Regularization Coefficient Research Articles

Online Tree-Structure-Constrained RPCA for Background Subtraction of X-ray Coronary Angiography Images

Background and objectiveBackground subtraction of X-ray coronary angiograms (XCA) can significantly improve the diagnosis and treatment of coronary vessel diseases. The XCA background is complex and dynamic due to structures with different intensities and independent motion patterns, making XCA background subtraction challenging. MethodsThe current work proposes an online tree-structure-constrained robust PCA (OTS-RPCA) method to subtract the XCA background. A morphological closing operation is used as a pre-processing step to remove large-scale structures like the spine, chest and diaphragm. In the following, the XCA sequence is decomposed into three different subspaces: low-rank background, residual dynamic background and vascular foreground. A tree-structured norm is introduced and applied to the vascular submatrix to guarantee the vessel spatial coherency. Moreover, the residual dynamic background is separately extracted to remove noise and motion artifacts from the vascular foreground. The proposed algorithm also employs an adaptive regularization coefficient that tracks the vessel area changes in the XCA frames. ResultsThe proposed method is evaluated on two datasets of real clinical and synthetic low-contrast XCA sequences of 38 patients using the global and local contrast-to-noise ratio (CNR) and structural similarity index (SSIM) criteria. For the real XCA dataset, the average values of global CNR, local CNR and SSIM are 6.27, 3.07 and 0.97, while these values over the synthetic low-contrast dataset are obtained as 5.15, 2.69 and 0.94, respectively. The implemented quantitative and qualitative experiments verify the superiority of the proposed method over seven selected state-of-the-art methods in increasing the coronary vessel contrast and preserving the vessel structure. ConclusionsThe proposed OTS-RPCA background subtraction method accurately subtracts backgrounds from XCA images. Our method might provide the basis for reducing the contrast agent dose and the number of needed injections in coronary interventions.

Prediction model of coal seam gas content based on kernel principal component analysis and IDBO-DHKELM

Abstract Accurate coal seam gas content assists in the effective prevention of coal and gas outburst accidents. To solve this problem, an IDBO-DHKELM coal seam gas content prediction model is proposed by combining improved Dung Beetle optimization algorithm (IDBO) with a deep hybrid kernel extreme learning machine (DHKELM). First, the index factors of the coupled gas content are determined according to the influence factors of coal seam gas content and the actual situation of mine production. The correlation of index factors is analyzed by SPSS 27 software via Pearson correlation coefficient matrix. Then, the principal components of the original data are extracted using the principal component analysis method (KPCA). Second, sine chaotic mapping, fusion improved sinusoidal algorithm, and fusion adaptive Gauss–Cauchy hybrid mutation perturbation are introduced to improve the Dung Beetle optimization algorithm (DBO) to enhance its global search capability. Third, IDBO is used to optimize the number of hidden layer nodes, regularization coefficient, penalty coefficient, and kernel parameter in DHKELM, which improves the prediction accuracy and further avoid the phenomenon of overfitting. Finally, the principal component extracted by KPCA is taken as the model’s input, and the gas content as the model’s output. The results are compared and analyzed with those of PSO-BPNN, GA-BPNN, PSO-SVM, and DPO-DHKELM models. The results demonstrate that the IDBO-DHKELM model’s performance is the best in each performance index. Compared with other models, the mean absolute error of test samples in the IDBO-DHKELM model is reduced by 0.402, 0.4407, 0.3554, and 0.0646, respectively. The mean absolute percentage error is decreased by 3.67%, 4.07%, 8.27%, and 6.35%, respectively. The root mean square error decreased by 0.7861, 0.7148, 0.3384, and 0.1186, respectively. The coefficient of determination (R 2) is increased by 0.1544, 0.1404, 0.0955, and 0.0396, respectively. Finally, the IDBO-DHKELM model and other models are applied to an experimental mine. The resulting IDBO-DHKELM model is the closest to the actual value, which further verifies the universality and reliability of the model. Therefore, the model is more suitable for the prediction of coal seam gas content.

