Article
Computer Science, Hardware & Architecture
Juan-juan Fu, Xing-lan Zhang
Summary: This paper proposes a feature fusion technique based on gradient importance enhancement to improve the accuracy and generalization ability of the intrusion detection model in the current unstable network security situation.
Article
Environmental Sciences
Qi Zhang, Yao Lu, Sicheng Shao, Li Shen, Fei Wang, Xuetao Zhang
Summary: Remote sensing change detection aims to identify changed pixels from two temporal images of the same location. Existing models use encoder-decoder structures and Siamese networks, but face challenges related to symmetry of change features, underutilization of the encoder parameters, and problems with sample balance and edge region detection. To address these issues, this paper proposes a mutual feature-aware network (MFNet) that includes a symmetric change feature fusion module, a mutual feature-aware module, and a loss function for edge regions. Experimental results demonstrate the effectiveness of MFNet, outperforming advanced algorithms with F1 scores of 83.11% and 91.52% on SYSU-CD and LEVIR-CD datasets, respectively.
Article
Medicine, General & Internal
Ali H. Al-Timemy, Laith Alzubaidi, Zahraa M. Mosa, Hazem Abdelmotaal, Nebras H. Ghaeb, Alexandru Lavric, Rossen M. Hazarbassanov, Hidenori Takahashi, Yuantong Gu, Siamak Yousefi
Summary: In this study, a deep learning model is proposed to accurately and robustly detect early clinical keratoconus (KCN). By extracting features from three different corneal maps using Xception and InceptionResNetV2 deep learning architectures, and then fusing the features, subclinical forms of KCN can be detected with high accuracy. The model achieved an AUC of 0.99 and an accuracy range of 97-100% in distinguishing normal eyes from eyes with subclinical and established KCN. The model was further validated on an independent dataset with an AUC of 0.91-0.92 and an accuracy range of 88-92%. This model is a step toward improving the detection of clinical and subclinical forms of KCN.
Article
Automation & Control Systems
Haifeng Wang, Lvjiyuan Jiang, Qian Zhao, Hao Li, Kai Yan, Yang Yang, Songlin Li, Yungang Zhang, Lianliu Qiao, Cuilian Fu, Hong Yin, Yun Hu, Haibin Yu
Summary: Deep learning-based target detection techniques have had a significant impact on daily life, with the feature pyramid being a widely utilized method for multiscale target detection. However, issues such as multiscale feature alignment and non-local feature fusion exist in the pyramid structure. The proposed progressive network structure effectively addresses these problems and shows improved performance compared to other state-of-art methods.
ENGINEERING APPLICATIONS OF ARTIFICIAL INTELLIGENCE
(2021)
Article
Computer Science, Artificial Intelligence
Jin Zhang, Qiuwei Liang, Qianqian Guo, Jinyu Yang, Qing Zhang, Yanjiao Shi
Summary: The article introduces a Residual Refinement Network (R(2)Net) method for salient object detection, which improves the performance of salient object detection through the fusion strategy of multi-scale features and contextual features. Experimental results demonstrate that the proposed method performs excellently on multiple benchmark datasets.
IMAGE AND VISION COMPUTING
(2022)
Article
Chemistry, Analytical
Faseela Abdullakutty, Pamela Johnston, Eyad Elyan
Summary: This paper presents a feature-fusion method that combines features extracted by pre-trained deep learning models with traditional color and texture features to improve the performance of face presentation attack detection. Extensive experiments show that enriching the feature space can enhance the detection rate, opening up future research directions for exploring new characterizing features and fusion strategies.
Article
Computer Science, Artificial Intelligence
Shufeng Xiong, Guipei Zhang, Vishwash Batra, Lei Xi, Lei Shi, Liangliang Liu
Summary: Compared to ordinary news, fake news spreads faster with lower production cost, causing significant social harm. Detecting fake news efficiently and accurately has become a research focus due to these reasons. We propose a Two-Round Inconsistency-based Multi-modal fusion Network (TRIMOON) for fake news detection, consisting of feature extraction, fusion, and classification modules. By performing two-fold inconsistency detection, we effectively filter noise generated during the fusion process. Experimental results demonstrate the superiority of our TRIMOON model over state-of-the-art approaches on Chinese and English datasets.
