Faster Convergence Speed Research Articles

Deep learning for geological mapping in the overburden area

This paper aims to achieve bedrock geologic mapping in the overburden area using big data, distributed computing, and deep learning techniques. First, the satellite Bouguer gravity anomaly with a resolution of 2′×2′ in the range of E66°-E96°, N40°-N55° and 1:5000000 Asia-European geological map are used to design a dataset for bedrock prediction. Then, starting from the gravity anomaly formula in the spherical coordinate system, we deduce the non-linear functional between rock density ρ and rock mineral composition m, content p, buried depth h, diagenesis time t and other variables. We analyze the feasibility of using deep neural network to approximate the above nonlinear generalization. The problem of solving deep neural network parameters is transformed into a non-convex optimization problem. We give an iterative, gradient descent-based solution algorithm for the non-convex optimization problem. Utilizing neural architecture search (NAS) and human-designed approach, we propose a geological-geophysical mapping network (GGMNet). The dataset for the network consists of both gravity anomaly and a priori geological information. The network has fast convergence speed and stable iteration during the training process. It also has better performance than a single neural network search or human-designed architectures, with the mean pixel accuracy (MAP) = 63.1% and the frequency weighted intersection over union (FWIoU) = 42.88. Finally, the GGMNet is used to predict the rock distribution of the Junggar Basin.

Research on Surveying and Mapping Data Processing Based on Nonlinear Mathematical Models and Deep Learning Optimization

In order to deeply understand the mapping data processing of nonlinear mathematical model optimization, this paper uses nonlinear model optimization theory to process mapping data. When the precision of parameter approximation is high, the calculation results of each algorithm are the same, and the iteration times of the fastest descent method and simplex method are significantly increased compared with the other three algorithms. Therefore x01 = 5.42 in the parameter is kept unchanged near the truth value, and the convergence range of Newton method, the fastest descent method, conjugate gradient method and simplex method away from the truth value of parameter x02 = −0.25 is investigated. When the approximate value of undetermined parameters is high, the results of Newton algorithm and simplex algorithm are consistent, but the Newton iterative algorithm has faster convergence speed and higher computational efficiency than the simplex algorithm. Lower value when the undetermined parameter approximation precision, namely the undetermined parameter approximation and its true value is far off, may complete failure type Newton iteration algorithm, the simplex algorithm can be a supplement of the Newton method for the most part, the simplex method significantly reduces the requirements for initial value of parameters, data calculation efficiency has improved significantly.

LMFLS: A new fast local multi-factor node scoring and label selection-based algorithm for community detection

Community detection is still regarded as one of the most applicable approaches for discovering latent information in complex networks. To meet the needs of processing large networks in today's world, it is important to propose fast methods that have low execution time and fast convergence speed, while maintaining algorithmic accuracy. To overcome these issues, a fast local multi-factor node scoring and label selection-based (LMFLS) method with low time complexity and fast convergence is proposed. Node scoring step incorporates diverse metrics to better assess impact of nodes from different aspects and obtain more meaningful order of nodes. In second step, to construct and stabilize initial structure of communities, an efficient label assignment technique based on the selection of the most similar neighbor is suggested. Moreover, two label selection strategies are proposed to significantly enhance the accuracy and improve convergence of the algorithm. During the label selection step, each node in graph tends to choose the most appropriate label based on a multi-criteria label influence from its surrounding nodes. Finally, by utilizing a novel merge method, small group of nodes are merged to form the final communities. Meanwhile, since drug repositioning is one of the popular research fields in therapeutics, to extend the application of the proposed algorithm in practical context, the LMFLS algorithm is applied on Drug-Drug network to find potential repositioning for drugs. Thorough experiments are conducted on both actual real-world networks and synthetic networks to assess the algorithm's performance and accuracy. The findings demonstrate that the proposed method outperforms state-of-the-art algorithms in terms of both accuracy and execution time.