Doppler Positioning with LEO Mega-Constellation: Equation Properties and Improved Algorithm

Doppler positioning, as an early form of positioning, has regained significant research interest in the context of Low Earth Orbit (LEO) satellites.Given the LEO mega-constellation scenario, the objective function of Doppler positioning manifests significant nonlinearity, leading to ill-conditioning challenges for prevalent algorithms like iterative least squares (LS) estimation, especially in cases where inappropriate initial values are selected. In this study, we investigate the causes of ill-posed problems from two perspectives. Firstly, we analyze the linearization errors of the Doppler observation equations in relation to satellite orbital altitude and initial value errors, revealing instances where traditional algorithms may fail to converge. Secondly, from an optimization theory perspective, we demonstrate the occurrence of convergence to locally non-unique solutions for Doppler positioning. Subsequently, to address these ill-conditioning issues, we introduce Tikhonov regularization terms in the objective function to constrain algorithm divergence, with a fitted model for the regularization coefficient. Finally, we conduct comprehensive simulation experiments in both dynamic and static scenarios to validate the performance of the proposed algorithm. On the one hand, when the initial values are set to 0, our algorithm achieves high-precision positioning, whereas the iterative LS fails to converge. On the other hand, in certain simulation scenarios, the iterative LS converges to locally non-unique solutions, resulting in positioning errors exceeding 50 km in the north and east directions, several hundred kilometers in the vertical direction, and velocity errors surpassing 120 m/s. In contrast, our algorithm demonstrates typical errors of a position error of 6.8462 m, velocity error of 0.0137 m/s, and clock drift error of 8.3746 × 10−6 s/s.This work provides an effective solution to the sensitivity issue of initial points in Doppler positioning and can serve as a reference for the algorithm design of Doppler positioning receivers with LEO mega-constellations.

Optimizing projectile aerodynamic parameter identification of kernel extreme learning machine based on improved Dung Beetle Optimizer algorithm

Accurately obtaining the aerodynamic parameters of a projectile is crucial for improving aircraft performance, optimizing design and control, and enhancing weapon strike accuracy. In this study, the Dung Beetle Optimization (DBO) algorithm was utilized to optimize the kernel parameters and regularization coefficients of the Kernel Extreme Learning Machine (KELM). To fully exploit the performance of the DBO algorithm and enhance identification accuracy, four improvement measures were proposed for the DBO algorithm. The enhanced DBO algorithm, known as the Improved Dung Beetle Optimization (IDBO) algorithm, was developed and validated using the CEC2017 test function. The proposed IDBO-KELM algorithm was applied to the aerodynamic parameter identification of projectiles. Comparative analysis with results from ELM, IDBO-ELM, and other models revealed that single ELM struggles to accurately identify aerodynamic parameters. While the identification results of IDBO-ELM were 3–5 orders of magnitude higher than those of ELM, a certain gap still existed between the identification results and the true values in the highly nonlinear transonic region. Notably, the IDBO-KELM model yielded the best identification results, with an overall improvement of 1–2 orders of magnitude compared to IDBO-ELM, effectively reducing errors in identification within the transonic region. Experimental findings demonstrate that the proposed algorithm combines the advantages of high accuracy and stability.

Using enhanced Variational Modal Decomposition and Dung Beetle Optimization Algorithm optimization-kernel Extreme Learning Machine model to forecast short-term wind power

With the widespread use of clean energy, the forecasting of wind power has become increasingly significant. Extracting time series from sophisticated wind power data and thus improving the prediction accuracy of wind power generation has become the key to short-term forecasting. This paper proposes a Kernel Extreme Learning Machine (KELM) prediction model based on Crested Porcupine Optimizer (CPO), Variational Modal Decomposition (VMD) with Improved Dung Beetle Optimization Algorithm (QHDBO). Firstly, by analyzing the correlation between the variables in the wind power data, the CPO algorithm is used to optimize the modal decomposition number and the penalty parameter of the VMD and extract the time-series information of the wind power series, and smooth it to obtain a number of sub-sequences with strong regularity. Then, the KELM prediction model is constructed for different subsequences, and the kernel parameters and regularization coefficients of KELM are optimized using the QHDBO algorithm. Finally, the predicted values of different subsequences are reconstructed to obtain the final prediction results. The simulation results show that the CVMD-QHDBO-KELM model can effectively improve the wind power prediction performance, and the MAPE values are all below 9 %, and most of the RE values are below 10 %.

