Computational Intelligence最新文献

{"title":"Click Fraud Detection With Recurrent Neural Networks Optimized by an Adapted Version of Variable Neighborhood Search Algorithm","authors":"Vico Zeljkovic,&nbsp;Miodrag Zivkovic,&nbsp;Fadi Al-Turjman,&nbsp;Jelena Gajic,&nbsp;Lepa Babic,&nbsp;Aleksandar Djordjevic,&nbsp;Milos Antonijevic,&nbsp;Nebojsa Bacanin","doi":"10.1111/coin.70283","DOIUrl":"https://doi.org/10.1111/coin.70283","url":null,"abstract":"<div>\u0000 \u0000 <p>The revenue generated from online ads has become quite significant, and as with the advancement in any sort of business, this one brings fraudsters with it. However, different forms of fraud can be performed as this work tackles the problems of click fraud in advertisements. In this case, the fraud can be performed by the party that bought the advertisement to boost its revenue, or by other malicious parties that tend to exhaust the resources for the said ad, for example. Due to this and many other scenarios, a robust solution for detecting such cases must be established. However, existing click fraud detection approaches either rely on static rule-based systems or deep learning models with manually tuned hyperparameters, which may result in limited adaptability and suboptimal performance in complex sequential environments. Therefore, there is a need for an adaptive optimization strategy capable of effectively tuning sequential models for improved fraud detection accuracy. This work proposes three different types of recurrent neural networks (RNNs) that are combined with the attention mechanism. Furthermore, in each of the three different experiments, the networks were optimized by strong metaheuristics optimizers, the results of which were compared to establish the strongest one. This was done with the purpose of confirming the improvements to the variable neighborhood search (VNS) algorithm, which was proposed by the authors in this work. The best synthesized RNN model tuned by the suggested modified optimizer attained accuracy of 0.806569, with Matthews correlation coefficient of 0.613209.</p>\u0000 </div>","PeriodicalId":55228,"journal":{"name":"Computational Intelligence","volume":"42 4","pages":""},"PeriodicalIF":1.9,"publicationDate":"2026-07-22","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148615675","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":4,"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":"Mutatis Mutandis: Revisiting the Comparator in Discrimination Testing","authors":"José M. Álvarez,&nbsp;Salvatore Ruggieri","doi":"10.1111/coin.70238","DOIUrl":"https://doi.org/10.1111/coin.70238","url":null,"abstract":"<p>Testing for individual discrimination involves deriving a profile, the comparator, similar to the one making the discrimination claim, the complainant, based on a protected attribute, such as race or gender, and comparing their decision outcomes. The complainant–comparator pair is central to discrimination testing. Most discrimination testing tools rely on this pair to establish evidence for discrimination. In this work, we revisit the role of the comparator in discrimination testing. We first argue for the inherent causal modeling nature of deriving the comparator. We then introduce a two-kind classification for the comparator: the <i>ceteris paribus</i> or “with all else equal” (CP) comparator and the <i>mutatis mutandis</i> or “with the appropriate adjustments being made” (MM) comparator. The CP comparator is the standard comparator, representing an idealized comparison for establishing discrimination as it aims for a complainant–comparator pair that only differs in membership in the protected attribute. As an alternative to the CP comparator, we define the MM comparator, which requires a comparator that represents the “what would have been” of the complainant without the effects of the protected attribute on the non-protected attributes. Under the MM comparator, the complainant–comparator pair can be dissimilar in terms of the non-protected attributes, departing from the idealized comparison imposed by the CP comparator. Notably, the MM comparator denotes a more complex object and its implementation offers an impactful venue for machine learning methods. We illustrate these two comparators and their impact on discrimination testing using a real-world example.</p>","PeriodicalId":55228,"journal":{"name":"Computational Intelligence","volume":"42 4","pages":""},"PeriodicalIF":1.9,"publicationDate":"2026-07-16","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://onlinelibrary.wiley.com/doi/epdf/10.1111/coin.70238","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148467423","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":4,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"HoloIntelligent: A Deep Learning Framework for Comparative Analysis of Holographic and Bright-Field Imaging in Head and Neck Carcinoma Classification","authors":"Asifa Nazir,&nbsp;Ahsan Hussain,&nbsp;Mandeep Singh,&nbsp;Deepika Mishra,&nbsp;Vivek Nayyar,&nbsp;Muzafar A. Macha,&nbsp;Assif Assad","doi":"10.1111/coin.70280","DOIUrl":"https://doi.org/10.1111/coin.70280","url":null,"abstract":"<div>\u0000 \u0000 <p>Head and Neck Cancer (HNC) remains a major global health challenge, with late-stage diagnosis limiting treatment outcomes. Conventional methods, including histopathology