Applied Soft Computing最新文献

{"title":"Quantum alternating operator ansatz with PSO optimizer for portfolio optimization problem","authors":"Wenjie Liu ,&nbsp;Yue Ma ,&nbsp;Yuchen Gu ,&nbsp;Jiajun Cheng ,&nbsp;Qingshan Wu","doi":"10.1016/j.asoc.2025.113419","DOIUrl":"10.1016/j.asoc.2025.113419","url":null,"abstract":"<div><div>Recently, Quantum Approximate Optimization Algorithm (QAOA) and Quantum Alternating Operator ansatz (QAOAz) are utilized to solve the Mean Variance (MV) model for the portfolio optimization problem (QAOAz-MV), which shows performance advantages over classical algorithms on this huge search space. For the more complex and comprehensive risk parity (RP) model, a novel QAOAz solution (QAOAz-RP) is proposed. We begin by defining the RP model for the portfolio optimization problem. Next, we detail the QAOAz algorithm process, where the problem Hamiltonian with the ZZZZ term is derived, the corresponding quantum circuit is ingeniously constructed using the parity check method, and the whole quantum circuit containing the ring XY-mixer is given. Finally, to improve the optimization performance of QAOAz, a Particle Swarm Optimization (PSO) optimizer is introduced to tune the parameters of the quantum circuits, which is applicable to both QAOAz-MV and QAOAz-RP. The experiment conducted on multiple financial markets (e.g. Chinese, U.S., and European) demonstrate that PSO-QAOAz-RP is significantly better for portfolio optimization than ABC-LP, GWO, GA and QAOAz-RP on eight portfolios in all metrics. PSO-QAOAz-MV also has advantage for MV model over the quantum algorithms, including QAOA (improves approximate ratio by 54.79% on average) and QAOAz (improves approximate ratio by 15.38% on average). This study not only provides a breakthrough quantum solution for portfolio optimization, but also provides a reusable technology paradigm for the deep integration of quantum computing and financial engineering.</div></div>","PeriodicalId":50737,"journal":{"name":"Applied Soft Computing","volume":"181 ","pages":"Article 113419"},"PeriodicalIF":7.2,"publicationDate":"2025-06-14","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144322524","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Multi-Level semantic and spatial hierarchy reasoning for 3D face reconstruction and dense alignment in unconstrained environments","authors":"Lei Li ,&nbsp;Fuqiang Liu ,&nbsp;Junyuan Wang ,&nbsp;Yanni Wang ,&nbsp;Zhitao Zhang ,&nbsp;Jiahao Li ,&nbsp;Qi Wang","doi":"10.1016/j.asoc.2025.113327","DOIUrl":"10.1016/j.asoc.2025.113327","url":null,"abstract":"<div><div>3D face reconstruction and dense alignment tasks encounter significant challenges when the samples are in highly unconstrained environments, particularly with large poses, extreme expressions, occlusions and complex backgrounds. Although carefully designing the neural network architecture can enhance the representation ability, its performance is still unsatisfactory due to the absence of accurate facial semantic and spatial hierarchy information. To address this challenge, we propose an approach that uses neural networks to capture multi-level facial semantic and spatial hierarchy knowledge so as to guide the learning process. Specifically, our approach, referred to as Multi-Level Semantic and Spatial Hierarchy Reasoning Network (MSHRNet), leverages a point-to-space level progressive face structure loss function to precisely learn the semantic and spatial hierarchy knowledge of different facial parts. This knowledge is then injected into the backbone network through multi-level hierarchy knowledge matrices to incorporate structural reasoning knowledge, which can suppress the effects of large poses, extreme expressions, and occlusions. Moreover, we introduce a sample-to-dataset level data augmentation module that effectively yields rich and diverse semantic and spatial hierarchy information to inhibit occlusions and complex backgrounds while learning fine-grained local details. Extensive quantitative and qualitative experiments on benchmark datasets demonstrate that our MSHRNet outperforms the state-of-the-art methods in terms of both accuracy and computational complexity at the cost of little increase in the number of parameters. Codes and all data are publicly available at <span><span>https://github.com/Ray-tju/MSHRNet</span><svg><path></path></svg></span>.