Adaptive Weighting Scheme Research Articles

Real-time infrared and visible image fusion network using adaptive pixel weighting strategy

Efficiently fusing infrared and visible images is critical for their real-time application in various electro-optic systems. In recent years, deep learning-based methods have significantly improved fusion quality by leveraging feature-based fusion. However, large-scale networks are usually time and resource-consuming. Although current lightweight networks have improved efficiency by tailoring network size, image fusion quality and processing speed are still unsatisfactory especially when deployed on embedded systems. In this paper, we propose to use pixel-based fusion strategy to design simple yet efficient network that learns pixel-by-pixel weights adaptively for image fusion. Based on the proposed fusion network, we further combine it with a detection model to construct a joint optimization framework which optimizes low-level and high-level tasks cooperatively. The fusion quality and processing speed of various infrared and visible image fusion networks are evaluated on multiple datasets and platforms. The evaluation results demonstrate that, the proposed method yields comparable or even better quantitative and qualitative results than current state-of-the-art large-scale and lightweight networks. Besides, the joint optimization is effective for producing better fusion quality and detection accuracy. More importantly, the proposed fusion network only requires ∼27 ms to fuse a pair of infrared and visible images with the resolution of 512 × 512 on Jeston Xavier NX, which is only a third of the time required by the previous fastest network. Therefore, the proposed method is an efficient solution for real-time infrared and visible image fusion on embedded systems.

Adaptive decoder-block selection with filter reweighting for medical image segmentation

Image segmentation models are mostly designed for specific tasks and trained on large-scale datasets. Medical image segmentation tasks, however, are highly diverse and usually have a small training dataset, making it difficult for a segmentation model to generalize well on other tasks. This architecture bias motivates a lot of research on designing a task-adaptive segmentation network with the most suitable settings including the network’s depth and width and feature fusion strategy. In this paper, we propose the adaptive Decoder-block Selection with Filter Reweighting (DeSFiR) algorithm to generate an efficient and effective medical image segmentation model for each specific dataset. This algorithm is a two-step iterative process. In the adaptive decoder-block selection step, we develop an auxiliary indicator function layer to bypass the decoder block whose individual responses are lower than a pre-defined threshold, aiming to generate a lightweight model. In the filter reweighing step, we replace invalid layers in the segmentation network with the external information from a simultaneously trained twin network. To this end, we utilize an entropy-based criterion to measure the informativeness of each layer and an adaptive weighting strategy to balance the replaced information between two networks. We have evaluated our algorithm on three tasks, including the segmentation of neural structures, nuclei, and liver. Our results indicate that the medical image segmentation model generated by the proposed DeSFiR algorithm has a significantly reduced complexity and comparable accuracy.

State of Charge Estimation for Batteries Based on Common Feature Extraction and Transfer Learning

The state of charge (SOC) of a battery is a key parameter of electrical vehicles (EVs). However, limited by the lack of computing resources, the SOC estimation strategy used in vehicle-mounted battery management systems (V-BMS) is usually simplified. With the development of the new energy vehicle big data platforms, it is possible to obtain the battery SOC through cloud-based BMS (C-BMS). In this paper, a battery SOC estimation method based on common feature extraction and transfer learning is proposed for C-BMS applications. Considering the diversity of driving cycles, a common feature extraction method combining empirical mode decomposition (EMD) and a compensation strategy for C-BMS is designed. The selected features are treated as the new inputs of the SOC estimation model to improve the generalization ability. Subsequently, a long short-term memory (LSTM) recurrent neural network is used to construct a basic model for battery SOC estimation. A parameter-based transfer learning method and an adaptive weighting strategy are used to obtain the C-BMS battery SOC estimation model. Finally, the SOC estimation method is validated on laboratory datasets and cloud platform datasets. The maximum root-mean-square error (RMSE) of battery SOC estimation with the laboratory dataset is 2.2%. The maximum RMSE of battery pack SOC estimation on two different electric vehicles is 1.3%.

Traffic Speed Sequence Prediction by Adaptive Weighted Long Short-Term Memory With Classification-Type Loss

Accurate traffic speed prediction is necessary to promote the development of intelligent transportation systems. The construction of consummate models is challenging owing to nonlinearity, nonstationarity, and long-term dependence in traffic speed prediction. This study proposed an ensemble long short-term memory (LSTM) model that was based on adaptive weighting, in which ensemble learning was the main solution. First, a data preprocessing model based on a seasonal statistical model was introduced to reconcile the long- and short-term dependence of the data. Second, the LSTM time step was considered during training, and a classification-type loss was designed to calculate the error rate in the ensemble system. Last, an adaptive weighting strategy was constructed to integrate a series of LSTMs generated in the system to obtain an ensemble model for traffic speed prediction. The experimental results showed that the proposed method was more stable and accurate than individual methods and existing ensemble learning methods.

