When addressing constrained multi-objective optimization problems (CMOPs), complex feasible region structures often hinder existing constrained multi-objective evolutionary algorithms (CMOEAs) from maintaining a proper balance between convergence and diversity, leading to incomplete coverage of constrained Pareto front (CPF). To address this issue, we developed a sparse region aware based constrained multi-objective optimization algorithm (SRACMO) using a dual-population co-evolutionary framework. We define a metric called region potential metric (RPM) to quantify the exploration value of different regions in the objective space and incorporate RPM into the parent selection process, enabling the algorithm actively to identify and utilize individuals located in sparse regions and boundaries, thereby enhancing the coverage ability of unexplored regions. In addition, a population state detection mechanism based on multi-index is designed to distinguish between progressive state and stagnant state. When the population gets stuck in stagnation, a recovery strategy combining regional collaborative mating and adaptive ϵ -constraint relaxation is introduced to enhance the regional search ability. Experimental results on 33 benchmark problems show that SRACMO achieves competitive overall performance compared with other 8 classic CMOEAs, especially performing stronger frontier coverage ability in complex feasible region structures.

In recent years, numerous algorithms for constrained multi-objective optimization based on the Competitive Swarm Optimizer (CSO) have demonstrated remarkably rapid convergence. However, after the swarm converges to the Constrained Pareto Front (CPF), these algorithms often encounter significant challenges in effective local exploitation. To enhance fine-grained search performance in the vicinity of the CPF, this paper proposes a clustering-based dynamic adaptive search strategy. The proposed strategy partitions the swarm into multiple sub-swarms via a clustering process, with the number of clusters adaptively adjusted throughout evolution. It dynamically modulates the swarm’s search dynamics — switching between cohesive global convergence and partitioned local exploitation — to effectively navigate constrained landscapes while maintaining both high convergence and well-distributed diversity along the CPF. Furthermore, a diversity enhancement strategy is designed to effectively balance solution diversity and constraint feasibility, enabling the swarm to achieve more uniform convergence toward the CPF and enhancing search effectiveness within the feasible solution region. Based on the proposed methodology, a novel algorithm for solving constrained multi-objective optimization problems (CMOPs) is developed. Extensive experimental results on 37 benchmark instances and 16 real-world engineering applications demonstrate that the proposed algorithm significantly outperforms nine state-of-the-art approaches, achieving either superior or highly competitive performance.

Constrained multi-objective optimization problems (CMOPs) are prevalent across various fields. In recent years, numerous constrained multi-objective evolutionary algorithms (CMOEAs) have been developed to tackle these challenges. However, existing CMOEAs still encounter significant challenges when addressing complex CMOPs. For instance, tackling CMOPs characterized by very narrow or disconnected feasible regions poses a significant challenge for current algorithms, particularly in simultaneously ensuring solution feasibility, diversity, and convergence. To address this limitation, this paper proposes DDCEO, a dual-stage dual-population evolutionary algorithm using new adaptive environmental selection method. Within DDCEO, the main population primarily focuses on ensuring solution feasibility, while the auxiliary population assists in enhancing diversity and convergence. The evolutionary process of the auxiliary population is divided into two distinct phases: Phase 1 focuses on global exploration by disregarding constraints to help locate potential feasible domains, while Phase 2 introduces an adaptive environmental selection method that enhances diversity and convergence by strategically retaining non-dominated infeasible solutions positioned far from the main population yet close to the constraint boundaries. Finally, comprehensive experiments on three benchmark test suites demonstrate that the proposed DDCEO algorithm shows superior performance against seven state-of-the-art CMOEAs, namely tDEACPBI, DDCMOEA, ToP, CCMO, BiCo, MTCMO, and CMOEMT. Statistical tests using the IGD and HV metrics confirm that DDCEO holds a significant advantage on the majority of complex test problems.

Existing decomposition-based constrained multi-objective evolutionary algorithms (CMOEAs) use fixed decomposition methods to partition the search space, which may limit their ability in solving some certain constrained multi-objective optimization problems (CMOPs). To address this issue, this paper introduces the concept of hyper-feasible solutions, which are extracted from feasible solutions and promising infeasible solutions. Based on these solutions, we propose a novel algorithm called HSWU, which adaptively partitions the search space and guides the search direction to enhance the efficiency of solution searching. Experimental results on three benchmark test suites demonstrate that HSWU outperforms five state-of-the-art CMOEAs in terms of performance.