INFORMATION FUSION
(2023)
Article
Computer Science, Artificial Intelligence
Ioannis Kansizoglou, Loukas Bampis, Antonios Gasteratos
Summary: One of the prominent attributes of Neural Networks is their ability to extract robust and descriptive features from high dimensional data. However, neural networks induce biases that are difficult to deal with and lack knowledge for describing the intra-layer properties, which limits the applicability of the extracted features. This paper presents a novel way of visualizing and understanding the vector space before the output layer of neural networks, aiming to enlighten the properties of deep feature vectors under classification tasks and improve the obtained results.
IEEE TRANSACTIONS ON PATTERN ANALYSIS AND MACHINE INTELLIGENCE
(2022)
Article
Computer Science, Information Systems
Vinayakumar Ravi, Rajasekhar Chaganti
Summary: A survey of literature reveals that using deep learning-based models for image classification, specifically applying convolutional neural networks (CNN), has been recognized as a significant direction for malware detection and classification. The image-based Android malware detection is effective in detecting both unpacked and packed malware. In this study, various CNN-based pretrained models are employed and their features are extracted and fused for classification using a meta-classifier.
MULTIMEDIA TOOLS AND APPLICATIONS
(2023)
Article
Automation & Control Systems
Chong Ma, Liguo Weng, Min Xia, Haifeng Lin, Ming Qian, Yonghong Zhang
Summary: This paper proposes a dual-branch network for change detection, which mitigates detection errors, omissions, and obtains sharper edges through the use of multi-scale strip convolution module, spatial attention module, and feature fusion network. Experimental results demonstrate that the proposed method outperforms other algorithms.
ENGINEERING APPLICATIONS OF ARTIFICIAL INTELLIGENCE
(2023)
Article
Multidisciplinary Sciences
Zhaoguo Li, Xiumei Wei, M. Hassaballah, Xuesong Jiang
Summary: This paper presents a deep learning model for industrial defect detection, which achieves effective feature fusion through the design of a separated prediction head and enhanced feature extraction structures. Extensive experiments on three public datasets demonstrate its competitive performance in industrial defect detection.
ADVANCED THEORY AND SIMULATIONS
(2023)
Article
Computer Science, Artificial Intelligence
Kunfeng Wang, Yadong Wang, Shuqin Zhang, Yonglin Tian, Dazi Li
Summary: This article proposes a self-learning multi-scale object detection network, named SLMS-SSD, which balances the semantic information and spatial information to effectively improve the accuracy of object detection, especially for small object detection.
EXPERT SYSTEMS WITH APPLICATIONS
(2022)
Article
Computer Science, Information Systems
Chao Dai, Chen Pan, Wei He
Summary: This paper proposes a feature extraction and fusion network (EFNet) that effectively integrates high-level semantic features and low-level image features, improving the performance of salient object detection.
MULTIMEDIA TOOLS AND APPLICATIONS
(2022)
Article
Chemistry, Analytical
Yi Li, Lingna Wang, Zeji Wang
Summary: In this study, a feature-enhancement- and channel-attention-guided single-shot detector (FCSSD) was proposed to improve object detection performance. By utilizing four modules, contextual and semantic information were explored, multi-scale features were refined, and channel weights were balanced, resulting in excellent detection performance for multi-scale object detection.
Article
Engineering, Multidisciplinary
Chao Zhao, Xin Shu, Xi Yan, Xin Zuo, Feng Zhu
Summary: Steel surfaces may have defects due to imperfect manufacturing techniques and external factors, which significantly impact their lifespan and usability. Therefore, surface defect detection is crucial in industrial production. However, traditional detection algorithms suffer from low accuracy and speed. In this study, we propose a model called RDD-YOLO, which is based on YOLOv5, for steel surface defect detection. Our model utilizes Res2Net blocks for feature extraction, a double feature pyramid network for enhanced representations, and a decoupled head for improved detection precision. Experimental results demonstrate that RDD-YOLO achieves accuracies of 81.1 mAP on NEU-DET and 75.2 mAP on GC10-DET, surpassing YOLOv5 by 4.3% and 5.8% respectively. Our proposed model shows comprehensive performance in steel surface defect detection.
Article
Mathematics, Interdisciplinary Applications
Jinshuai Bai, Timon Rabczuk, Ashish Gupta, Laith Alzubaidi, Yuantong Gu
Summary: In this paper, a modified loss function called LSWR loss function is proposed for the Physics-Informed Neural Network (PINN) in computational solid mechanics. Through testing and comparison in 2D and 3D problems, the effectiveness, robustness, and accuracy of the PINN based on the LSWR loss function in predicting displacement and stress fields are demonstrated.