</div></div>","PeriodicalId":50737,"journal":{"name":"Applied Soft Computing","volume":"180 ","pages":"Article 113327"},"PeriodicalIF":7.2,"publicationDate":"2025-06-14","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144298346","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Federated semi-supervised learning with contrastive representations against noisy labels","authors":"Wenjie Mao ,&nbsp;Bin Yu ,&nbsp;Yihan Lv ,&nbsp;Yu Xie ,&nbsp;Chen Zhang","doi":"10.1016/j.asoc.2025.113421","DOIUrl":"10.1016/j.asoc.2025.113421","url":null,"abstract":"<div><div>Federated semi-supervised learning presents a pragmatic scenario wherein a centralized model is trained utilizing a server with access to labeled data, while participating clients lack any labeled data. In this context, the inaccuracy of real-world labels on the server available for training poses a huge challenge to the federated semi-supervised learning. These inaccuracies can have a detrimental impact on the overall performance of the system and impose limitations on its use. In this paper, we propose a novel Federated Semi-supervised learning framework with Contrastive Representations, called FedCR, with the aim of addressing the aforementioned ubiquitous problems in the field of image classification tasks. Firstly, our approach employs contrastive representation learning to build memory representations of images, which can learn an image’s general features from an augmented view without relying on negative pairs and prevent the model from memorizing noise. Then we take a cautious approach during model updates to prevent any potential leakage to ensure the privacy and security of the clients’ information. Additionally, for the sake of improving robustness of the model, a contrastive regularization function is applied to preserve information connected to true labels while filtering out information associated with wrong labels. Furthermore, we mitigate the negative impact of mislabeled data during supervised learning by utilizing an improved cross-entropy loss function. Extensive experiments on prevalent datasets for image classification tasks show that the proposed method surpasses previously established state-of-the-art federated semi-supervised learning algorithms and efficiently alleviates the issue of model over-fitting to erroneous labels, especially when label noise is present.</div></div>","PeriodicalId":50737,"journal":{"name":"Applied Soft Computing","volume":"180 ","pages":"Article 113421"},"PeriodicalIF":7.2,"publicationDate":"2025-06-14","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144298343","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"A new link prediction model for grain trade networks based on improved variational graph autoencoder and genetic algorithm","authors":"Yanhui Li ,&nbsp;Yuzhi Song ,&nbsp;Qi Yao ,&nbsp;Xu Guan","doi":"10.1016/j.asoc.2025.113336","DOIUrl":"10.1016/j.asoc.2025.113336","url":null,"abstract":"<div><div>Food security is related to the national economy and people’s livelihood, and exploring potential cooperative relationships with oneself is one of the most effective strategies to prevent and mitigate the risk of food import supply chain disruptions. How to use prior information in the grain import trade network to obtain better potential representations of nodes is a key issue in link prediction tasks. Under the theoretical framework of the variational graph autoencoder, this paper creates a new link prediction model, IVGAE-GA. Two feature extraction modules and a feature fusion module are designed to mine effective information in the trade networks. Specifically, a dynamic adaptive graph attention (DAGAN) module is proposed to extract high-order feature information from trade networks. Then, the neighborhood feature information of each node is captured through the graph convolutional neural network (GCN) to strengthen the guiding effect of the initial prior information on the prediction results. In addition, an average feature fusion (AVFF) module is designed to further refine the latent representation of nodes by mixing these non-local and local feature information. The entire IVGAE framework is optimized through cross-entropy loss and KL loss. Finally, the genetic algorithm (GA) is utilized for hyperparameter selection to help the model perform better. Extensive experimental results on two widely used publicly available datasets and four real grain trade networks illustrate that our model achieves better prediction performance compared to some existing methods. The proposed link prediction framework can be a good option for predicting potential cooperative relationships.