When addressing constrained multi-objective optimization problems, the presence of complex constraints often results in a non-connected feasible region, segmenting the Pareto front into multiple discrete segments. This fragmentation can significantly limit population diversity. To tackle this issue, we have designed two mechanisms aimed at preserving population diversity and have developed a constrained multi-objective co-evolutionary algorithm (DESCA) based on the framework of a two-population co-evolutionary algorithm. The proposed algorithm consists of two populations: a main population dedicated to exploring the constrained Pareto front and an auxiliary population tasked with exploring the unconstrained Pareto front. To sustain the diversity within both populations, the algorithm dynamically adjusts the genetic operator based on the observed states of the populations. Moreover, when the main population encounters stagnation, a regional mating mechanism is employed between the main population and the auxiliary population, accompanied by a relaxation of the constraints on the main population. Conversely, when the auxiliary population experiences stagnation, a diversity-first individual selection strategy is implemented; this strategy utilizes a regional distribution index to assess individual diversity and mitigates population stagnation by enhancing diversity. The performance of DESCA has been evaluated across 33 benchmark problems and 6 real-world problems. Experimental results demonstrate that DESCA exhibits strong competitiveness compared to seven other typical state-of-the-art algorithms.

This research aims to address critical challenges in multivariate time series forecasting: capturing complex temporal patterns and overcoming the exponential decay issue in the long-term memory of Long Short-Term Memory (LSTM) models. To address these, we propose a new deep learning model called EL-LSTM. The model combines a simplified Leaky Integrate-and-Fire (LIF) neuron model with LSTM and significantly enhances the model's ability to capture complex temporal dependencies through local and global point attention mechanisms. The EL-LSTM model can effectively retain key information. We have validated the effectiveness of the model in multivariate time series prediction tasks, especially in the fields of finance and traffic flow forecasting. Experimental results show that the EL-LSTM model has achieved significant improvements in forecasting accuracy, particularly when dealing with data in concept-drift environments. This study offers a new perspective for the field of time series prediction and demonstrates the potential for handling complex time series data in practical applications.

Both dual-population and two-phase strategies are effective for utilizing infeasible solution information and significantly enhancing the ability of algorithms to solve constrained multiobjective optimization problems. However, most existing algorithms tend to underperform when facing problems with complex constraints. To address these issues, a constrained multiobjective evolutionary algorithm named DPTPEA, which combines dual-population and two-phase strategies, is proposed in this article. DPTPEA employs two collaborative populations [the exploitive population (expPop) and the tractive population (tracPop)] and divides the evolutionary process of the tracPop into two phases (Phase 1 and Phase 2). In Phase 1, the tracPop ignores constraints and drags the expPop across the infeasible region by sharing offspring information. In Phase 2, the tracPop adopts the epsilon-constrained method to converge toward the constrained Pareto front and to guide the expPop exploiting different feasible regions. Moreover, a dynamic cooperation strategy, a boundary point direction sampling strategy, and a dynamic environmental selection are proposed to improve the exploration ability of tracPop for solving complex problems. Comprehensive experiments on three popular test suites demonstrate that DPTPEA outperforms seven state-of-the-art algorithms on most test problems.

Background: Colorectal cancer (CRC) with peritoneal metastasis (PM) is associated with poor prognosis. The Peritoneal Cancer Index (PCI) is used to evaluate the extent of PM and to select Cytoreductive Surgery (CRS). However, PCI score is not accurate to guide patient's selection for CRS. Objective: We have developed a novel AI framework of decoupling feature alignment and fusion (DeAF) by deep learning to aid selection of PM patients and predict surgical completeness of CRS. Methods: 186 CRC patients with PM recruited from four tertiary hospitals were enrolled. In the training cohort, deep learning was used to train the DeAF model using Simsiam algorithms by contrast CT images and then fuse clinicopathological parameters to increase performance. The accuracy, sensitivity, specificity, and AUC by ROC were evaluated both in the internal validation cohort and three external cohorts. Results: The DeAF model demonstrated a robust accuracy to predict the completeness of CRS with AUC of 0.9 (95 % CI: 0.793-1.000) in internal validation cohort. The model can guide selection of suitable patients and predict potential benefits from CRS. The high predictive performance in predicting CRS completeness were validated in three external cohorts with AUC values of 0.906(95 % CI: 0.812-1.000), 0.960(95 % CI: 0.885-1.000), and 0.933 (95 % CI: 0.791-1.000), respectively. Conclusion: The novel DeAF framework can aid surgeons to select suitable PM patients for CRS and predict the completeness of CRS. The model can change surgical decision-making and provide potential benefits for PM patients.