A method for predicting methane production from anaerobic digestion of kitchen waste under small sample conditions.

Anaerobic digestion (AD) has the great potential to treat organic waste and achieve remarkable results effectively. However, it is very tough to establish an accurate mechanistic model for this process. Data-driven modeling technology has opened a new door to solving this problem. While when the sample set is small, traditional data-driven modeling methods are often powerless. In this paper, an effective method is proposed for data-driven high-precision modeling in small sample scenarios. A time series generative adversarial network (TimeGAN) is first utilized to augment the original high-quality small-sample data collected during the AD methane production. A novel hybrid kernel extreme learning machine (HKELM) is then designed to form a better structure of the data-driven model, whose regularization coefficient C0 is optimized by the sparrow search algorithm (SSA). Finally, this semi-finished model (SSA-HKELM) is trained by the augmented data to form the final mathematical model (TimeGAN-SSA-HKELM) for the AD methane generation process. Comparative experiments of the methane daily production prediction error haveverified the effectiveness of the method, which can be extended to other similar small sample data-driven modeling scenarios.

Piecewise regularity results for linear elliptic systems with piecewise regular coefficients

Piecewise regularity results for linear elliptic systems with piecewise regular coefficients

Remaining useful life prediction method of rolling bearings based on improved 3σ and DBO-HKELM

Abstract An improved 3σ method and dung beetle algorithm optimization hybrid kernel extreme learning machine-based (DBO-HKELM) approach for predicting the remaining useful life (RUL) of rolling bearings was suggested in order to increase prediction accuracy. Firstly, multi-dimensional degradation feature data is extracted from bearing vibration data. Considering the influence of noise signal on the prediction accuracy, an improved kernel principal component analysis method is proposed to reduce the noise of degraded features. Then, an improved 3σ method is proposed to determine the starting point of bearing degradation by combining bearing vibration signal data. Lastly, a DBO-HKELM life prediction model was put forth. The parameters of hybrid kernel extreme learning machine were optimized by dung beetle algorithm, and appropriate kernel parameters and regularization coefficient were selected. The feature set of degradation indicators is input into the trained model to output the bearing RUL prediction results starting from the determined degradation starting point. Multiple data sets were used to verify that the new RUL prediction method significantly improves the prediction accuracy.

Iterative robust peak-aware guided filter for signal smoothing

Due to the imperfection of experimental measurements, the measured spectrum signals are inevitably corrupted by random error (noise). It is very useful in applications to remove the noise in the measured data while preserving some significant features (such as peaks) by a smoothing filtering process. This paper presents a peak-preserving smoothing technique in the framework of guided filtering for spectrum signals, called robust peak-aware guided filter (RPAGF). Firstly, the robust peak-aware weight (RPAW) is introduced by means of hyperbolic tangent with threshold parameter, where the parameter is estimated by the median of absolute deviations from the median of local variances of the signal under consideration. Then the RPAGF is presented by incorporating the RPAW of a guidance signal into the cost function of guided filter as the regularization coefficient. The RPAGF uses the Gaussian-weighted averaging to deal with the problem of overlapping windows involved in computing filtering output, while the guidance signal is generated by a time fractional diffusion filtering from the input signal. Finally, the self-iteration scheme of RPAGF called iterative RPAGF (IRPAGF) is designed to provide a more faithful guidance signal for RPAGF at each iteration. The IRPAGF is evaluated by experiments on simulated and real spectrum signals, in comparison to the relevant state-of-the-art methods. Results show that the IRPAGF achieves higher performance in terms of peak-preserving smoothing of spectrum signals.

Prediction of Geometric Dimensions of Deposited Layer Produced Using Laser-Arc Hybrid Additive Manufacturing.