Spatio-temporal deep recurrent convolutional neural network for infrared focal plane arrays non-uniformity correction

The inherent non-uniformity generated in infrared focal plane arrays (IRFPA) due to inconsistent responses of detectors, and it significantly degrades the performance of infrared imaging systems. Traditional Non-Uniformity Correction (NUC) methods mainly leverage the temporal low-frequency property of non-uniformity to remove it. In recent years, Convolutional Neural Network (CNN) has gained attention as a powerful tool for conducting NUC using solely spatial information from a single frame. In this study, we propose a deep neural network with a recurrent spatio-temporal feature fusion module to enhance CNN performance using inter-frame information. Initially, the proposed network uses CNN to coarsely extract FPN features solely from the spatial domain. Subsequently, the network merges it with the historical estimation under the guidance of two gating signals, which calculated from spatio-temporal information with gated convolutional layers. Both stages are jointly trained to ensure that the entire network operates end-to-end and adapts to different scenarios, eliminating complex parameter tuning efforts. Experiments demonstrated that the proposed network shows both the advantages of traditional and CNN-based NUC methods. It has a higher convergence starting point, faster convergence speed, and a 1 ∼ 2 dB improvement in PSNR compared to both types of methods.

A Study on An MPPT Control Approach Using Artificial Intelligence and the Perturb and Observe Method

A maximum power point tracking (MPPT) control procedure constructed based on the Artificial intelligence optimization algorithm is proposed. A mathematical model of photovoltaic cells was established under varying light intensity and ambient temperature. Maximal power tracking control for iterative search using an artificial intelligence (AI) optimization technique Excellent, the artificial intelligence-based MPPT algorithm's principle is presented. Simulation results it shows that the artificial intelligence algorithm can faster and exactly track the maximum power point and remains stable, and compared to Perturb and Observe algorithm under dynamic shadow conditions and artificial intelligence MPPT control method, which has higher tracking accuracy and faster convergence speed. Faster and smaller oscillation amplitude, which is the best response to sunlight conditions for photovoltaic maximum power point tracking technology Flexibility. Using Matlab/Simulink software to build a simulation model of an independent photovoltaic system. It controls variables by keeping the temperature continuous and changes the light intensity to simulate different lighting environments. The identical comparison was conducted between all based MPPT methods such as the fuzzy control approach, demonstrating that the fuzzy control based technique exhibits higher results in terms of efficiency.

Stochastic configuration networks with improved supervisory mechanism

The stochastic configuration networks (SCNs) have been shown to have great potential for developing fast learning models, making them especially suitable for industrial devices. To achieve low consumption during the SCNs training process, this paper proposes an improved version of SCNs (ISCN-III). First, guided by the function dictionary set and the objective function space, a dynamic hyperparameter is proposed to replace the nonnegative sequence of the supervisory mechanism. Second, based on the dynamic hyperparameters and function dictionary set, an improved supervisory mechanism is proposed to further reduce computational consumption. Finally, the improved supervisory mechanism with the dynamic hyperparameter proves the universal approximation property of ISCN-III. ISCN-III is compared with SCNs and its improved versions using six benchmark datasets and subsequently applied to a grinding process and a continuous stirred tank reactor. The experimental results prove that ISCN-III has the advantages of low computational cost and fast convergence speed.

Implementation of an Enhanced Crayfish Optimization Algorithm.

This paper presents an enhanced crayfish optimization algorithm (ECOA). The ECOA includes four improvement strategies. Firstly, the Halton sequence was used to improve the population initialization of the crayfish optimization algorithm. Furthermore, the quasi opposition-based learning strategy is introduced to generate the opposite solution of the population, increasing the algorithm's searching ability. Thirdly, the elite factor guides the predation stage to avoid blindness in this stage. Finally, the fish aggregation device effect is introduced to increase the ability of the algorithm to jump out of the local optimal. This paper performed tests on the widely used IEEE CEC2019 test function set to verify the validity of the proposed ECOA method. The experimental results show that the proposed ECOA has a faster convergence speed, greater performance stability, and a stronger ability to jump out of local optimal compared with other popular algorithms. Finally, the ECOA was applied to two real-world engineering optimization problems, verifying its ability to solve practical optimization problems and its superiority compared to other algorithms.