This kind of algorithm composed of multiple operators, when solving different constrained multi-objective optimization problems (CMOP5), always has operators with good effects guiding the population to seek a better Pareto Front (PF). However, in the evolutionary process of such algorithms, there exist operators that have no effect but still generate offspring, thereby slowing down the convergence speed of the algorithm. To accelerate the convergence speed of the algorithm, a constrained multi-objective co-Evolutionary algorithm based on operator score and reward (SRCA) is presented in this paper, this SRCA algorithm has proposed an operator evaluation and operator reward mechanism which attempt to select operators that are beneficial to the convergence and diversity of the population for reproduction. The experimental results demonstrate that SRCA algorithm can effectively expedite the convergence speed and enhance the diversity of the population.

Targeting the poor precision, limited real-time and high model complexity of defects and exotic objects detection in solar photovoltaic panels, a new intelligent detection algorithm, SPP YOLO, is proposed. Expanded solar photovoltaic panels data from StyleGAN2-ADA. Building upon the YOLOv11 architecture, the proposed SPP YOLO method integrates Dynamic Snake Convolution (DSC) operations within the backbone’s CBS modules, resulting in the formation of DBS modules that leverage adaptive convolutional processing. By enhancing the global feature focus, this integration preserves the key information related to different global morphologies and improves the precision of target detection in the model. In addition, the coordinate attention mechanism is integrated into the C3K2 module to enhance the spatial perception of the model and reduce feature duplication. The use of the lightweight upsampling operator CARAFE in the feature extraction network allows contextual information to be collected across a wide range of sensory domains, improving the feature extraction and fusion capabilities of the model. A learning rate optimisation strategy based on Sparrow search algorithm (SSA) is used during model training to further improve the detection accuracy of the model. The proposed SPP YOLO algorithm, which helps to achieve a better balance between efficiency and accuracy in solar panel inspection, shows significant overall effectiveness and provides theoretical support for industrial smart manufacturing.

Using real-time and precise detection methods for maize leaf disease can significantly reduce economic losses in agriculture. Practical implementation often faces challenges such as the large volume of leaf disease data, low identification accuracy, and inefficiencies in production environments. To address these issues, this study introduces YOLO-MSM, a maize leaf disease detection algorithm that integrates multi-scale variable kernel convolution. In the YOLO MSM algorithm, we introduce an innovative convolutional method, MKConv (Multi-scale Variable Kernel Convolution), which offers diverse parameter configuration options and adapts flexibly to sample shapes with specific data characteristics. This design significantly enhances the network’s overall performance. Additionally, to highlight critical features and mitigate the influence of environmental noise, we develop the C2f-SK module, leveraging the SK (Selective Kernel) attention mechanism to optimize feature extraction and representation. The loss function is optimized using MPDIoU (Minimum Point Distance Intersection over Union) to enhance the algorithm’s capability in accurately locating densely occluded targets. The findings from the experiments indicate that the YOLO MSM algorithm reaches a real-time detection rate of 279.56 fps. In comparison to the baseline algorithm, the algorithm improves the precision and recall by 0.66% and 1.61%, respectively. Moreover, YOLO MSM algorithm is effectively lightweight compared to the series of cutting-edge algorithm models, which are only 5.4 MB in size, and the number of parameters and Flops are also reduced significantly. Therefore, YOLO MSM algorithm has an obvious light-weight advantage, which can achieve a good balance between precision and speed, and lay a theoretical foundation for identifying and detecting leaf disease on mobile devices.

In the field of evolutionary computation, gene expression programming (GEP) has been favored by the public because of its straightforward encoding method. Due to the coding rules for the head and tail on gene, the ability of gene to be constructed as an expression is limited, thereby limiting the expressiveness of individuals. In this article, a variant that uses the standard GEP individual structure and is based on code reuse strategy is proposed to maintain the concise individual representation advantage of GEP and improve the expressive ability of individual. This method can improve the accuracy of the individual and reduce the expression tree complexity. In addition, experiments conducted to predict the recurrence of cervical cancer and malignancy of breast cancer have demonstrated that the predicted performance of GEP and gene expression programming with structured reusability (SR-GEP) has performed better than that of C5.0, but SR-GEP performs best.