and standard imaging, are time-consuming, heavily reliant on expert interpretation, and introduce variability with diagnostic delays. To address these limitations, this study integrates digital holographic imaging with advanced Deep Learning (DL) architectures for automated HNC classification. A novel dataset comprising 3915 holographic images from 291 patients is utilized to evaluate two proposed models: a <i>Convolutional Block Attention Module (CBAM)-Integrated U-Net classifier</i> and a custom-designed <i>HoloIntelligent framework</i>. Class imbalance is addressed through synthetic oversampling, and performance is evaluated using both discriminative metrics and error-based measures. Both models achieved high classification accuracy, with the CBAM-Integrated U-Net attaining 97.29% and HoloIntelligent achieving 97.63%. HoloIntelligent outperformed the CBAM-Integrated U-Net in classifying both normal and abnormal cases, demonstrating superior discriminative performance on holographic data. To further validate the proposed models, ablation studies were conducted to analyze the contribution of each model component. The study is further strengthened through statistical evaluation and comparative analysis with corresponding bright-field data across multiple baseline models. In addition, the explainability analysis improved the interpretability and credibility of the findings. The study also discusses current limitations, key challenges, and future research directions. Finally, the study concludes with a summary of its major contributions. Overall, the findings demonstrate that HoloIntelligent achieves robust and reliable performance on holographic data, underscoring its potential as a valuable tool for early detection and clinical decision-making in HNC.</p>\u0000 </div>","PeriodicalId":55228,"journal":{"name":"Computational Intelligence","volume":"42 4","pages":""},"PeriodicalIF":1.9,"publicationDate":"2026-07-15","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148467399","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":4,"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":"Globally Valid Rule-Based Explanations for Black-Box Models","authors":"Van Quoc Phuong Huynh,&nbsp;Florian Beck,&nbsp;Josef Küng,&nbsp;Johannes Fürnkranz","doi":"10.1111/coin.70278","DOIUrl":"https://doi.org/10.1111/coin.70278","url":null,"abstract":"<p>Black-box explanation methods such as <span>Lime</span> define a neighborhood around the query example, learn an interpretable local surrogate model in this neighborhood, and use this local surrogate model for its explanation. <span>Lore</span> and <span>Anchors</span> are two such methods, which deliver local explanations in the form of IF-THEN rules. In this article, we argued that such explanations are incomplete because they can miss the specification of the local neighborhood in which they are valid. To counter this, we proposed <span>Glori</span>, an alternative approach which learns globally valid explanations for single examples using the recently proposed <span>Lord</span> rule learner. Our experimental evaluation confirms its improved fidelity, not only for the case when fidelity is measured on an algorithm-specific neighborhood. Moreover, <span>Glori</span> is considerably more efficient than its neighborhood-based alternatives.</p>","PeriodicalId":55228,"journal":{"name":"Computational Intelligence","volume":"42 4","pages":""},"PeriodicalIF":1.9,"publicationDate":"2026-07-14","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://onlinelibrary.wiley.com/doi/epdf/10.1111/coin.70278","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148467471","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":4,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Image Captioning Across Object Categories: A CNN-LSTM and BLEU Score-Based Study","authors":"Garima Salgotra,&nbsp;Pawanesh Abrol","doi":"10.1111/coin.70281","DOIUrl":"https://doi.org/10.1111/coin.70281","url":null,"abstract":"<div>\u0000 \u0000 <p>Image captioning aims to automatically generate semantic descriptions for visual data and is widely applied in assistive systems, content filtering, and recommendation engines. This paper presents a hybrid encoder–decoder architecture integrating Convolutional Neural Networks (CNNs) for visual feature extraction and Long Short-Term Memory (LSTM) networks for sequence generation. An attention mechanism is incorporated to dynamically weight salient image regions during decoding, improving contextual representation and object interaction modeling. The proposed model is trained and evaluated on the Flickr8k dataset, categorized by object complexity. Performance is assessed using BLEU, METEOR, and ROUGE metrics. Experimental results demonstrate superior performance on low-complexity images, achieving BLEU scores of 0.61, 0.60, and 0.38 for single-, double-, and multi-object images, respectively. Comparative analysis indicates that the attention-enhanced model outperforms standard approaches, particularly in complex scenes.