Study on DNA Storage Encoding Based IAOA under Innovation Constraints.

With the informationization of social processes, the amount of related data has greatly increased, making traditional storage media unable to meet the current requirements for data storage. Due to its advantages of a high storage capacity and persistence, deoxyribonucleic acid (DNA) has been considered the most prospective storage media to solve the data storage problem. Synthesis is an important process for DNA storage, and low-quality DNA coding can increase errors during sequencing, which can affect the storage efficiency. To reduce errors caused by the poor stability of DNA sequences during storage, this paper proposes a method that uses the double-matching and error-pairing constraints to improve the quality of the DNA coding set. First, the double-matching and error-pairing constraints are defined to solve problems of sequences with self-complementary reactions in the solution that are prone to mismatch at the 3' end. In addition, two strategies are introduced in the arithmetic optimization algorithm, including a random perturbation of the elementary function and a double adaptive weighting strategy. An improved arithmetic optimization algorithm (IAOA) is proposed to construct DNA coding sets. The experimental results of the IAOA on 13 benchmark functions show a significant improvement in its exploration and development capabilities over the existing algorithms. Moreover, the IAOA is used in the DNA encoding design under both traditional and new constraints. The DNA coding sets are tested to estimate their quality regarding the number of hairpins and melting temperature. The DNA storage coding sets constructed in this study are improved by 77.7% at the lower boundary compared to existing algorithms. The DNA sequences in the storage sets show a reduction of 9.7-84.1% in the melting temperature variance, and the hairpin structure ratio is reduced by 2.1-80%. The results indicate that the stability of the DNA coding sets is improved under the two proposed constraints compared to traditional constraints.

Controllable Image Caption Based on Adaptive Weight and Optimization Strategy

Controllable Image Caption Based on Adaptive Weight and Optimization Strategy

A Novel Decomposition-Based Multi-Objective Evolutionary Algorithm with Dual-Population and Adaptive Weight Strategy

Multi-objective evolutionary algorithms mainly include the methods based on the Pareto dominance relationship and the methods based on decomposition. The method based on Pareto dominance relationship will produce a large number of non-dominated individuals with the increase in population size or the number of objectives, resulting in the degradation of algorithm performance. Although the method based on decomposition is not limited by the number of objectives, it does not perform well on the complex Pareto front due to the fixed setting of the weight vector. In this paper, we combined these two different approaches and proposed a Multi-Objective Evolutionary Algorithm based on Decomposition with Dual-Population and Adaptive Weight strategy (MOEA/D-DPAW). The weight vector adaptive adjustment strategy is used to periodically change the weight vector in the evolution process, and the information interaction between the two populations is used to enhance the neighborhood exploration mechanism and to improve the local search ability of the algorithm. The experimental results on 22 standard test problems such as ZDT, UF, and DTLZ show that the algorithm proposed in this paper has a better performance than the mainstream multi-objective evolutionary algorithms in recent years, in solving two-objective and three-objective optimization problems.

BSRWPSO-FKNN: A boosted PSO with fuzzy K-nearest neighbor classifier for predicting atopic dermatitis disease.

Atopic dermatitis (AD) is an allergic disease with extreme itching that bothers patients. However, diagnosing AD depends on clinicians' subjective judgment, which may be missed or misdiagnosed sometimes. This paper establishes a medical prediction model for the first time on the basis of the enhanced particle swarm optimization (SRWPSO) algorithm and the fuzzy K-nearest neighbor (FKNN), called bSRWPSO-FKNN, which is practiced on a dataset related to patients with AD. In SRWPSO, the Sobol sequence is introduced into particle swarm optimization (PSO) to make the particle distribution of the initial population more uniform, thus improving the population's diversity and traversal. At the same time, this study also adds a random replacement strategy and adaptive weight strategy to the population updating process of PSO to overcome the shortcomings of poor convergence accuracy and easily fall into the local optimum of PSO. In bSRWPSO-FKNN, the core of which is to optimize the classification performance of FKNN through binary SRWPSO. To prove that the study has scientific significance, this paper first successfully demonstrates the core advantages of SRWPSO in well-known algorithms through benchmark function validation experiments. Secondly, this article demonstrates that the bSRWPSO-FKNN has practical medical significance and effectiveness through nine public and medical datasets. The 10 times 10-fold cross-validation experiments demonstrate that bSRWPSO-FKNN can pick up the key features of AD, including the content of lymphocytes (LY), Cat dander, Milk, Dermatophagoides Pteronyssinus/Farinae, Ragweed, Cod, and Total IgE. Therefore, the established bSRWPSO-FKNN method practically aids in the diagnosis of AD.

