by surface average curvature, is employed to determine the optimal position and orientation of the electrode. Additionally, an elite dataset, generated using the Monte Carlo method and Arena's Principle, is utilized to train an artificial neural network (ANN). This network predicts hyperparameters for the nonlinear optimization problem. Based on the proposed method, a multi-objective optimization model is formulated for an integral shrouded blisk, considering minimization of iteration count, axial motion, and maximization of machining surface quality. The Pareto front is utilized to obtain the optimal EDM trajectory. Experimental results demonstrate a 17.38 % reduction in the overall machining cycle duration using this trajectory, and the surface roughness and profile accuracy satisfy the design specifications, which proves the effectiveness of this method. • A machine learning-driven multi-objective optimization method for machining trajectory planning is proposed. • A discrete point trajectory planning method for EDM of highly distorted closed surfaces is deduced. • An elite dataset construction method base on the Arena's principle and Monte Carlo method is proposed. • The structural design of shaped electrodes and trajectory planning are transformed into a multi-objective decision problem. [ABSTRACT FROM AUTHOR]

This paper aims to investigate the performance enhancement of a novel X-type rotary engine through hydrogen doping and proposes an optimal control strategy under part-load conditions. For this reason, an integrated model combining a dual zone model with recurrent neural network model for a hydrogen-doping X-type rotary engine is developed. Then, the effects of operating parameters on the comprehensive performance of engine are analyzed. Finally, an optimal multi-variable control strategy (OMCS) is introduced to achieve optimal performance under part-load conditions. The results show that the optimization of hydrogen is more pronounced at high rotational speeds, small ignition angles, or lean combustion conditions. Compared with single-variable control strategies, OMCS offers a broader regulation range and can maintain thermal efficiency above 0.27. At high loads, OMCS achieves high thermal efficiency by adjusting the equivalence ratio and ignition timing, albeit with a slight compromise on NO emission performance. • A novel hydrogen-doping X-type rotary engine with special structure is proposed. • A novel integrated model combining dual zone model and RNN is established. • Effects of multi-parameters on power, thermal efficiency and emission are analyzed. • The engine optimal control strategy under part-load conditions is investigated. [ABSTRACT FROM AUTHOR]

To enhance solution accuracy and training efficiency in neural network approximation to partial differential equations, partitioned neural networks can be used as a solution surrogate instead of a single large and deep neural network defined on the whole problem domain. In such a partitioned neural network approach, suitable interface conditions or subdomain boundary conditions are combined to obtain a convergent approximate solution. However, there has been no rigorous study on the convergence and parallel computing enhancement on the partitioned neural network approach. In this paper, iterative algorithms are proposed to enhance parallel computation performance in the partitioned neural network approximation. Our iterative algorithms are based on classical additive Schwarz domain decomposition methods. For the proposed iterative algorithms, their convergence is analyzed under an error assumption on the local and coarse neural network solutions. Numerical results are also included to show the performance of the proposed iterative algorithms. [ABSTRACT FROM AUTHOR]

Forward–backward doubly stochastic differential equations (FBDSDEs) serve as a probabilistic interpretation of stochastic partial differential equations (SPDEs) with diverse applications. Coupled FBDSDEs encounter numerous challenges in numerical approximation compared to forward–backward stochastic differential equations (FBSDEs) and decoupled FBDSDEs, including ensuring the measurability of the numerical solutions, accounting for the mutual influences between forward and backward processes, and considering the relationship with respect to SPDEs rather than PDEs. This paper introduces, for the first time, a numerical method for solving multi-dimensional coupled FBDSDEs. By integrating an optimal control-based approach with deep neural networks, it effectively addresses the coupling-related challenges between forward and backward equations. Computational examples of coupled FBDSDEs with explicit solutions demonstrate that the proposed deep learning-based numerical algorithm achieves commendable performance in terms of both accuracy and efficiency. [ABSTRACT FROM AUTHOR]

Physics-informed neural networks (PINNs) have shown remarkable achievements in solving partial differential equations (PDEs). However, their performance is limited when encountering oscillatory part in the solutions of PDEs. Therefore, this paper proposes a multi-scale deep neural network with periodic activation function to achieve high-frequency to low-frequency conversion, which can capture the oscillation part of the solution of PDEs. Moreover, the use of adaptive sampling method can adaptively change the location and distribution of residual points, improving the performance of the network. Additionally, the gradient-enhanced strategy is also utilized to embed the gradient information of the PDEs into the loss function of the neural network, which further improves the accuracy of PINNs. Through the numerical experiments verification, it is found that our method is better than PINNs in terms of accuracy and efficiency. • A MscaleDNNs structure with periodic activation functions is proposed. • The proposed neural network has better stability and robustness. • The effectiveness of the method was verified through numerical experiments. [ABSTRACT FROM AUTHOR]