To enhance the precision of damage monitoring of cave paintings in cultural heritage protection and to realize fast real-time detection, this paper proposes an intelligent monitoring algorithm using computer vision technology, called YOLO CP(Cave Paintings). The C2f-FasterEMA Block module is developed by the algorithm, which also refines the deep layer’s residual module in the backbone network. This improvement boosts the capability to extract target features while minimizing the number of parameters. Moreover, the RepGD (Rep Gather-and-Distribute) mechanism has been integrated into the feature fusion network, thereby boosting the capability of cross-layer feature information fusion and enhancing the model’s detection accuracy. Ultimately, Inner-SIoU (Linear Spatial Intersection over Union) is presented to enhance the loss function. This approach utilizes a more appropriate aspect ratio metric and addresses the shortcomings of the original loss function, thereby speeding up the convergence of the model. The findings from the experiment indicate that the YOLO CP algorithm decreases the parameters and floating-point operations by 8.14% and 7.14%, respectively, when compared with the original YOLOv10n baseline algorithm in the task of detecting damage in cave paintings. At the same time, the YOLO CP model shows enhancements in Precision and Recall by 1.82 and 7.05 percentage points, respectively. The speed of detection in real-time achieves 277.39 FPS. The proposed algorithm demonstrates a considerable enhancement in performance, offering both theoretical and technical backing for the automated and intelligent detection of cave painting damage. Additionally, it holds the promise of quick implementation on embedded devices, serving as a crucial technical assurance for the preservation of cultural artifacts and heritage.

Feature selection is a complicated optimization problem with significant practical applications. Nevertheless, its strength lies in significantly reducing the size of datasets and increasing classification efficiency. Several evolutionary algorithms (EAs) have been used to solve feature selection problems. However, most EAs are unsuitable to deal with real-world problems with multiple objectives. The multi-objective feature selection problems normally consist of two objectives: minimizing the number of feature selections and minimizing the classification errors. In this paper, a replication analysis method based on the evolutionary algorithm (RAEA) is proposed to classify bi-objective selected features. The proposed method has improved the framework of dominance-based EA from two viewpoints: firstly, modify the reproduction procedure to improve the features of offspring, and secondly, the replication analysis technique has been proposed to filter out unnecessary solutions. We conducted some experiments with the proposed method and compared it with five traditional MOEAs. The experimental results show that RAEA performs better results on most datasets, indicating that RAEA not only performs best in the optimization process but also performs better results in generalization and classification.

With the development of technology, unmanned aerial vehicles (UAVs) and Internet of Things devices are widely used in smart agriculture, resulting in significant energy consumption. In this paper, the optimization problem for UAV-assisted mobile computing in smart agriculture is modeled as a constrained multi-objective optimization problem. By jointly optimizing the deployment position of UAVs, the offloading location of the tasks, the transmit power of the devices, and the resource allocation of the UAVs, two optimization objectives (total delay and energy consumption) are minimized simultaneously. In view of the complex constraints, a constrained multiobjective algorithm named JO-DPTS is proposed. The algorithm adopts dual-population and two-stage approach to improve population convergence and diversity. The simulation results substantiate that JO-DPTS exhibits superior performance compared to the other three state-of-the-art constrained multi-objective evolutionary algorithms.

Achieving the balance between convergence and diversity is a key and challenging issue in many-objective optimization. Reference vector guided selection is an exemplary method for decomposition-based many-objective evolutionary algorithms (MaOEAs). However, there are some problems with it such as insufficient number of obtained solutions and inefficient convergence evaluation metric. Aiming at solving or alleviating these problems, this paper proposes a many-objective evolutionary algorithm based on reference vector guided selection and two diversity and convergence enhancement strategies. The proposed algorithm introduces two new strategies namely adaptive sparse region detection and convergence-only selection. The former is to adaptively detect sparse regions of current elite population, while the latter is to prevent the elimination of solutions with prominent convergence performance. Together with a newly proposed elite retention strategy, these two strategies can achieve diversity and convergence enhancement on the basis on reference vector guided selection. Besides, A new selection criterion for reference vector guided selection is proposed to better measure the convergence of solutions in high dimensionality. Experimental results on widely used test problem suites up to 15 objectives indicate that the proposed algorithm is highly competitive in comparison with seven state-of-the-art MaOEAs.