Optimizing Image Fusion Using Modified Principal Component Analysis Algorithm and Adaptive Weighting Scheme

Image fusion is an important technique for combining two or more images to produce a single, high-quality image. Principal component analysis (PCA) is a commonly used method for image fusion. However, existing PCA-based image fusion algorithms have some limitations, such as sensitivity to noise and poor fusion quality. In this paper, we propose a modified PCA algorithm for image fusion that uses an adaptive weighting scheme to improve the fusion quality. The proposed algorithm optimizes the fusion process by selecting the principal components that contain the most useful information and weighing them appropriately. Experimental results show that the proposed algorithm outperforms existing PCA-based image fusion algorithms in terms of fusion quality, sharpness, and contrast. Keywords - Image fusion, principle components analysis, adaptive weighting scheme, optimization, fusion quality, sharpness,contrast.

Multistage Relation Network With Dual-Metric for Few-Shot Hyperspectral Image Classification

Recently, few-shot learning (FSL) has exhibited great potentials in hyperspectral image (HSI) classification due to its promising performance under few training samples. Although existing FSL methods have achieved great success, some limitations can still be witnessed. On the one hand, current methods mainly rely on the single metric to identify, which cannot effectively represent the class distribution with few labeled samples. On the other hand, existing methods usually only use the last deep feature of feature extractor, which may lead to the under-utilization of scarce labeled samples. To overcome the above issues, a novel multistage relation network with dual-metric (DM-MRN) is proposed for few-shot HSI classification. Firstly, a sample recombination strategy is designed to increase the variety of classification tasks in training period. Secondly, an embedding module is employed to extract deep features of the input image patches. Thirdly, we propose two relation modules: image-to-class (I2C) block and image-to-image (I2I) block. I2C block is designed to compute I2C-level relation score between second-order features, and I2I block is conceived to generate I2I-level relation score between first-order features. Finally, DM-MRN is constructed by integrating one embedding module, two I2C blocks, and one I2I block. In addition, an adaptive weighting strategy is designed to fuse the obtained relation scores, and classification can be achieved by assigning each query sample to the class with the highest value of the fused relation score. Extensive experiments carried out on five popular HSI data sets demonstrate that the proposed method outperforms other traditional and advanced models under few training samples in terms of classification accuracy and generalization performance, i.e., the performance improvement in terms of OA is around 0.30%-27.98% under 10 labeled samples per class.

MFGNet: Dynamic Modality-Aware Filter Generation for RGB-T Tracking

Many RGB-T trackers attempt to attain robust feature representation by utilizing an adaptive weighting scheme (or attention mechanism). Different from these works, we propose a new dynamic modality-aware filter generation module (named MFGNet) to boost the message communication between visible and thermal data by adaptively adjusting the convolutional kernels for various input images in practical tracking. Given the image pairs as input, we first encode their features with the backbone network. Then, we concatenate these feature maps and generate dynamic modality-aware filters with two independent networks. The visible and thermal filters will be used to conduct a dynamic convolutional operation on their corresponding input feature maps respectively. Inspired by residual connection, both the generated visible and thermal feature maps will be summarized with input feature maps. The augmented feature maps will be fed into the RoI align module to generate instance-level features for subsequent classification. To address issues caused by heavy occlusion, fast motion and out-of-view, we propose to conduct a joint local and global search by exploiting a new direction-aware target driven attention mechanism. The spatial and temporal recurrent neural network is used to capture the direction-aware context for accurate global attention prediction. Extensive experiments on three large-scale RGB-T tracking benchmark datasets validated the effectiveness of our proposed algorithm.

Toward Physically Plausible Data-Driven Models: A Novel Neural Network Approach to Symbolic Regression

Many real-world systems can be described by mathematical models that are human-comprehensible, easy to analyze and can be helpful in explaining the system’s behavior. Symbolic regression is a method that can automatically generate such models from data. Historically, symbolic regression has been predominantly realized by genetic programming, a method that evolves populations of candidate solutions that are subsequently modified by genetic operators crossover and mutation. However, the approach suffers from several deficiencies: it does not scale well with the number of variables and samples in the training data – models tend to grow in size and complexity without an adequate accuracy gain, and it is hard to fine-tune the inner model coefficients using just genetic operators. Recently, neural networks have been applied to learn the whole analytic model, i.e., its structure as well as the coefficients, by means of gradient-based optimization algorithms. In this paper, we propose a novel neural network-based symbolic regression method that constructs physically plausible models based on very small training data sets and prior knowledge about the system. The method employs an adaptive weighting scheme to effectively deal with multiple loss function terms and an epoch-wise learning process to reduce the chance of getting stuck in poor local optima. Furthermore, we propose a parameter-free method for choosing the model with the best interpolation and extrapolation performance out of all models generated throughout the whole learning process. We experimentally evaluate the approach on four test systems: the TurtleBot 2 mobile robot, the magnetic manipulation system, the equivalent resistance of two resistors in parallel, and the longitudinal force of the anti-lock braking system. The results clearly show the potential of the method to find parsimonious models that comply with the prior knowledge provided.