The scale of meteorological data products does not match the requirements of application scenarios, which limits their application. It is suggested that large-scale reanalysis data must be downscaled before use. Attention mechanism is the key to high-performance downscaling models. However, in different application scenarios and different locations on the network, the attention mechanism is not always beneficial. In this paper, we propose a dynamic attention module that can adaptively generate weights for each branch based on input features, thereby dynamically suppressing unnecessary attention adjustments. At the same time, we propose a layer attention module, which can independently and adaptively aggregate the feature representation of different network layers. In addition, we design a unique loss function based on homoscedasticity uncertainty, which can directly guide the model to learn the numerical mapping relationship from low resolution to high resolution at the pixel level, and implicitly motivate the model to better reconstruct the data distribution of each meteorological field by guiding the model to learn the distribution difference between different meteorological fields. Experiments show that our model is more robust in time dimension, with an MAE average reduction of about 40% compared to VDSR and other methods in downscaling composite meteorological data. It can more accurately reconstruct multivariate high-resolution meteorological fields. Codes available at https://github.com/HitKarry/SDDN. [ABSTRACT FROM AUTHOR]

Change detection in high-resolution Earth observation is a fundamental Earth vision task to understand the subtle temporal dynamics of Earth's surface, significantly promoted by generic vision technologies in recent years. Vision Transformer is a powerful component to learning spatiotemporal representation but with enormous computation complexity, especially for high-resolution images. Besides, there is still lacking principles in designing macro architectures integrating these advanced vision components for various change detection tasks. In this paper, we present a deep probabilistic change model (DPCM) to provide a unified, interpretable, modular probabilistic change process modeling to address multiple change detection tasks, including binary change detection, one-to-many semantic change detection, and many-to-many semantic change detection. DPCM describes any complex change process as a probabilistic graphical model to provide theoretical evidence for macro architecture design and generic change detection task modeling. We refer to this probabilistic graphical model as the probabilistic change model (PCM), where DPCM is the PCM parameterized by deep neural networks. For parameterization, the PCM is factorized into many easy-to-solve distributions based on task-specific assumptions, and then we can use deep neural modules to parameterize these distributions to solve the change detection problem uniformly. In this way, DPCM has both theoretical macro architecture from PCM and strong representation capability of deep networks. We also present the sparse change Transformer for better parameterization. Inspired by domain knowledge, i.e., the sparsity of change and the local correlation of change, the sparse change Transformer computes self-attention within change regions to model spatiotemporal correlations, which has a quadratic computational complexity of the change region size but independent of image size, significantly reducing computation overhead for high-resolution image change detection. We refer to this instance of DPCM with sparse change Transformer as ChangeSparse to demonstrate their effectiveness. The experiments confirm ChangeSparse's superiority in speed and accuracy for multiple real-world application scenarios, such as disaster response and urban development monitoring. The code is available at https://github.com/Z-Zheng/pytorch-change-models. More resources can be found in http://rsidea.whu.edu.cn/resource_sharing.htm. [ABSTRACT FROM AUTHOR]

• An effective strategy to improve face recognition of low-quality facial images. • Validation of this strategy under forensically relevant challenging scenarios. • Cleaned versions of the BFW and Adience datasets are provided. • A new verification protocol for the Quis-Campi dataset. In forensic facial comparison, questioned-source images are usually captured in uncontrolled environments, with non-uniform lighting, and from non-cooperative subjects. The poor quality of such material usually compromises their value as evidence in legal proceedings. On the other hand, in forensic casework, multiple images of the person of interest are usually available. In this paper, we propose to aggregate deep neural network embeddings from various images of the same person to improve the performance in forensic comparison of facial images. We observe significant performance improvements, especially for low-quality images. Further improvements are obtained by aggregating embeddings of more images and by applying quality-weighted aggregation. We demonstrate the benefits of this approach in forensic evaluation settings with the development and validation of common-source likelihood ratio systems and report improvements in C llr both for CCTV images and for social media images. [ABSTRACT FROM AUTHOR]