</p>\u0000 </div>","PeriodicalId":55228,"journal":{"name":"Computational Intelligence","volume":"42 4","pages":""},"PeriodicalIF":1.9,"publicationDate":"2026-07-12","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148466981","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":4,"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-Viewed Graph Representation Learning Through Graph Neural Network and Rich-Spatial Local Feature Embedding","authors":"Phu Pham","doi":"10.1111/coin.70279","DOIUrl":"https://doi.org/10.1111/coin.70279","url":null,"abstract":"<div>\u0000 \u0000 <p>For many years, graph representation learning plays a pivotal role in bioinformatics and cheminformatics; as a result, supporting a wide range of tasks such as drug discovery, toxicity prediction, and compound–protein interaction analysis. However, existing approaches often focus solely on either sequential molecular fingerprints or graph-based structural features, which limit their ability to capture both local chemical substructures and global molecular topology. To address this issue, we propose MM2Vec, a novel multi-viewed molecular representation learning framework that integrates local rich-feature embedding with graph neural network (GNN)-based structural learning. Specifically, each molecular graph is first processed through an MLP-based embedding layer that encodes sub-structural fingerprint information extracted from radius-based subgraphs, capturing fine-grained chemical and physiochemical features. Simultaneously, a multi-layered GNN encoder learns topological relationships from the molecular graph structure; therefore, focusing more on geometric and relational information among atoms. The outputs from both embedding branches are then fused using a learnable linear mechanism to produce unified, high-quality molecular embeddings in a shared latent space. These fused representations are used to drive task-specific prediction layers for addressing various learning objectives. We validate the proposed MM2Vec model on multiple graph learning tasks, including drug-induced liver injury (DILI) classification and lethal dose (LD) molecular regression problems. Experimental results show that MM2Vec consistently outperforms classical machine learning (ML)-based models and recent state-of-the-art deep learning (DL)/GNN-based methods in terms of accuracy, robustness, and generalization. Our findings in this highlight the importance of combining both sub-structural and graph-structural perspectives and demonstrate the versatility and effectiveness of our MM2Vec model for a wide range of molecular analysis tasks.</p>\u0000 </div>","PeriodicalId":55228,"journal":{"name":"Computational Intelligence","volume":"42 4","pages":""},"PeriodicalIF":1.9,"publicationDate":"2026-07-06","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148462756","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":4,"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":"Balanced Fidelity Network for Blind Hyperspectral Unmixing","authors":"Wen Yan,&nbsp;Jie Huang","doi":"10.1111/coin.70275","DOIUrl":"https://doi.org/10.1111/coin.70275","url":null,"abstract":"<div>\u0000 \u0000 <p>Hyperspectral image (HSI) unmixing is a popular field in HSI processing. Its main task is to decompose mixed pixels into pure spectral signatures called <i>endmembers</i> and their corresponding fractions called <i>abundances</i>. Some unmixing algorithms introduce subspaces and use regularizations to characterize the properties of subspaces to improve the estimated abundance accuracy. While we shift our focus to the balance of the original space and subspace information. We propose a balanced fidelity regularization to realize it. The subspace denoted by a projected matrix is learned through the proposed network without extra trial of designing specialized regularization. The proposed network consists of two parts. One is an autoencoder, estimating abundances, endmembers, and reconstructions, and the other is a convolutional layer projecting the reconstructions and the input HSI into a subspace. Specifically, the balanced fidelity term ensuring the consistence in both original space and the projected subspace to exploit and balance the information in two spaces. In addition, an adapted contraction regularization and a sparse regularization are incorporated to the loss function to further improve the unmixing performance. We conduct simulated and real-data experiments to show that the proposed network greatly improves the estimation accuracy of endmembers and abundances compared with some state-of-the-art algorithms.</p>\u0000 </div>","PeriodicalId":55228,"journal":{"name":"Computational Intelligence","volume":"42 4","pages":""},"PeriodicalIF":1.9,"publicationDate":"2026-07-05","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148462870","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":4,"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 Misalignment: A Critical Issue in Cross-Architecture KD","authors":"Gousia Habib,&nbsp;Gaurav Harjule,&nbsp;I. A. Malik","doi":"10.1111/coin.70102","DOIUrl":"https://doi.org/10.1111/coin.70102","url":null,"abstract":"<div>\u0000 \u0000 <p>This article addresses the challenge of deploying complex deep learning models on resource-constrained edge devices for real-time security applications. The core objective is to transfer the capabilities of a high-computation model to a lightweight model through feature-based KD (KD) and optimization techniques like quantization and pruning. The teacher model used for experiments was DeiT-base, pretrained on ImageNet. It was then fine-tuned to custom