Fault Assessment for Mechanical Equipment with Adaptive Weights Incorporating Sensitivity and Monotonicity

Driven by the increasing needs in industrial processes for condition-based maintenance, this paper present an adaptive weighting strategy based on sensitivity and monotonicity for fault assessment. Since the sensitivities of features vary with the changes of fault severity, adaptive weight coefficients are designed based on sensitivities to strengthen the feature information. Meanwhile, considering the irreversibility of fault evolution and the difference of sensitive features to different faults, unique monotonic multi-domain feature set with high sensitivity can be selected. Finally, a monotonic health index (HI) is fused based on adaptive weight coefficients for fault assessment which satisfies the needs for intuitiveness in industrial sites. Moreover, the effectiveness of the proposed method is demonstrated by rolling bearing test rig. Results show that the average assessment accuracy can reach 87.5%, 95.375% and 95.375%.

A cloud computing load prediction hybrid model with adaptive weight strategy

A cloud computing load prediction hybrid model with adaptive weight strategy

Aero-Engine Rotor Assembly Process Optimization Based on Improved Harris Hawk Algorithm

Multi-stage disc rotor assembly is an important part of the aero-engine rotor manufacturing process. To solve the problem that excessive unbalance of assembly affects the vibration of the whole machine, this paper presents an optimization method for aero-engine rotor assembly balance based on an improved Harris Hawk algorithm. Firstly, the assembly sequence model of the single-stage disc blade and the phase assembly model of a multi-stage disc of the engine rotor is established. Secondly, by using the initial population generation based on dynamic opposing learning and the escape energy function of the non-linear logarithmic convergence factor, the search mechanism of the whale optimization algorithm is introduced in the global exploration, and the adaptive weight strategy and mutation strategy of the genetic algorithm is introduced in the development to improve the algorithm. Then, the effectiveness of the algorithm is verified by experiments and compared with particle swarm optimization, genetic algorithm, and Harris Hawk algorithm, the unbalance of the optimal blade assembly sequence is reduced by 91.75%, 99.82%, and 83.39%, respectively. The algorithm comparison and analysis are carried out for all disc-blade assembly optimization of the rotor. The optimal unbalance of the improved Harris Hawk optimization algorithm is reduced by 79.71%, 99.48%, and 54.92% on average. The unbalance of the algorithm in this paper is the best. Finally, the improved Harris Hawk algorithm is used to find the best assembly phase, and the optimized unbalanced force and moment are reduced by 84.22% and 98.05%, respectively. The results of this study prove that the improved Harris Hawk algorithm for aero-engine rotor assembly balance optimization can effectively reduce the unbalance of rotor disc blade assembly and rotor unbalance and provide a powerful solution for solving engine vibration.

Improved Whale Optimization Algorithm for Solving Microgrid Operations Planning Problems

Microgrid operations planning is one of the keys to ensuring the safe and efficient outputs of distributed energy resources (DERs) and the stable operation of a power system in a microgrid (MG). In this study, for the symmetry in renewable energy and microgrid systems, and coordinated control based on a storage battery system, an MG dispatching model with DER conditions and integrated costs is established in grid-connected mode, on the basis of MG operation costs, interaction costs, and pollutant emissions costs. Moreover, an optimization objective for minimizing integrated costs is established. Therefore, based on the original whale optimization algorithm (WOA), an improved whale optimization algorithm (IWOA) with adaptive weight strategy and Levy flight trajectory is proposed in this paper, to solve the optimal operations planning problem of MGs. Finally, in computing comparisons with methods such as the genetic algorithm (GA), particle swarm algorithm (PSO), WOA, wild horse optimizer (WHO), and enhanced whale optimization algorithm (EWOA), the results show that the IWOA computation had lower integrated costs and higher operational efficiency. Moreover, it is verified that the IWOA performed better in solving the MG operations planning problem.