COMPUTATIONAL MECHANICS
(2023)
Review
Computer Science, Information Systems
Sabah Abdulazeez Jebur, Khalid A. Hussein, Haider Kadhim Hoomod, Laith Alzubaidi, Jose Santamaria
Summary: Due to the continuous advancement of technology, human behavior detection and recognition have become important research in computer vision. Anomaly detection (AD) is one of the most challenging problems in this field due to the complex environment and difficulty in extracting specific features. Deep learning, particularly convolution neural networks and recurrent neural networks, have shown excellent performance in addressing AD tasks and other domains. This review provides a comprehensive overview of DL methods and architectures for video AD, including different classifications of anomalies, analysis of DL methods, and discussion of applications and future directions.
Article
Engineering, Multidisciplinary
Jinshuai Bai, Hyogu Jeong, C. P. Batuwatta-Gamage, Shusheng Xiao, Qingxia Wang, C. M. Rathnayaka, Laith Alzubaidi, Gui-Rong Liu, Yuantong Gu
Summary: This work extends the Physics-informed neural network (PINN) method to computational solid mechanics problems. Various formulation and programming techniques for implementing the governing equations are investigated, and two commonly used loss functions for PINN-based computational solid mechanics are examined. Numerical examples demonstrate the performance of PINN-based computational solid mechanics from 1D to 3D solid problems. The provided Python programs with step-by-step explanations can be extended for more challenging applications.
INTERNATIONAL JOURNAL OF COMPUTATIONAL METHODS
(2023)
Article
Chemistry, Analytical
Methaq A. Shyaa, Zurinahni Zainol, Rosni Abdullah, Mohammed Anbar, Laith Alzubaidi, Jose Santamaria
Summary: Concept drift refers to the change in the statistical distribution of data over time in data streaming scenarios. This article proposes an extended variant of the genetic programming combiner (GPC) to handle concept drift in data stream classification. Experimental results demonstrate that the proposed method outperforms traditional GPC and other existing methods in handling various types of concept drift.
Article
Chemistry, Analytical
A. H. Alamoodi, O. S. Albahri, A. A. Zaidan, H. A. Alsattar, B. B. Zaidan, A. S. Albahri, Amelia Ritahani Ismail, Gang Kou, Laith Alzubaidi, Mohammed Talal
Summary: An intelligent remote prioritization method is proposed in this research for patients with high-risk multiple chronic diseases, based on emotion and sensory measurements and multi-criteria decision making. The methodology consists of a case study using a multi-criteria decision matrix and a modified technique for reorganizing opinion order. The results highlight the importance of chronic heart disease and emotion-based criteria, as well as the significance of Peaks as a sensor-based criterion and chest pain as an emotion criterion. Low blood pressure disease is identified as the most important criterion for patient prioritization, with severe cases being prioritized. The results are evaluated through systematic ranking and sensitivity analysis.
Article
Biology
Fouad H. Awad, Murtadha M. Hamad, Laith Alzubaidi
Summary: Big-medical-data classification and image detection are important in healthcare. Logistic regression and YOLOv4 have limitations with big medical data. We proposed a robust approach using logistic regression and YOLOv4, enhanced by parallel k-means pre-processing and a neural engine processor. Our approach accurately classified medical data and detected medical images, improving the performance of logistic regression and YOLOv4 and offering a promising solution for healthcare.
Review
Health Care Sciences & Services
A. S. Albahri, Z. T. Al-qaysi, Laith Alzubaidi, Alhamzah Alnoor, O. S. Albahri, A. H. Alamoodi, Anizah Abu Bakar
Summary: The significance of deep learning techniques in SSVEP-based BCI applications is assessed through a systematic review. Relevant articles were gathered from three reliable databases and classified into five categories based on their deep learning methods. The study examines the findings and challenges in existing applications, and provides recommendations for researchers and developers.