Laser-arc hybrid additive manufacturing (LAHAM) holds substantial potential in industrial applications, yet ensuring dimensional accuracy remains a major challenge. Accurate prediction and effective control of the geometrical dimensions of the deposited layers are crucial for achieving this accuracy. The width and height of the deposited layers, key indicators of geometric dimensions, directly affect the forming precision. This study conducted experiments and in-depth analysis to investigate the influence of various process parameters on these dimensions and proposed a predictive model for accurate forecasting. It was found that the width of the deposited layers was positively correlated with laser power and arc current and negatively correlated with scanning speed, while the height was negatively correlated with laser power and scanning speed and positively with arc current. Quantitative analysis using the Taguchi method revealed that the arc current had the most significant impact on the dimensions of the deposited layers, followed by scanning speed, with laser power having the least effect. A predictive model based on extreme gradient boosting (XGBoost) was developed and optimized using particle swarm optimization (PSO) for tuning the number of leaf nodes, learning rate, and regularization coefficients, resulting in the PSO-XGBoost model. Compared to models enhanced with PSO-optimized support vector regression (SVR) and XGBoost, the PSO-XGBoost model exhibited higher accuracy, the smallest relative error, and performed better in terms of Mean Relative Error (MRE), Mean Square Error (MSE), and Coefficient of Determination R2 metrics. The high predictive accuracy and minimal error variability of the PSO-XGBoost model demonstrate its effectiveness in capturing the complex nonlinear relationships between process parameters and layer dimensions. This study provides valuable insights for controlling the geometric dimensions of the deposited layers in LAHAM.

The Method of Restoring Lost Information from Sensors Based on Auto-Associative Neural Networks

The research aims to develop a neural network-based lost information restoration method when the complex nonlinear technical object (using the example of helicopter turboshaft engines) sensors fail during operation. The basis of the research is an auto-associative neural network (autoencoder), which makes it possible to restore lost information due to the sensor failure with an accuracy of more than 99%. An auto-associative neural network (autoencoder)-modified training method is proposed. It uses regularization coefficients that consist of the loss function to create a more stable and common model. It works well on the training sample of data and can produce good results on new data. Also, it reduces its overtraining risk when it adapts too much to the training data sample and loses its ability to generalize new data. This is especially important for small amounts of data or complex models. It has been determined based on the computational experiment results (the example of the TV3-117 turboshaft engine) that lost information restoration based on an auto-associative neural network provides a data restoring error of no more than 0.45% in the case of single failures and no more than 0.6% in case of double failures of the engine parameter registration sensor event.

Mine water inrush source discrimination model based on KPCA-ISSA-KELM.

In order to ensure the safety of coal mine production, a mine water source identification model is proposed to improve the accuracy of mine water inrush source identification and effectively prevent water inrush accidents based on kernel principal component analysis (KPCA) and improved sparrow search algorithm (ISSA) optimized kernel extreme learning machine (KELM). Taking Zhaogezhuang mine as the research object, firstly, Na+, Ca2+, Mg2+, Cl-, SO2- 4 and HCO- 3 were selected as evaluation indexes, and their correlation was analyzed by SPSS27 software, with reducing the dimension of the original data by KPCA. Secondly, the Sine Chaotic Mapping, dynamic adaptive weights, and Cauchy Variation and Reverse Learning were introduced to improve the Sparrow Search Algorithm (SSA) to strengthen global search ability and stability. Meanwhile, the ISSA was used to optimize the kernel parameters and regularization coefficients in the KELM to establish a mine water inrush source discrimination model based on the KPCA-ISSA-KELM. Then, the mine water source data are input into the model for discrimination in compared with discrimination results of KPCA-SSA-KELM, KPCA-KELM, ISSA-KELM, SSA-KELM and KELM models. The results of the study show as follows: The discrimination results of the KPCA-ISSA-KELM model are in agreement with the actual results. Compared with the other models, the accuracy of the KPCA-ISSA-KELM model is improved by 8.33%, 12.5%, 4.17%, 21.83%, and 25%, respectively. Finally, when these models were applied to discriminate water sources in a coal mine in Shanxi, and the misjudgment rates of each model were 28.57%, 19.05%, 14.29%, 23.81%, 9.52% and 4.76%, respectively. From this, the KPCA-ISSA-KLEM model is the most accurate about discrimination and significantly better than other models in other evaluation indicators, verifying the universality and stability of the model. It can be effectively applied to the discrimination of inrush water sources in mines, providing important guarantees for mine safety production.