</div></div>","PeriodicalId":50737,"journal":{"name":"Applied Soft Computing","volume":"180 ","pages":"Article 113336"},"PeriodicalIF":7.2,"publicationDate":"2025-06-14","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144306998","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"An optimized classifier chains‐based deep learning framework for Inter-Turn Fault diagnosis in Permanent Magnet Synchronous Motors","authors":"Amir Hossein Baharvand ,&nbsp;Sina Hossein Beigi Fard ,&nbsp;Amir Hossein Poursaeed ,&nbsp;Meysam Doostizadeh","doi":"10.1016/j.asoc.2025.113482","DOIUrl":"10.1016/j.asoc.2025.113482","url":null,"abstract":"<div><div>Inter-Turn Faults (ITF) of Permanent Magnet Synchronous Motor (PMSM) pose a major challenge. Early detection of these faults improves PMSM performance for predictive maintenance, preventing performance drops and reducing maintenance costs. This paper introduces a new model for the automatic detection of ITF, utilizing an optimized convolutional neural network (CNN). The proposed model incorporates convolutional layers for feature extraction, normalization layers to achieve better convergence, dropout layers to avoid overfitting, and bi-long short-term memory layers (LSTM) to preserve temporal dependencies. The LSTM layers of CNN aid in time series data analysis. Furthermore, Bayesian optimization is used to automatically select and optimize the CNN model’s parameters and improve its performance. This system has several outputs to identify the fault types and their exact location. The classifier chain technique is utilized to maintain independence between different outputs, thereby increasing the system’s accuracy and efficiency. The data used in this study includes the phase currents of the PMSM in healthy and faulty conditions with different intensities. Our proposed model is designed as a multi-output system and can detect both the fault type, such as the fault from phase A to ABC, and the fault locations in three phases, ranging from 10 % to 90 %. Additionally, this model’s performance, along with other models considered for comparison, has been evaluated using various criteria such as accuracy and F1-score to testify to the effectiveness of the proposed method. The results indicate that the proposed optimized CNN model can automatically detect stator ITFs with an accuracy higher than 95 %.</div></div>","PeriodicalId":50737,"journal":{"name":"Applied Soft Computing","volume":"180 ","pages":"Article 113482"},"PeriodicalIF":7.2,"publicationDate":"2025-06-14","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144313333","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Adversarial hard negative samples for continual relation extraction","authors":"Shunyu Yao ,&nbsp;Jinyu Guo ,&nbsp;Jijie Li ,&nbsp;Jie Ou ,&nbsp;Yufei Feng ,&nbsp;Jie Hu ,&nbsp;Dan Liu","doi":"10.1016/j.asoc.2025.113365","DOIUrl":"10.1016/j.asoc.2025.113365","url":null,"abstract":"<div><div>Continual relation extraction (CRE) is a crucial task in continuous learning, aiming to train a model continually on data of new relations to extract relations between entities from unstructured text. This process involves learning newly emerged relations while avoiding catastrophic forgetting of previously learned relations. Existing works have demonstrated that storing a few typical samples of old relations in memory and replaying them during subsequent training for new relations is helpful. It can assist the model to maintain a stable understanding of old relations, thus effectively avoiding forgetting them. However, most prior research has focused on efficiently utilizing memory samples while neglecting their learning difficulty, which refers to the challenge in mastering these samples. In this paper, we propose an adversarial hard negative samples selection mechanism to increase the diversity of memory samples and dynamically adjust the number of samples among relations according to the performance of the model. Experimental results show that our method consistently improves the performance of state-of-the-art CRE models without increasing the number of training samples on mainstream benchmarks.