Applied Intelligence最新文献

{"title":"GPSP-CLIP: learning generic pseudo-state prompts for flexible zero-shot anomaly detection","authors":"Weiyu Hu,&nbsp;Shubo Zhou,&nbsp;Yongbin Gao,&nbsp;Xue-Qin Jiang","doi":"10.1007/s10489-025-06843-1","DOIUrl":"10.1007/s10489-025-06843-1","url":null,"abstract":"<div><p>Large-scale foundation models such as Contrastive Language-Image Pre-training (CLIP) have shown great potential in zero-shot anomaly detection (ZSAD) task, allowing a single model to generalize to unseen categories without fine-tuning on specific classes. However, existing ZSAD methods often rely on rigid prompt designs, which makes them difficult to adapt to the diverse characteristics of industrial products. Additionally, the need to manually define category-specific and state-specific prompts limits their scalability and generalization. This paper proposes a generic pseudo-state prompting model based on CLIP (<i>GPSP-CLIP</i>) to address these challenges. The motivation behind <i>GPSP-CLIP</i> is to develop a flexible prompting method capable of representing normal and anomalous conditions across various applications without relying on predefined text prompts. Technically, <i>GPSP-CLIP</i> employs fully learnable parameters to generate broad, pseudo-state text features, enabling generalization across different industrial contexts. By employing distinct prompt learning strategies for anomaly classification and segmentation, <i>GPSP-CLIP</i> optimizes each task independently. This enables the model to effectively capture high-level semantics through global prompts while identifying fine-grained defect patterns via local prompts. Experimental results on the well-known MVTec and VisA datasets demonstrate improved performance, with a 1.8% improvement in AP for anomaly classification and a 1.3% gain in AUPRO for anomaly segmentation compared to state-of-the-art methods.</p></div>","PeriodicalId":8041,"journal":{"name":"Applied Intelligence","volume":"55 15","pages":""},"PeriodicalIF":3.5,"publicationDate":"2025-10-07","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145256335","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"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":"AI-crafted narratives: an empirical study on generating interactive stories using generative pre-training transformers","authors":"Ana Carolina de Souza Mendes,&nbsp;Mason Adsero,&nbsp;Joshua Palicka,&nbsp;Nurulla Zholdoshov,&nbsp;Xin Zhao","doi":"10.1007/s10489-025-06833-3","DOIUrl":"10.1007/s10489-025-06833-3","url":null,"abstract":"<div><p>Interactive storytelling, which has long captivated audiences, is embracing evolution. Dungeons &amp; Dragons, a timeless example of open-ended interactive storytelling, has entered the digital realm, with platforms such as AI Dungeon employing large-language models to enhance the experience. Amid the growing utilization of Chat-GPT and concerns regarding AI’s potential to replace jobs, this empirical study examines the capability of GPT-3.5 and GPT-4 to autonomously generate an engaging interactive narrative, alongside the necessary Twine code for compiling the story. To assess the quality of the generated story, user perceptions of its authorship, and its future, an accompanying survey was conducted. The result shows that even though some concerns arise regarding authorship, AI is promising in the field of generating interactive stories.</p></div>","PeriodicalId":8041,"journal":{"name":"Applied Intelligence","volume":"55 15","pages":""},"PeriodicalIF":3.5,"publicationDate":"2025-10-06","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145256562","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"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":"Real-time task scheduling strategy for 3D printing cloud platforms in health scenes","authors":"Jianjia He,&nbsp;Jian Wu,&nbsp;Jingran Ni,&nbsp;Yuning Zhang,&nbsp;Keng Leng Siau","doi":"10.1007/s10489-025-06907-2","DOIUrl":"10.1007/s10489-025-06907-2","url":null,"abstract":"<div><p>In health scenes, 3D Printing Cloud Platform (3DPCP) needs to cope with unpredictable fluctuations in tasks and resources, but traditional scheduling methods have problems such as incomplete consideration of factors, poor optimization, and weak dynamic adaptability, which make it difficult to meet real-time scheduling requirements. To this end, the real-time task scheduling problem of 3DPCP for health scenes is defined, a real-time task scheduling model is established, the design time of user personalized services is considered, a rescheduling scheme is designed in combination with task variations and device variations, and a scheduling strategy that incorporates dynamic mechanisms and improved multi-objective greywolf optimization algorithms is proposed in order to minimize the integrated scheduling cost and the average delivery time of the product. The findings of simulation experiments show that when equipment changes are not considered, compared with the optimal heuristic algorithm in this field, the average cost of the proposed algorithm is reduced by 2014.1 yuan, and the average delivery time is shortened by 1.52 h. When equipment changes are considered, compared with the multi-objective Genetic Algorithm Dynamic Strategies (GADS), the average cost of the proposed algorithm is reduced by 2984.57 yuan, and the average delivery time is shortened by 0.39 h, which validates the effectiveness of the proposed method.