An online battery–state of charge estimation method using the varying forgetting factor recursive least square - unscented Kalman filter algorithm on electric vehicles

Accurate and fast estimation of the state of charge is important for the battery management system of electric vehicles. This paper proposes a method to estimate the state of charge of Lithium-ion batteries by the <span lang="EN-US">variable</span> forgetting factor recursive least square (VFFRLS) – unscented Kalman filter (UKF) algorithm in real-time without the off-line battery testing data. Since the state observation requires an accurate model, an equivalent circuit model was constructed. Then, the VFFRLS algorithm is used to identify online the battery model parameters based on voltage and current measurements. An advantage of this algorithm is that it requires less initial information and shorter identification time than offline parameter identification. After the model parameters are well identified, the unscented Kalman filter estimates the state of charge and minimizes noise characteristics and uncertainty in the parameter identification process. The VFFRLS algorithm applied in this paper has shown a good result with the model output error of less than 1%, and the identification achieves real-time response. The state of charge obtained by the UKF algorithm has shown satisfactory estimation results with fast convergence speed and small errors. The UKF filter provides the results with a 1.5% error from the reference and converges after 10 cycles.

Optimization and scheduling scheme of park-integrated energy system based on multi-objective Beluga Whale Algorithm

To improve the consumption rate of renewable energy and ensure the stability of power and heat supply within the industrial park, a park-integrated energy system (PIES), which incorporates electric heating equipment, energy storage devices, and multiple micro sources, is established. Based on the established PIES, a price-based demand response (PDR) mechanism is introduced to establish a multi-objective optimization and scheduling model with two time scales: day-ahead scheduling and intra-day scheduling, aiming to minimize the impact on the main grid and ensure the flexibility of system operation and real-time scheduling. With the objectives of minimizing operating costs and maximizing environmental benefits for PIES, a hybrid algorithm named Non-Dominated Sorting Beluga Whale Optimization (NSBWO), which combines the Non-Dominated Sorting Genetic Algorithm II (NSGA-II) and the Beluga Whale Optimization (BWO) algorithm is proposed to solve the optimization problem of the model. Simultaneously, an adaptive penalty function is introduced in the algorithm to handle constraint problems in optimization and additional penalty term for wind and solar curtailment is included within the constraints to improve the consumption rate of sustainable energy. The proposed NSBWO algorithm exhibits more robust search capabilities and faster convergence speed than the NSGA-II and MOPSO algorithm. Simulation results show that the proposed NSBWO algorithm can achieve low-carbon economic scheduling of the system in both long-term and short-term time scales and improve renewable energy consumption. Due to its flexibility, intra-day scheduling has lower operating costs than day-ahead scheduling.

Enhancing graph convolutional networks with progressive granular ball sampling fusion: A novel approach to efficient and accurate GCN training

Graph convolutional network (GCN) has gained considerable attention and has been widely utilized in graph data analytics. However, training large GCNs presents considerable challenges owing to the inherent complexity of graph-structured data. Previous training algorithms frequently struggle with slow convergence speed caused by full-batch gradient descent on entire graphs and reduced model performance due to inappropriate node sampling methods. To address these issues, we propose a novel framework called Progressive Granular Ball Sampling Fusion (PGBSF). PGBSF leverages granular ball sampling to partition the original graph into a collection of subgraphs, thereby enhancing both training efficiency and detail capture. Then, it applies a progressive approach accompanied by a parameter-sharing strategy for incremental GCN model training, which results in robust performance and rapid convergence speed. This simple yet effective strategy considerably enhances classification accuracy and memory efficiency. The experiment results show that our proposed architecture consistently outperforms other baseline models in terms of accuracy across almost all datasets with different label rates. In addition, PGBSF improves GCN performance significantly on large and complex datasets. Moreover, GCN+PGBSF reduces time complexity by training on subgraphs and achieves the fastest convergence speed among all models, with a relatively small variance in loss during training.

How to Improve Representation Alignment and Uniformity in Graph-Based Collaborative Filtering?