datasets. Student model, MobileNetV3 Large was trained from scratch using KD on custom datasets tailored for security tasks, including Eyegaze, Emotions, and Weapons. Through KD, the student model learned to replicate the teacher model's performance efficiently. Experiments demonstrated that the optimized student model performs effectively in real-time analysis of images and videos on edge devices. This provides a practical tool for security forces in threat detection and surveillance. The findings highlight the potential of KD and quantization to make advanced machine learning models deployable in real-world security applications, offering enhanced performance in resource-limited environments. Future work will explore multitask learning, quantization-aware training, MobileNetV4 applications, custom dataset development, and innovative loss functions for improved cross-architecture distillation.</p>\u0000 </div>","PeriodicalId":55228,"journal":{"name":"Computational Intelligence","volume":"42 4","pages":""},"PeriodicalIF":1.9,"publicationDate":"2026-06-30","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148386934","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":4,"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":"ODE-Based MAPPO Optimization for Guard Strategies in Artificial Attack Scenarios","authors":"Nengfei Cui,&nbsp;Haoyang Li,&nbsp;Zheng Yang,&nbsp;Zhicheng Dong","doi":"10.1111/coin.70243","DOIUrl":"https://doi.org/10.1111/coin.70243","url":null,"abstract":"<div>\u0000 \u0000 <p>This paper presents a reinforcement learning-based framework for optimizing guard strategies in artificial attack scenarios. The environment is modeled as a multi-agent confrontation game, in which guards collaborate to intercept an attacker while protecting evacuating pedestrians. Pedestrian movement is governed by an extended floor field model, reflecting panic-induced behavior under threat. To address the coordination and adaptability challenges in such high-stakes dynamic settings, we adopt the ordinary differential equation (ODE)-based Multi-Agent Proximal Policy Optimization (MAPPO) algorithm under a centralized training and decentralized execution paradigm. By training agents through iterative interaction with the environment, the proposed approach enables guards to learn robust and efficient defense policies. Simulation results demonstrate that with the MAPPO-trained policies, guards can effectively constrain the attacker, minimize pedestrian casualties, and enable efficient cooperative defense.</p>\u0000 </div>","PeriodicalId":55228,"journal":{"name":"Computational Intelligence","volume":"42 4","pages":""},"PeriodicalIF":1.9,"publicationDate":"2026-06-30","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148343355","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":4,"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 Novel Method for Drug Application on Signaling Pathways With Concurrent Faults Using PBNs","authors":"Tapan Chowdhury,&nbsp;Ekarsi Lodh,&nbsp;Shalini Majumder","doi":"10.1111/coin.70276","DOIUrl":"https://doi.org/10.1111/coin.70276","url":null,"abstract":"<div>\u0000 \u0000 <p>Cellular processes are tightly regulated by various growth and transcription factors through information fluxes. They initiate cellular division and involve interactions between proteins termed signaling pathways. Dysregulated signaling can contribute to proliferative cellular states, and such pathway-level alterations can be approximated using probabilistic network models. In this study, Boolean pathway representations are extended with probabilistic interaction weights to approximate uncertainty in pathway-level protein–protein interactions, under simulated dysregulated signaling conditions. The simulated proliferative output states can be reduced under selected modeled drug-combination scenarios. This study focuses on the signaling pathways' probabilistic nature, featuring concurrent multiple faults. Initially, modeling the signaling pathways to their respective Probabilistic Boolean Networks (PBNs), we observed their behavior in the presence of concurrent faults, followed by conventional drug therapy to mitigate the effect of these faults. The proposed framework introduces a <i>Condensed_Probabilistic_Score (CPS)</i>, which ranks modeled drug combinations according to their ability to reduce proliferative output states across simulated fault scenarios without requiring prior specification of the exact fault combination. Additionally, we propose a strategy for prioritizing candidate custom target nodes whose combinations produced higher CPS values than the modeled known-drug combinations in the simulated PBN framework. Candidate combinations based on these custom target nodes are further evaluated as in silico intervention hypotheses. These findings should be interpreted within the assumptions of the abstracted PBN model and require biological validation using independent pathway resources and experimental perturbation studies.</p>\u0000 </div>","PeriodicalId":55228,"journal":{"name":"Computational Intelligence","volume":"42 4","pages":""},"PeriodicalIF":1.9,"publicationDate":"2026-06-30","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148386892","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":4,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}