</p></div>","PeriodicalId":8041,"journal":{"name":"Applied Intelligence","volume":"55 15","pages":""},"PeriodicalIF":3.5,"publicationDate":"2025-10-04","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145210456","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"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":"Research on full lifecycle health management of permanent magnet synchronous electric drum driven by digital twin with dynamic update","authors":"Wei Chen,&nbsp;Weimin Wu,&nbsp;Mei Zhang,&nbsp;Huashun Li,&nbsp;Qing Shi","doi":"10.1007/s10489-025-06910-7","DOIUrl":"10.1007/s10489-025-06910-7","url":null,"abstract":"<div><p>Currently, research on fault Prognostics and Health Management (PHM) based on Digital Twin mainly focuses on integrating real-time data from various sources to facilitate comprehensive product inspection and health management. However, existing DT research faces three main theoretical bottlenecks: the lack of dynamic evolution mechanisms in multi-physics coupled modeling, static models’ difficulty in adapting to the drift of equipment degradation characteristics, and health status assessment’s reliance on prior fault samples. To address these issues, this paper proposes a comprehensive lifecycle dynamic management method for the Tubular Permanent Magnet Synchronous Electric Drum (TPMSED), by constructing a Dual-service Lifecycle Management Digital Twin Model (DSL-DT) that achieves deep integration of physical entities and virtual spaces. Firstly, a multi-physics coupled dynamic model is established, integrating the nonlinear interactions of electromagnetic fields, temperature fields, and dynamic fields. This is achieved through a combination of finite element simulation and data-driven approaches, addressing the challenge of characterizing equipment performance degradation under complex operating conditions. Secondly, an innovative dual dynamic adjustment mechanism for compensator updates and parameter updates is designed, utilizing ridge regression algorithms and adaptive gradient algorithms to achieve online optimization of model parameters, effectively suppressing model mismatch during the degradation process. Lastly, a health index (HI)-based state assessment method is proposed, which triggers model updates by comparing characteristic deviations with thresholds, significantly enhancing the accuracy of Remaining Useful Life (RUL) predictions. Experimental validation on an intelligent conveying system development platform demonstrates that this method can accurately track the performance evolution of equipment throughout its lifecycle, providing a new theoretical paradigm and technical pathway for health management of complex electromechanical systems.</p></div>","PeriodicalId":8041,"journal":{"name":"Applied Intelligence","volume":"55 15","pages":""},"PeriodicalIF":3.5,"publicationDate":"2025-10-04","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145210795","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"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 3D efficient and essentialized swin transformer network for alzheimer’s disease diagnosis","authors":"Shengchao Huang,&nbsp;Qun Dai","doi":"10.1007/s10489-025-06884-6","DOIUrl":"10.1007/s10489-025-06884-6","url":null,"abstract":"<div><p>Deep learning methods (e.g., convolutional neural networks, CNNs) have been widely applied to Alzheimer’s disease diagnosis based on structural magnetic resonance imaging (sMRI) data. However, CNN-based methods face significant Limitations in capturing the global feature distribution of the whole brain. Transformer-based models have shown promise in addressing this issue, but they often sacrifice local feature sensitivity. Moreover, the large number of parameters in Transformer-based models results in a strong dependence on large-scale datasets, which is difficult to satisfy in real-world 3D medical imaging scenarios. Through comprehensive consideration, we propose a 3D Efficient and Essentialized Swin Transformer Network (E2STN) to strike a balance between being lightweight and comprehensive feature extraction, thereby boosting Alzheimer’s disease diagnosis performance in 3D dataset scenarios. Specifically, E2STN includes four modules: an Efficient Swin Transformer (EST) module for identifying global structural information and being lightweight to reduce reliance on large-scale datasets, which is a novel task-oriented Transformer architecture; a Focused Feature Enhancement Convolution Unit (FFE-CU) for enhancing lesion details, thereby compensating for the limited perception of fine-grained pathological information by the Transformer; a Disease Risk Map generator (DRMg) for visualizing pathological regions; and an ROI-based classifier for precise categorization. Our proposed method has been validated by two diagnosis tasks (i.e., Alzheimer’s disease diagnosis and mild cognitive impairment conversion prediction) on the ADNI dataset. Compared to several state-of-the-art methods, our model demonstrates superior performance.