Collaborative filtering (CF) is a prevalent technique utilized in recommender systems (RSs), and has been extensively deployed in various real-world applications. A recent study in CF focuses on improving the quality of representations from the perspective of alignment and uniformity on the hyperspheres for enhanced recommendation performance. It promotes alignment to increase the similarity between representations of interacting users and items, and enhances uniformity to have more uniformly distributed user and item representations within their respective hyperspheres. However, although alignment and uniformity are enforced by two different optimized objectives, respectively, they jointly constitute the supervised signals for model training. Models trained with only supervised signals in labeled data can inevitably overfit the noise introduced by label sampling variance, even with i.i.d. datasets. This overfitting to noise further compromises the model's generalizability and performance on unseen testing data. To address this issue, in this study, we aim to mitigate the effect caused by the sampling variance in labeled training data to improve representation generalizability from the perspective of alignment and uniformity. Representations with more generalized alignment and uniformity further lead to improved model performance on testing data. Specifically, we model the data as a user-item interaction bipartite graph, and apply a graph neural network (GNN) to learn the user and item representations. This graph modeling approach allows us to integrate self-supervised signals into the RS, by performing self-supervised contrastive learning on the user and item representations from the perspective of label-irrelevant alignment and uniformity. Since the representations are less dependent on label supervision, they can capture more label-irrelevant data structures and patterns, leading to more generalized alignment and uniformity. We conduct extensive experiments on three benchmark datasets to demonstrate the superiority of our framework (i.e., improved performance and faster convergence speed). Our codes: https://github.com/zyouyang/AUPlus

Application of Improved Ant Colony Algorithm in Optimizing the Charging Path of Electric Vehicles

In current traffic congestion scenarios, electric vehicles (EVs) have the problem of reduced battery life and continuous decline in endurance. Therefore, this study proposes an optimization method for electric vehicle charging scheduling based onthe ant colony optimization algorithm with adaptive dynamic search (ADS-ACO), and conducts experimental verification on it. The experiment revealed that in the four benchmark functions, the research algorithm has the fastest convergence speed and can achieve convergence in most of them. In the validation of effectiveness, the optimal solution for vehicle time consumption under the ADS-ACO algorithm in the output of the algorithm with a stationary period and a remaining battery energy of 15 kW·h was 2.146 h in the regular road network. In the initial results of 15 kW·h under changes in road conditions from peak to peak periods, the total energy consumption of vehicles under the research algorithm was 4.678 kW·h and 4.656 kW·h under regular and irregular road networks, respectively. The change results were 4.509 kW·h and 4.656 kW·h, respectively. The initial results of 10 kW·h were 4.755 kW·h and 4.873 kW·h, respectively. The change results were 4.461 kW·h and 4.656 kW·h, respectively, which are lower than the comparison algorithm. In stability verification, research algorithms can find the optimal path under any conditions. The algorithm proposed in the study has been demonstrated to be highly effective and stable in electric vehicle charging path planning. It represents a novel solution for electric vehicle charging management and is expected to significantly enhance the range of electric vehicles in practical applications.

Dynamic Gaussian mutation beetle swarm optimization method for large-scale weapon target assignment problems

The weapon target assignment is a crucial issue for firepower resources optimization in modern warfare. Such a problem is complicated, multi-constrained, strongly nonlinear, NP-complete and the existing studies did not consider the suitability between different weapons and targets. In this paper, a novel weapon target assignment model is established that involves the weapon-target suitability and is closer to the real combat scenarios. Then, in view of that the conventional weapon target assignment methods are difficult to be applied in the large-scale problems efficiently, this work proposes a dynamic Gaussian mutation beetle swarm optimization algorithm with rule-based chaotic initialization. With the assistance of the dynamic parameter adjustment strategies and Gaussian mutation, the improved algorithm has fast convergence speed and high convergence accuracy, and it can solve the weapon target assignment problems with excellent optimization capabilities. Besides, the rule-based chaotic initialization strategy is embedded in this algorithm to generate high-quality population with better diversity. Finally, two comparative simulation cases of different initialization methods and algorithms for solving the large-scale weapon target assignment problems are designed. The results demonstrate that the proposed approach can provide more superior assignment schemes than its competitors with enhanced efficiency.