TNM classification of colorectal cancer is of great significance for doctors to make clinical decision, evaluate patient prognosis and improve treatment. The diagnosis results of TNM classification of colorectal cancer combined with multi-modal medical data are often more accurate than those based on single modal medical data. However, how to balance the redundancy and complementarity of multimodal medical data in deep learning is a difficult problem. Considering the expensive collection and labeling of medical data, we propose an improved self-supervised contrastive learning method for TNM classification of colorectal cancer. Self-supervised contrast learning guides the feature representation of the learning data according to its own supervised information. In the feature extraction stage, we used a deep convolutional neural network with a spatial-channel attention module to extract the features of MRI images, and incorporated clinicopathological parameters in the fully connected layer to achieve multi-modal data fusion. In the comparison of feature similarity, Mahalanobis distance measurement learning method is used to eliminate the scale interference of features between different models. In the contrast loss stage, we design a contrast loss function suitable for multimodal feature representation for model training. In order to verify the effectiveness of the proposed method, experiments were carried out on the collected data sets. The experimental results show that compared with traditional methods, the TNM staging diagnosis technique proposed in this study based on improved self-supervised comparative learning has achieved significant improvement in accuracy and recall rate.

Heterogeneous attribute data (also called mixed data), characterized by attributes with numerical and categorical values, occur frequently across various scenarios. Since the annotation cost is high, clustering has emerged as a favorable technique for analyzing unlabeled mixed data. To address the complex real-world clustering task, this paper proposes a new clustering method called Adaptive Micro Partition and Hierarchical Merging (AMPHM) based on neighborhood rough set theory and a novel hierarchical merging mechanism. Specifically, we present a distance metric unified on numerical and categorical attributes to leverage neighborhood rough sets in partitioning data objects into fine-grained compact clusters. Then, we gradually merge the current most similar clusters to avoid incorporating dissimilar objects into a similar cluster. It turns out that the proposed approach breaks through the clustering performance bottleneck brought by the pre-set number of sought clusters k and cluster distribution bias, and is thus capable of clustering datasets comprising various combinations of numerical and categorical attributes. Extensive experimental evaluations comparing the proposed AMPHM with state-of-the-art counterparts on various datasets demonstrate its superiority.

With the rapid development of Internet information technology, vast amounts of textual data are constantly emerging. However, raw data is often unstructured and lacks readability, posing significant challenges for data mining. Feature selection, as a crucial step in machine learning, is essential for enhancing model performance. In this paper, we propose an Improved Binary Particle SwarmOptimization (IBPSO) algorithm that addresses the limited search capability of traditional swarm intelligence algorithms in high-dimensional optimization problems. By incorporating a dynamic nonlinear decreasing inertia weight update strategy and a local probability crossover mutation strategy, IBPSO effectively enhances algorithm performance while overcoming the limitations of conventional approaches. The proposed algorithm demonstrates superior performance through testing on benchmark functions.

As the number of objectives increases, many-objective optimization problems (MaOPs) become increasingly complex. Traditional indicator-based many-objective evolutionary algorithms can often ensure the convergence of the population but tend to struggle with maintaining its diversity. In this paper, an enhanced diversity indicator-based many-objective evolutionary algorithm with shape-conforming convergence metric, namely, MaOEA-DISC, is presented to relieve this weakness. In MaOEA-DISC, firstly, we focus on the inter-individual spacing relationships within the population based on I-& varepsilon;+ Indicator, proposing a novel enhanced diversity I-& varepsilon;+ Indicator to ensure the enhancement of population diversity while converging. Secondly, we propose a new metric of individual convergence, which calculates the convergence of individuals based on the shape of Pareto front, reducing the errors caused by the shape of the Pareto front when measuring individual convergence, thereby assessing individual convergence more accurately. Finally, to further improve the convergence speed of the population, different mating strategies are employed for mating in the parental generation. MaOEA-DISC is compared with other algorithms on various benchmark MaOPs ranging from 3 to 10 objectives, as well as a real-world MaOPs. Experimental results demonstrate that when dealing with MaOPs, MaOEA-DISC not only achieves excellent population convergence and diversity but also effectively maintains a balance between them, showing promising practical value.