</div></div>","PeriodicalId":50737,"journal":{"name":"Applied Soft Computing","volume":"181 ","pages":"Article 113365"},"PeriodicalIF":7.2,"publicationDate":"2025-06-14","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144322521","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Fault diagnosis of helicopter accessory gearbox under multiple operating conditions based on feature mode decomposition and multi-scale convolutional neural networks","authors":"Anping Wan ,&nbsp;Zengzhen Zhu ,&nbsp;Khalil AL-Bukhaiti ,&nbsp;Xiaomin Cheng ,&nbsp;Xiaosheng Ji ,&nbsp;Jinglin Wang ,&nbsp;Tianmin Shan","doi":"10.1016/j.asoc.2025.113403","DOIUrl":"10.1016/j.asoc.2025.113403","url":null,"abstract":"<div><div>The expanding global civil helicopter market has intensified the need for advanced fault diagnosis techniques in helicopter engines. Traditional fault classification methods, such as Variational Mode Decomposition (VMD), have limitations in decomposing complex signals and separating different signal components, which can impede accurate fault feature extraction and compromise diagnostic accuracy and reliability. This paper introduces a novel fault diagnosis method that combines Feature Mode Decomposition (FMD) with a Multi-Scale Convolutional Neural Network (MCNN) to address these challenges. The approach begins by collecting signals from helicopter accessory gearboxes under simulated ground operation conditions. The FMD technique is applied to decompose the gear vibration signals, and the decomposed signals from different sensors are reconstructed and normalized. This preprocessed data is then fed into the MCNN network at various scales, enabling simultaneous extraction and fusion of multi-scale features. The final fault classification is performed using a <span><math><mi>softmax</mi></math></span> classifier. Experimental results demonstrate the efficacy of the proposed method in extracting fault features. Under the given conditions, a fault diagnosis accuracy of up to 100 % was achieved for helicopter accessory gearboxes, marking a significant improvement of 3.1 % compared to VMD-based methods. This enhancement in accuracy represents a substantial advancement in aviation safety and reliability. The study showcases the superior performance of FMD in decomposing complex mechanical signals, particularly its ability better to capture both periodic and impulsive characteristics of fault signals. The integration of MCNN allows for more effective multi-sensor data processing, enhancing the model's capacity to detect and classify faults across various scales and conditions. The FMD-MCNN approach improves the accuracy and efficiency of fault diagnosis and demonstrates significant potential for practical application in aviation maintenance technology.</div></div>","PeriodicalId":50737,"journal":{"name":"Applied Soft Computing","volume":"180 ","pages":"Article 113403"},"PeriodicalIF":7.2,"publicationDate":"2025-06-13","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144291170","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Advances in EEG-based emotion recognition: Challenges, methodologies, and future directions","authors":"Jichi Chen ,&nbsp;Yuguo Cui ,&nbsp;Chunfeng Wei ,&nbsp;Kemal Polat ,&nbsp;Fayadh Alenezi","doi":"10.1016/j.asoc.2025.113478","DOIUrl":"10.1016/j.asoc.2025.113478","url":null,"abstract":"<div><div>Emotion recognition plays a pivotal role in affective computing and human-computer interaction, especially in the fields of mental health care, auxiliary medicine, and intelligent system design. As a non-invasive and time-sensitive neural signal, electroencephalogram (EEG) has become an important means of emotion recognition research. However, due to its susceptibility to noise and individual differences, EEG-based emotion recognition still faces major challenges. This review systematically summarizes the latest progress in EEG-based emotion recognition, sorts out the research paradigm of EEG-based emotion recognition, including public datasets, signal preprocessing techniques, feature extraction methods, and recognition models, and focuses on the end-to-end modeling advantages of deep learning methods in this field in recent years. Through a comparative analysis of representative literature, this study concludes that although deep learning models have promoted the development of this field, their generalization ability, interpretability, and applicability in real-world scenarios are still limited. In addition, current EEG datasets are often limited by small sample size, lack of diversity, and inconsistent labeling standards. In summary, future research should focus on cross-subject recognition techniques, small sample learning strategies, and the development of real-time, deployable emotion recognition systems. These directions are expected to bridge the gap between academic research and practical applications and further promote the advancement of EEG-based emotion recognition technology.