Optimization on the peroxide ratio control strategy of PEMFC system based on immune algorithm

The air supply system is an essential component of a proton exchange membrane fuel cell (PEMFC), and proper control of the oxygen stoichiometry ratio(λo2) is crucial to prevent oxygen starvation in the PEMFC while maintaining efficient and stable performance output. In response to the issues of poor dynamic performance and slow response in traditional λo2 control strategies, this study was conducted to establish a dynamic oxygen flow model of the PEMFC and proposed a λo2 control strategy based on an immune algorithm-fuzzy PID (IAFPID). The performance of the IAFPID control strategy in λo2 control was analyzed under different load requirements, simulated vehicle dynamic loads, and temperature disturbances. The results demonstrate that the IAFPID control strategy has advantages such as fast convergence speed, strong robustness, and good dynamic performance. The efficiency of achieving the optimal stoichiometry ratio for high power output in PEMFC under disturbances was nearly 50% higher with IAFPID control compared to traditional PID, Fuzzy-PID, Genetic Algorithm PID, and Immune Algorithm PID strategies. Furthermore, this method ensures a lower average error in maintaining stable λo2 under dynamic disturbances compared to the IAPID and PID control strategies, with reductions of 12.93% and 37.75%, respectively.

Special Trans Function based exact expressions for the double and triple diode models of solar cells: Superior fitness, accuracy and convergence

Inspired by the fact that the Special Trans Function (STF) is more accurate than Lambert W function (LWF), this paper rigorously derives two types of STF-based exact expressions for the double-diode model (DDM) and triple-diode model (TDM) of solar cells. The former involves multi-STF (mSTF), while the latter contains single-STF (sSTF), bridging the remaining gap in describing the nonlinear I–V characteristic of solar cells using STF. The proposed mSTF and sSTF are closely linked with but different from the multi-Lambert W function (mLWF), single-LWF (sLWF) and implicit exponential function (IEF) based expressions, particularly in terms of fitness and parameter extraction. Through a fitness test involving 203 cases in the space of STF branch x∈R+, it was found that under the same parameter values, the proposed mSTF consistently outperforms in fitting the measured I–V data of various solar cell/modules, followed by the proposed sSTF, mLWF, sLWF, and finally IEF-based expression. Notably, all optimal STF branch x fall within the interval of x∈[0,6] rather than the preset x∈[0,20]. A normalized trust-region-reflective (NTRR) algorithm and a V-shaped selection strategy are developed to improve the parameter extraction of proposed mSTF and sSTF. Results from the derivative-dependent V-shaped NTRR algorithm and population-based Rcr-IJADE algorithm indicate that the proposed sSTF consistently achieves the highest accuracy and fastest convergence speed, followed by the mSTF, sLWF, mLWF and IEF-based expression. Moreover, the proposed sSTF exhibits fewer degeneracies from TDM to DDM compared to other expressions. Given these advantages, the proposed mSTF and sSTF show great promise for PV simulation and therefore deserve serious attention.

Enhancing quality and speed in database-free neural network reconstructions of undersampled MRI with SCAMPI.

We present SCAMPI (Sparsity Constrained Application of deep Magnetic resonance Priors for Image reconstruction), an untrained deep Neural Network for MRI reconstruction without previous training on datasets. It expands the Deep Image Prior approach with a multidomain, sparsity-enforcing loss function to achieve higher image quality at a faster convergence speed than previously reported methods. Two-dimensional MRI data from the FastMRI dataset with Cartesian undersampling in phase-encoding direction were reconstructed for different acceleration rates for single coil and multicoil data. The performance of our architecture was compared to state-of-the-art Compressed Sensing methods and ConvDecoder, another untrained Neural Network for two-dimensional MRI reconstruction. SCAMPI outperforms these by better reducing undersampling artifacts and yielding lower error metrics in multicoil imaging. In comparison to ConvDecoder, the U-Net architecture combined with an elaborated loss-function allows for much faster convergence at higher image quality. SCAMPI can reconstruct multicoil data without explicit knowledge of coil sensitivity profiles. Moreover, it is a novel tool for reconstructing undersampled single coil k-space data. Our approach avoids overfitting to dataset features, that can occur in Neural Networks trained on databases, because the network parameters are tuned only on the reconstruction data. It allows better results and faster reconstruction than the baseline untrained Neural Network approach.