Prediction of total volatile basic nitrogen (TVB-N) in fish meal using a metal-oxide semiconductor electronic nose based on the VMD-SSA-LSTM algorithm.

The total volatile basic nitrogen (TVB-N) is the main indicator for evaluating the freshness of fish meal, and accurate detection and monitoring of TVB-N is of great significance for the health of animals and humans. Here, to realize fast and accurate identification of TVB-N, in this article, a self-developed electronic nose (e-nose) was used, and the mapping relationship between the gas sensor response characteristic information and TVB-N value was established to complete the freshness detection. The TVB-N variation curve was decomposed into seven subsequences with different frequency scales by means of variational mode decomposition (VMD). Each subsequence was modelled using different long short-term memory (LSTM) models, and finally, the final TVB-N prediction result was obtained by adding the prediction results based on different frequency components. To improve the performance of the LSTM, the sparrow search algorithm (SSA) was used to optimize the number of hidden units, learning rate and regularization coefficient of LSTM. The prediction results indicated that the high accuracy was obtained by the VMD-LSTM model optimized by SSA in predicting TVB-N. The coefficient of determination (R2), the root-mean-squared error (RMSE) and relative standard deviation (RSD) between the predicted value and the actual value of TVBN were 0.91, 0.115 and 6.39%. This method improves the performance of e-nose in detecting the freshness of fish meal and provides a reference for the quality detection of e-nose in other materials. © 2024 Society of Chemical Industry.

Identification of bridge influence line and multiple-vehicle loads based on physics-informed neural networks

Influence lines (ILs) and vehicle loads identification are critical in the design, health monitoring, and damage detection of bridges. Traditionally, the approach used in most existing literature has been to solve the system of equations directly. However, these approaches require complex calculations such as matrix decomposition and regularization coefficient optimization, making them difficult to implement. In addition, there are difficulties in obtaining accurate axle information and effectively separating the bridge response due to each vehicle. Thus, the improvement of identification algorithms for ILs and multi-vehicle loads remains of significant importance. To address these issues, this paper presents a novel approach that integrates prior physical equations and neural networks. This is achieved by integrating the equation that reflects the relationship between axle loads and bridge response into the neural network, utilizing existing methods for acquiring axle information of vehicles. To validate the effectiveness of the proposed method, it was first applied to theoretical and simulation data. The study then investigated the impact of noise and dynamic effects on the accuracy of the results, as well as the range of the neural network layers and sampling intervals. Finally, the method was implemented for identifying multiple-vehicle loads. The findings of the study confirm the feasibility and numerical stability of the proposed approach. The proposed method eliminates the need for complex computational processes, including matrix decomposition, diagonalization, regularization coefficient optimization, and solution vector smoothing fitting. As a result, the implementation of the algorithm is significantly less challenging, and identification accuracy is improved. It is important to note, however, that the proposed method is relatively more time-consuming due to the iterative learning and training required by the neural network.

Study on creep characteristics and reduction coefficient of geomembrane under top pressure

A new top-pressure creep test device was developed to simulate the stress conditions of geomembrane in a real working state. CBR puncture and top pressure creep tests under different load levels were conducted using this device to study the creep characteristics of the geomembrane at top pressure. The top pressure creep reduction coefficients under different load levels and design lives were calculated and compared with the regular tensile creep reduction coefficients. Quantitative relationships between the creep-reduction coefficient, design life, and failure strain were established to predict the top-pressure creep-reduction coefficient. The results indicated that the top pressure creep reduction coefficient was greater than that of regular tensile creep. The creep reduction coefficient changed minimally with an increase in the design life. The top pressure creep reduction coefficient remained stable with an increase in the top pressure area strain for different design lives. Based on the results, to improve the safety and durability of engineering structures, the top pressure creep reduction coefficient should be used in the design of geosynthetics affected by the top pressure.