An Improved NSGA-II Algorithm Based on Adaptive Weighting and Searching Strategy

Non-dominated sorting genetic algorithm II is a classical multi-objective optimization algorithm but it suffers from poor diversity and the tendency to fall into a local optimum. In this paper, we propose an improved non-dominated sorting genetic algorithm, which aims to address the issues of poor global optimization ability and poor convergence ability. The improved NSGA-II algorithm not only uses Levy distribution for global search, which enables the algorithm to search a wider range, but also improves the local search capability by using the relatively concentrated search property of random walk. Moreover, an adaptive balance parameter is designed to adjust the respective contributions of the exploration and exploitation abilities, which lead to a faster search of the algorithm. It helps to expand the search area, which increases the diversity of the population and avoids getting trapped in a local optimum. The superiority of the improved NSGA-II algorithm is demonstrated through benchmark test functions and a practical application. It is shown that the improved strategy provides an effective improvement in the convergence and diversity of the traditional algorithm.

Particle swarm optimization with Chebychev functional-link network model for engineering design problems

In recent years, many improved particle swarm optimization (PSO) algorithms have been developed to improve the performance of PSO. These improved algorithms have greatly improved PSO performance, but PSO still has some shortcomings. So, an particle swarm optimization with Chebychev functional-link network model is proposed (APSOCFLN) in this paper. Firstly, in order to make up for the shortcomings of the canonical PSO update search, a novel Chebychev functional-link network (CFLN) elite guidance strategy is proposed. Two different update mechanisms are executed alternately, increasing the diversity of the algorithm population and making the algorithm more capable of jumping out of local optimization. Secondly, in order to better balance the exploration and exploitation of the algorithm, an adaptive probability strategy is proposed. Thirdly, an adaptive weighting strategy is proposed. It will be used for CFLN elite guidance strategy and give different weights to the two alternating update strategies, which can effectively solve the problem of premature convergence of PSO. Fourthly, the APSOCFLN and comparison algorithms are used to solve the practical engineering problems, and the results show that APSOCFLN has high precision, fast convergence , and excellent performance. Finally, the performance of the algorithm is tested with CEC2017 and CEC2022 benchmark functions. The comparative algorithm includes the classical PSO improvement algorithms and the classical other algorithms, and the experimental results show the effectiveness of the proposed strategy and the good performance of the improved algorithm.

A semi-supervised adaptive Markov Gaussian embedding process (SAMGEP) for prediction of phenotype event times using the electronic health record

While there exist numerous methods to identify binary phenotypes (i.e. COPD) using electronic health record (EHR) data, few exist to ascertain the timings of phenotype events (i.e. COPD onset or exacerbations). Estimating event times could enable more powerful use of EHR data for longitudinal risk modeling, including survival analysis. Here we introduce Semi-supervised Adaptive Markov Gaussian Embedding Process (SAMGEP), a semi-supervised machine learning algorithm to estimate phenotype event times using EHR data with limited observed labels, which require resource-intensive chart review to obtain. SAMGEP models latent phenotype states as a binary Markov process, and it employs an adaptive weighting strategy to map timestamped EHR features to an embedding function that it models as a state-dependent Gaussian process. SAMGEP’s feature weighting achieves meaningful feature selection, and its predictions significantly improve AUCs and F1 scores over existing approaches in diverse simulations and real-world settings. It is particularly adept at predicting cumulative risk and event counting process functions, and is robust to diverse generative model parameters. Moreover, it achieves high accuracy with few (50–100) labels, efficiently leveraging unlabeled EHR data to maximize information gain from costly-to-obtain event time labels. SAMGEP can be used to estimate accurate phenotype state functions for risk modeling research.

Marine Predators Algorithm Based on Adaptive Weight and Chaos Factor and Its Application

Aiming at the problems of uneven distribution of initialized populations and unbalanced exploration and exploitation leading to slow convergence, low convergence accuracy, and easy to fall into local optimality of marine predators algorithm (MPA), a marine predators algorithm based on adaptive weight and chaos factor is proposed (ACMPA), the algorithm is applied to the traveling salesman problem (TSP), and the shortest path planning and research are carried out for the traveling salesman problem. Firstly, the improved adaptive weight strategy is used to balance the exploration and exploitation stage of the algorithm and improve the convergence accuracy of the algorithm. Secondly, the chaos factor is used to replace the random factor, and the ergodicity of the chaos factor is used to make it easier for predators to jump out of local optimization and enhance the optimization ability of the algorithm. Finally, 10 benchmark test functions, the CEC2015 test set, and the CEC2017 test set are used to evaluate the effectiveness of the ACMPA. The results show that, compared with the other four intelligent optimization algorithms, the improved ACMPA achieves better results in both mean and standard deviation, and the algorithm has a better effect on the shortest path problem.