Detection and classification of wilting status in leaf images based on VGG16 with EfficientNet V3 algorithm

The aim of this paper is to explore the importance of leaf wilting status detection and classification in agriculture to meet the demand for monitoring and diagnosing plant growth conditions. By comparing the performance of the traditional VGG16 image classification algorithm and the popular EfficientNet V3 algorithm in leaf image wilting status detection and classification, it is found that EfficientNet V3 has faster convergence speed and higher accuracy. As the model training process proceeds, both algorithms show a trend of gradual convergence of Loss and Accuracy and increasing accuracy. The best training results show that VGG16 reaches a minimum loss of 0.288 and a maximum accuracy of 96% at the 19th epoch, while EfficientNet V3 reaches a minimum loss of 0.331 and a maximum accuracy of 97.5% at the 20th epoch. These findings reveal that EfficientNet V3 has a better performance in leaf wilting status detection, which provides a more accurate and efficient means of plant health monitoring for agricultural production and is of great research significance.

Multi-Strategy Improved Dung Beetle Optimization Algorithm and Its Applications.

The dung beetle optimization (DBO) algorithm, a swarm intelligence-based metaheuristic, is renowned for its robust optimization capability and fast convergence speed. However, it also suffers from low population diversity, susceptibility to local optima solutions, and unsatisfactory convergence speed when facing complex optimization problems. In response, this paper proposes the multi-strategy improved dung beetle optimization algorithm (MDBO). The core improvements include using Latin hypercube sampling for better population initialization and the introduction of a novel differential variation strategy, termed "Mean Differential Variation", to enhance the algorithm's ability to evade local optima. Moreover, a strategy combining lens imaging reverse learning and dimension-by-dimension optimization was proposed and applied to the current optimal solution. Through comprehensive performance testing on standard benchmark functions from CEC2017 and CEC2020, MDBO demonstrates superior performance in terms of optimization accuracy, stability, and convergence speed compared with other classical metaheuristic optimization algorithms. Additionally, the efficacy of MDBO in addressing complex real-world engineering problems is validated through three representative engineering application scenarios namely extension/compression spring design problems, reducer design problems, and welded beam design problems.

Accelerating regularized tensor decomposition using the alternating direction method of multipliers with multiple Nesterov’s extrapolations

Tensor decomposition is an essential tool for multiway signal processing. At present, large-scale high-order tensor data require fast and efficient decomposing algorithms. In this paper, we propose accelerated regularized tensor decomposition algorithms using the alternating direction method of multipliers with multiple Nesterov’s extrapolations in the block coordinate descent framework. We implement the acceleration in three cases: only in the inner loop, only in the outer loop, and in both the inner and outer loops. Adaptive safeguard strategies are developed following the acceleration to guarantee monotonic convergence. Afterwards, we utilize the proposed algorithms to accelerate two types of conventional decomposition: nonnegative CANDECOMP/PARAFAC (NCP) and sparse CANDECOMP/PARAFAC (SCP). The experimental results on synthetic and real-world tensors demonstrate that the proposed algorithms achieve significant acceleration effects and outperform state-of-the-art algorithms. The accelerated algorithm with extrapolations in both the inner and outer loops has the fastest convergence speed and takes almost one-third of the running time of typical algorithms.

Deep unsupervised learning using spike-timing-dependent plasticity

Spike-timing-dependent plasticity (STDP) is an unsupervised learning mechanism for spiking neural networks that has received significant attention from the neuromorphic hardware community. However, scaling such local learning techniques to deeper networks and large-scale tasks has remained elusive. In this work, we investigate a Deep-STDP framework where a rate-based convolutional network, that can be deployed in a neuromorphic setting, is trained in tandem with pseudo-labels generated by the STDP clustering process on the network outputs. We achieve 24.56% higher accuracy and 3.5 × faster convergence speed at iso-accuracy on a 10-class subset of the Tiny ImageNet dataset in contrast to a k-means clustering approach.