Fault Diagnosis of Hydraulic Components Based on Multi-Sensor Information Fusion Using Improved TSO-CNN-BiLSTM.

In order to realize the accurate and reliable fault diagnosis of hydraulic systems, a diagnostic model based on improved tuna swarm optimization (ITSO), optimized convolutional neural networks (CNNs), and bi-directional long short-term memory (BiLSTM) networks is proposed. Firstly, sensor selection is implemented using the random forest algorithm to select useful signals from six kinds of physical or virtual sensors including pressure, temperature, flow rate, vibration, motor power, and motor efficiency coefficient. After that, fused features are extracted by CNN, and then, BiLSTM is applied to learn the forward and backward information contained in the data. The ITSO algorithm is adopted to adaptively optimize the learning rate, regularization coefficient, and node number to obtain the optimal CNN-BiLSTM network. Improved Chebyshev chaotic mapping and the nonlinear reduction strategy are adopted to improve population initialization and individual position updating, further promoting the optimization effect of TSO. The experimental results show that the proposed method can automatically extract fusion features and effectively utilize multi-sensor information. The diagnostic accuracies of the plunger pump, cooler, throttle valve, and accumulator are 99.07%, 99.4%, 98.81%, and 98.51%, respectively. The diagnostic results of noisy data with 0 dB, 5 dB, and 10 dB signal-to-noise ratios (SNRs) show that the ITSO-CNN-BiLSTM model has good robustness to noise interference.

Gait variability, fractal dynamics, and statistical regularity of treadmill and overground walking recorded with a smartphone

BackgroundThe nonlinear variability present during steady-state gait may provide a signature of health and showcase one’s walking adaptability. Although treadmills can capture vast amounts of walking data required for estimating variability within a small space, gait patterns may be misrepresented compared to an overground setting. Smartphones may provide a low-cost and user-friendly estimate of gait patterns among a variety of walking settings. However, no study has investigated differences in gait patterns derived from a smartphone between treadmill walking (TW) and overground walking (OW). Research questionThis study implemented a smartphone accelerometer to compare differences in temporal gait variability and gait dynamics between TW and OW. MethodsSixteen healthy adults (8F; 24.7 ± 3.8 years) visited the laboratory on three separate days and completed three 8-minute OW and three TW trials, at their preferred speed, during each visit. The inter-stride interval was calculated as the time difference between right heel contact events located within the vertical accelerometery signals recorded from a smartphone while placed in participants front right pant pocket during walking trials. The inter-stride interval series was used to calculate stride time standard deviation (SD) and coefficient of variation (COV), statistical persistence (fractal scaling index), and statistical regularity (sample entropy). Two-way analysis of variance compared walking condition and laboratory visits for each measure. ResultsCompared to TW, OW displayed significantly (p < 0.01) greater stride time SD (0.014 s, 0.020 s), COV (1.26 %, 1.82 %), fractal scaling index (0.70, 0.79) and sample entropy (1.43, 1.63). No differences were found between visits for all measures. SignificanceSmartphone-based assessment of gait provides the ability to distinguish between OW and TW conditions, similar to previously established methodologies. Furthermore, smartphones may be a low-cost and user-friendly tool to estimate gait patterns outside the laboratory to improve ecological validity, with implications for free-living monitoring of gait among various populations.

Automatic Segmentation of Bone Marrow Lesions on MRI Using a Deep Learning Method.