</p></div>","PeriodicalId":8041,"journal":{"name":"Applied Intelligence","volume":"55 15","pages":""},"PeriodicalIF":3.5,"publicationDate":"2025-10-04","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145210468","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"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":"Feature selection based on information entropy with variable precision fuzzy mixed granularity","authors":"Jiaxin Wang,&nbsp;Jingqian Wang,&nbsp;Xiaohong Zhang,&nbsp;Jun Liu","doi":"10.1007/s10489-025-06890-8","DOIUrl":"10.1007/s10489-025-06890-8","url":null,"abstract":"<div><p>Fuzzy rough set theory allows defining different fuzzy relationships for different attribute types to quantify the similarity between objects. Meanwhile, information entropy, a powerful tool for quantifying uncertainty, is further extended within this framework to fuzzy rough set-based information entropy. The granularity division of traditional fuzzy entropy usually relies on fuzzy similarity relationships. In this paper, we first define variable precision mixed fuzzy granularity, combine it with fuzzy entropy to construct the information entropy based on variable precision mixed fuzzy granularity, and define fuzzy granularity entropy (FGe), fuzzy granularity joint entropy (FGJe), fuzzy granularity conditional entropy (FGCe), and fuzzy granularity mutual information (FGMI), and study the relationship and related properties among them. Then the importance function for evaluating the importance of features is constructed using FGMI, which lays the foundation for the feature selection (FS) algorithm. To evaluate the performance of the algorithm, numerical experiments are conducted on 15 public datasets and compared with other algorithms. The experimental results show that the method shows good adaptability and FS ability for handling different types of data.</p></div>","PeriodicalId":8041,"journal":{"name":"Applied Intelligence","volume":"55 15","pages":""},"PeriodicalIF":3.5,"publicationDate":"2025-10-04","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145210702","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"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":"Question answering system based on the combination of large language model and knowledge graph","authors":"Jihong Wang,&nbsp;Yichen Zhang,&nbsp;Wei Liu","doi":"10.1007/s10489-025-06828-0","DOIUrl":"10.1007/s10489-025-06828-0","url":null,"abstract":"<div><p>In recent years, large language models (LLMs) have achieved remarkable progress in natural language processing. However, their application in question-answering systems continues to face challenges such as insufficient credibility and interpretability of responses, as well as high computational resource demands. To address these issues, this paper proposes a question-answering system that integrates knowledge graphs with lightweight LLMs. Specifically, a lightweight front-end model based on BERT and T5 is employed to extract and transform logical forms from natural language queries, which are then executed on a knowledge graph. Subsequently, a smaller-scale LLM generates credible and interpretable answers based on these query results. Experimental results show that the proposed method achieves F1 scores of 75.6% and 76.8% on the WebQSP and GRAILQA datasets, respectively, surpassing other representative approaches. Furthermore, integrating the extracted knowledge with the ChatGLM-6B model significantly improves answer quality, increasing ratings by 112.8% for simple questions and 77.4% for complex questions, thus validating the effectiveness of our approach.</p></div>","PeriodicalId":8041,"journal":{"name":"Applied Intelligence","volume":"55 15","pages":""},"PeriodicalIF":3.5,"publicationDate":"2025-10-04","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145210457","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"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":"Three-way density peak clustering in incomplete information systems","authors":"Zhao Li,&nbsp;Ju-sheng Mi,&nbsp;Lei-jun Li","doi":"10.1007/s10489-025-06911-6","DOIUrl":"10.1007/s10489-025-06911-6","url":null,"abstract":"<div><p>The absence of data in the information systems will result in uncertainty in the classification of objects, and the fringe region of the cluster in the three-way clustering reflects the uncertainty of the clustering results, which can fit the incomplete information system well. Consequently, this paper proposes a three-way clustering framework in incomplete information systems, drawing on the density peak clustering algorithm and the model of three-way decision. Firstly, this paper defines the reflexive binary relation in incomplete information systems and determines the center of the corresponding object class. Then, the density peak clustering algorithm is utilized to identify the optimal clustering centers among all object class centers. Subsequently, the membership degree and relative loss function matrix of the object under each cluster center are defined according to the distance relationship between each object and all cluster centers. Finally, the clustering rules are obtained by the minimum risk decision theory, and the initial clustering results are processed to meet the three-way clustering criteria. In the experimental section of this paper, two sets of experiments are designed to show the clustering accuracy of the proposed algorithm and the influence of parameters on the clustering results.