</div></div>","PeriodicalId":50737,"journal":{"name":"Applied Soft Computing","volume":"180 ","pages":"Article 113478"},"PeriodicalIF":7.2,"publicationDate":"2025-06-13","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144298237","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Constraint Consensus assisted Evolutionary Algorithm for large-scale constrained optimization","authors":"Noha Hamza ,&nbsp;Saber Elsayed ,&nbsp;Ruhul Sarker ,&nbsp;Daryl Essam","doi":"10.1016/j.asoc.2025.113383","DOIUrl":"10.1016/j.asoc.2025.113383","url":null,"abstract":"<div><div>Large-scale constrained optimization problems present significant challenges due to their large number of variables and many constraints. Improper handling of these constraints can lead to suboptimal or infeasible solutions. Many existing approaches overlook this aspect. In this paper, we integrate a constraint-objective cooperative coevolution framework with a Constraint Consensus method, known as DBmax (Maximum Direction-based Method), into differential evolution. In this framework, a problem is decomposed into a number of smaller subproblems (subcomponents) using the Recursive Differential Grouping technique, where interactive variables are allocated to one subproblem. By assessing the impact of each group on the objective function and constraint violation, the most suitable group is selected for evolution. Subsequently, the DBmax method is applied adaptively to the infeasible solutions within the chosen group for improving their feasibility. The algorithm was evaluated on 12 test problems, with the experimental results consistently demonstrating its effectiveness by outperforming existing state-of-the-art methods, in terms of the solution’s feasibility and quality.</div></div>","PeriodicalId":50737,"journal":{"name":"Applied Soft Computing","volume":"181 ","pages":"Article 113383"},"PeriodicalIF":7.2,"publicationDate":"2025-06-13","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144313355","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Random-key algorithms for optimizing integrated Operating Room Scheduling","authors":"Bruno Salezze Vieira ,&nbsp;Eduardo Machado Silva ,&nbsp;Antônio Augusto Chaves","doi":"10.1016/j.asoc.2025.113368","DOIUrl":"10.1016/j.asoc.2025.113368","url":null,"abstract":"<div><div>Efficient surgery room scheduling is essential for hospital efficiency, patient satisfaction, and resource utilization. This study addresses the challenge as a combinatorial optimization problem that incorporates multi-room scheduling, equipment scheduling, and complex availability constraints for rooms, patients, and surgeons, facilitating rescheduling and enhancing operational flexibility. To solve such a problem, we introduce multiple algorithms based on a Random-Key Optimizer (RKO), coupled with relaxed formulations to compute lower bounds efficiently, rigorously tested on literature and new, real-world-based instances. The RKO approach decouples the problem from the solving algorithms through an encoding/decoding layer, making it possible to use the same solving algorithms to multiple room scheduling problems case studies from multiple hospitals, given the particularities of each place, even other optimization problems. Among the possible RKO algorithms, we design the heuristics Biased Random-Key Genetic Algorithm with <span><math><mi>Q</mi></math></span>-Learning, Simulated Annealing, and Iterated Local Search for use within an RKO framework, employing a single decoder function. The proposed heuristics, complemented by the lower-bound formulations, provided optimal gaps for evaluating the effectiveness of the heuristic results. Our results demonstrate significant lower- and upper-bound improvements for the literature instances, notably in proving one optimal result. Our strong statistical analysis shows the effectiveness of our implemented heuristic search mechanisms. Furthermore, the best-proposed heuristic efficiently generates schedules for the newly introduced instances, even in highly constrained scenarios. This research offers valuable insights and practical solutions for improving surgery scheduling processes, delivering tangible benefits to hospitals by optimizing resource allocation, reducing patient wait times, and enhancing overall operational efficiency.</div></div>","PeriodicalId":50737,"journal":{"name":"Applied Soft Computing","volume":"180 ","pages":"Article 113368"},"PeriodicalIF":7.2,"publicationDate":"2025-06-11","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144288754","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}