INTERNATIONAL JOURNAL OF TELEMEDICINE AND APPLICATIONS
(2023)
Article
Computer Science, Theory & Methods
Laith Alzubaidi, Jinshuai Bai, Aiman Al-Sabaawi, Jose Santamaria, A. S. Albahri, Bashar Sami Nayyef Al-dabbagh, Mohammed. A. A. Fadhel, Mohamed Manoufali, Jinglan Zhang, Ali. H. H. Al-Timemy, Ye Duan, Amjed Abdullah, Laith Farhan, Yi Lu, Ashish Gupta, Felix Albu, Amin Abbosh, Yuantong Gu
Summary: Data scarcity is a major challenge in training deep learning models due to the need for a large amount of labeled data. Manual labeling is costly and time-consuming, and many applications lack sufficient data for training. This paper presents a comprehensive overview of state-of-the-art techniques to address the issue of data scarcity in deep learning and provides recommendations for data acquisition and ensuring the trustworthiness of training datasets.
JOURNAL OF BIG DATA
(2023)
Article
Engineering, Multidisciplinary
Jinshuai Bai, Gui-Rong Liu, Ashish Gupta, Laith Alzubaidi, Xi-Qiao Feng, YuanTong Gu
Summary: Our study reveals that physics-informed neural networks (PINN) are often local approximators after training. This led to the development of a novel physics-informed radial basis network (PIRBN), which maintains the local approximating property throughout the training process. Unlike deep neural networks, PIRBN consists of only one hidden layer and a radial basis activation function. Under appropriate conditions, we demonstrated that PIRBNs can converge to Gaussian processes using gradient descent methods. Furthermore, we investigated the training dynamics of PIRBN using the neural tangent kernel (NTK) theory and explored various initialization strategies. Numerical examples showed that PIRBN is more effective than PINN in solving nonlinear partial differential equations with high-frequency features and ill-posed computational domains. Moreover, existing PINN numerical techniques such as adaptive learning, decomposition, and different loss functions can be applied to PIRBN. The reproducible code for all numerical results is available at https://github.com/JinshuaiBai/PIRBN.
COMPUTER METHODS IN APPLIED MECHANICS AND ENGINEERING
(2023)
Review
Computer Science, Artificial Intelligence
A. S. Albahri, Ali M. Duhaim, Mohammed A. Fadhel, Alhamzah Alnoor, Noor S. Baqer, Laith Alzubaidi, O. S. Albahri, A. H. Alamoodi, Jinshuai Bai, Asma Salhi, Jose Santamaria, Chun Ouyang, Ashish Gupta, Yuantong Gu, Muhammet Deveci
Summary: In recent years, there has been a significant shift in the healthcare sector towards embracing artificial intelligence (AI) to improve disease diagnosis accuracy and mitigate health risks. However, the development of trustworthy and explainable AI (XAI) in healthcare is still in its early stages. This study provides a systematic review of the trustworthiness and explainability of AI applications in healthcare, focusing on quality, bias risk, and data fusion, to offer more accurate insights and recommendations.
INFORMATION FUSION
(2023)
Review
Computer Science, Artificial Intelligence
Laith Alzubaidi, Aiman Al-Sabaawi, Jinshuai Bai, Ammar Dukhan, Ahmed H. Alkenani, Ahmed Al-Asadi, Haider A. Alwzwazy, Mohamed Manoufali, Mohammed A. Fadhel, A. S. Albahri, Catarina Moreira, Chun Ouyang, Jinglan Zhang, Jose Santamaria, Asma Salhi, Freek Hollman, Ashish Gupta, Ye Duan, Timon Rabczuk, Amin Abbosh, Yuantong Gu
Summary: This article provides a comprehensive review of the synergy between AI and DM, with a focus on the importance of trustworthiness. It addresses the need for trustworthy AI, the key requirements for establishing trustworthiness, methods of obtaining trustworthy data, and priorities for challenging applications. The article emphasizes the necessity of addressing trustworthiness in AI systems before deployment.
INTERNATIONAL JOURNAL OF INTELLIGENT SYSTEMS
(2023)
Meeting Abstract
Ophthalmology
Rossen M. Hazarbassanov, Laith Al-Zubaidi, Zahraa M. Mosa, Hazem Abdul Mutal, Alexandru Lavric, Hidenori Takahashi, Taneri Taneri, Siamak Yousefi, Ali Al-Timemy
INVESTIGATIVE OPHTHALMOLOGY & VISUAL SCIENCE
(2023)
Meeting Abstract
Ophthalmology
Rossen Mihaylov Hazarbassanov, Laith Al-Zubaidi, Zahraa Mosa, Hazem Abdul Mutal, Alexandru Lavric, Hidenori Takahashi, Suphi Taneri, Siamak Yousefi, Ali Al-Timemy
INVESTIGATIVE OPHTHALMOLOGY & VISUAL SCIENCE
(2023)