</p></div>","PeriodicalId":8041,"journal":{"name":"Applied Intelligence","volume":"55 15","pages":""},"PeriodicalIF":3.5,"publicationDate":"2025-10-04","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145210467","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"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":"Learning promotion policies with attention-based deep Q-networks","authors":"Yingnan Xu,&nbsp;Xuchun Wu,&nbsp;Zhenjun Li,&nbsp;Congli Liu,&nbsp;Yansheng Zhang","doi":"10.1007/s10489-025-06914-3","DOIUrl":"10.1007/s10489-025-06914-3","url":null,"abstract":"<div><p>In financial services, personalized promotion strategies are critical for sustaining customer engagement and driving asset growth. We present FAT-DQN, a deep reinforcement learning framework for off-line environments that models sequential decision-making as a Markov Decision Process (MDP), where promotional actions influence future changes in customer assets under management (AUM). FAT-DQN extends the standard Deep Q-Network (DQN) architecture with a multi-head self-attention mechanism over promotion–reward histories augmented by learnable temporal encodings, and applies Feature-wise Linear Modulation (FiLM) to incorporate customer-segment embeddings. To improve robustness, we employ per-customer reward normalization and evaluate policies with both ranking-based metrics and counterfactual off-policy estimators. Empirical results on real promotion logs show that FAT-DQN consistently outperforms baseline methods, yielding a higher mean NDCG@3 (0.7744) compared to Batch-Constrained deep Q-learning (BCQ, 0.7325) and DQN (0.6852). It further improves alignment between predicted and realized outcomes, achieving a Spearman correlation of 0.2584, compared to 0.1619 for BCQ and 0.1522 for DQN. Counterfactual evaluations further show that FAT-DQN delivers consistently strong off-policy estimates, confirming its robustness across evaluation settings. These findings demonstrate that attention-based architectures with modulation offer a more effective and interpretable alternative to standard reinforcement learning approaches for personalized promotion planning in financial services.</p></div>","PeriodicalId":8041,"journal":{"name":"Applied Intelligence","volume":"55 15","pages":""},"PeriodicalIF":3.5,"publicationDate":"2025-10-03","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145210886","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"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":"Multimodal prompt learning with selective feature fusion: towards robust cross-modal alignment","authors":"Jiabao Han,&nbsp;Yahui Wang,&nbsp;Wei Zhong,&nbsp;Ying Zhang,&nbsp;Xichao Yuan","doi":"10.1007/s10489-025-06919-y","DOIUrl":"10.1007/s10489-025-06919-y","url":null,"abstract":"<div><p>Vision–language models (VLMs) have shown impressive transferability but still struggle with robustness and generalization when applied to downstream tasks with limited supervision. To address these challenges, we propose a Selective Feature Fusion (SFF) framework that adaptively suppresses noisy visual regions and reinforces task-relevant cross-modal cues through lightweight, learnable gating. Our approach integrates text-guided visual masking and image-aware textual calibration into a unified pipeline, enabling more discriminative and semantically aligned multimodal representations. Comprehensive evaluations across nine widely used benchmarks demonstrate that our method consistently surpasses strong prompt-learning baselines under both few-shot and base-to-novel generalization settings. In particular, under the 8-shot scenario, our approach achieves the best overall accuracy, maintaining a clear margin over representative methods such as CoCoOp and MaPLe. These results highlight not only the robustness of our design but also its effectiveness in capturing cross-modal semantics under data-limited conditions. Further analyses, including ablation studies and qualitative visualizations, confirm that the proposed gating and calibration modules are complementary and play indispensable roles in improving performance. Taken together, this work provides a simple yet powerful strategy for enhancing the adaptability and generalization of VLMs in real-world scenarios.</p></div>","PeriodicalId":8041,"journal":{"name":"Applied Intelligence","volume":"55 15","pages":""},"PeriodicalIF":3.5,"publicationDate":"2025-10-03","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145210240","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}