Bone marrow lesion (BML) volume is a potential biomarker of knee osteoarthritis (KOA) as it is associated with cartilage degeneration and pain. However, segmenting and quantifying the BML volume is challenging due to the small size, low contrast, and various positions where the BML may occur. It is also time-consuming to delineate BMLs manually. In this paper, we proposed a fully automatic segmentation method for BMLs without requiring human intervention. The model takes intermediate weighted fat-suppressed (IWFS) magnetic resonance (MR) images as input, and the output BML masks are evaluated using both regular 2D Dice similarity coefficient (DSC) of the slice-level area metric and 3D DSC of the subject-level volume metric. On a dataset with 300 subjects, each subject has a sequence of 36 IWFS MR images approximately. We randomly separated the dataset into training, validation, and testing sets with a 70%/15%/15% split at the subject level. Since not every subject or image has a BML, we excluded the images without a BML in each subset. The ground truth of the BML was labeled by trained medical staff using a semi-automatic tool. Compared with the ground truth, the proposed segmentation method achieved a Pearson's correlation coefficient of 0.98 between the manually measured volumes and automatically segmented volumes, a 2D DSC of 0.68, and a 3D DSC of 0.60 on the testing set. Although the DSC result is not high, the high correlation of 0.98 indicates that the automatically measured BML volume is strongly correlated with the manually measured BML volume, which shows the potential to use the proposed method as an automatic measurement tool for the BML biomarker to facilitate the assessment of knee OA progression.

Automated Sparse and Low-Rank Shallow Autoencoders for Recommendation

Collaborative filtering (CF) works have demonstrated the robust capabilities of Shallow Autoencoders on implicit feedback, showcasing highly competitive performance with other reasonable approaches (e.g., iALS and VAE-CF). However, despite their dual advantages of high performance and simple construction, EASE still exhibits several major shortcomings that must be addressed. To be more precise, the scalability of EASE is limited by the number of items, which determines the storage and inversion cost of a large dense matrix; the square-loss optimization objective does not consistently meet the recommendation task’s requirement for predicting personalized rankings, resulting in sub-optimal outcomes; the regularization coefficients are sensitive and require re-calibration with different datasets, leading to an exhaustive and time-consuming fine-tuning process. In order to address these obstacles, we propose a novel approach called Similarity-Structure Aware Shallow Autoencoder (AutoS 2 AE) that aims to enhance both recommendation accuracy and model efficiency. Our method introduces three similarity structures: Co-Occurrence, KNN, and NSW graphs, which replace the large dense matrix in EASE with a sparse structure, thus facilitating model compression. Additionally, we optimize the model by incorporating a low-rank training component into the matrix and applying a weighted square loss for improved ranking-oriented approximations. To automatically tune the hyperparameters, we further design two validation losses on the validation set for guidance and update the hyperparameters using the gradients of these validation losses. Both theoretical analyses regarding the introduction of similarity structures and empirical evaluations on multiple real-world datasets demonstrate the effectiveness of our proposed method, which significantly outperforms competing baselines.

A Hybrid Soft Sensor Model for Measuring the Oxygen Content in Boiler Flue Gas.

As an indispensable component of coal-fired power plants, boilers play a crucial role in converting water into high-pressure steam. The oxygen content in the flue gas is a crucial indicator, which indicates the state of combustion within the boiler. The oxygen content not only affects the thermal efficiency of the boiler and the energy utilization of the generator unit, but also has adverse impacts on the environment. Therefore, accurate measurement of the flue gas's oxygen content is of paramount importance in enhancing the energy utilization efficiency of coal-fired power plants and reducing the emissions of waste gas and pollutants. This study proposes a prediction model for the oxygen content in the flue gas that combines the whale optimization algorithm (WOA) and long short-term memory (LSTM) networks. Among them, the whale optimization algorithm (WOA) was used to optimize the learning rate, the number of hidden layers, and the regularization coefficients of the long short-term memory (LSTM). The data used in this study were obtained from a 350 MW power generation unit in a coal-fired power plant to validate the practicality and effectiveness of the proposed hybrid model. The simulation results demonstrated that the whale optimization algorithm-long short-term memory (WOA-LSTM) model achieved an MAE of 0.16493, an RMSE of 0.12712, an MAPE of 2.2254%, and an R2 value of 0.98664. The whale optimization algorithm-long short-term memory (WOA-LSTM) model demonstrated enhancements in accuracy compared with the least squares support vector machine (LSSVM), long short-term memory (LSTM), particle swarm optimization-least squares support vector machine (PSO-LSSVM), and particle swarm optimization-long short-term memory (PSO-LSTM), with improvements of 4.93%, 4.03%, 1.35%, and 0.49%, respectively. These results indicated that the proposed soft sensor model exhibited more accurate performance, which can meet practical requirements of coal-fired power plants.