Article
Computer Science, Information Systems
Yuna Han, Byung-Woo Hong
Summary: The novel efficient Fourier convolutional neural network achieves competitive accuracy with traditional convolutional neural networks by using a new activation function, eliminating unnecessary processes, and reducing the number of learnable parameters. It performs with higher accuracy in both shallow and deep neural networks, while dramatically reducing the number of parameters. The proposed methods have the potential to be applied in all architecture based on convolutional neural networks.
Article
Computer Science, Artificial Intelligence
Jakub Zak, Anna Korzynska, Antonina Pater, Lukasz Roszkowiak
Summary: In this study, the combination of the Fourier Transform and Convolutional Neural Network methods was used to classify images in multiple datasets. By incorporating Fourier Transform Layer, the processing speed was increased without sacrificing accuracy, providing an alternative approach to Convolutional Neural Networks that reduces the need for GPU training. Experimental results showed that models with the proposed layer achieved comparable test accuracy to convolutional models for images of size 128 x 128 and larger, with a minimum of 27% reduction in training time per one epoch on Central Processing Unit.
APPLIED SOFT COMPUTING
(2023)
Article
Computer Science, Artificial Intelligence
Yong Wang, Xian Wei, Xuan Tang, Keping Yu, Lingkun Luo
Summary: In this paper, we propose a method for robust visual tracking by adaptively integrating information from RGB and thermal videos. We use convolutional neural network (CNN) representation with random projection to describe RGB and thermal images, and develop an adaptive fusion strategy based on a period of time to optimize the reliable weights of different modalities. Extensive experiments against other state-of-the-art methods demonstrate the effectiveness of the proposed method, and through analyzing quantitative tracking results, we provide basic insights in RGB and thermal data tracking.
EXPERT SYSTEMS WITH APPLICATIONS
(2023)
Article
Environmental Sciences
Cuiping Shi, Xinlei Zhang, Jingwei Sun, Liguo Wang
Summary: A lightweight convolutional neural network with multi-level feature fusion has been proposed, which maximizes information extraction and avoids information loss, achieving higher classification accuracy and lower model complexity, realizing a trade-off between model accuracy and running speed.
Article
Computer Science, Information Systems
Ben Chen, Feiwei Qin, Yanli Shao, Jin Cao, Yong Peng, Ruiquan Ge
Summary: The study proposes a novel method for diagnosing leukemia by classifying white blood cells in bone marrow using the WBC-GLAformer model. The model combines the features of convolutional neural networks and transformers to enrich the features and improve classification accuracy by selecting discriminative regions.
JOURNAL OF KING SAUD UNIVERSITY-COMPUTER AND INFORMATION SCIENCES
(2023)
Article
Computer Science, Artificial Intelligence
Gao Huang, Zhuang Liu, Geoff Pleiss, Laurens van der Maaten, Kilian Q. Weinberger
Summary: Recent work has shown that adding shorter connections in convolutional networks can make the network deeper, more accurate, and more efficient in training. This paper introduces Dense Convolutional Network (DenseNet), which connects each layer to every other layer in a feed-forward manner. DenseNets alleviate the vanishing-gradient problem, encourage feature reuse, and improve parameter efficiency, leading to significant improvements in object recognition tasks.
IEEE TRANSACTIONS ON PATTERN ANALYSIS AND MACHINE INTELLIGENCE
(2022)
Article
Environmental Sciences
Jiangbo Xi, Okan K. Ersoy, Ming Cong, Chaoying Zhao, Wei Qu, Tianjun Wu
Summary: This paper proposes a wide and deep Fourier network for efficient feature learning in hyperspectral remote sensing image (HSI) classification. The method utilizes pruned features extracted in the frequency domain to extract hierarchical features layer-by-layer. Experimental results show that the proposed method achieves excellent performance in HSI classification, with the ability to be implemented in lightweight embedded computing platforms.
Article
Computer Science, Artificial Intelligence
Brendan Kolisnik, Isaac Hogan, Farhana Zulkernine
Summary: We propose a hierarchical image classification model, Condition-CNN, which improves prediction accuracy and reduces training time by using the Teacher Forcing training algorithm and conditional probabilities. The validation results show that Condition-CNN achieves higher prediction accuracy for Level 1, 2, and 3 classes compared to other baseline CNN models.
EXPERT SYSTEMS WITH APPLICATIONS
(2021)
Article
Computer Science, Artificial Intelligence
Yingying Zhu, Yinghao Wang, Haonan Chen, Zemian Guo, Qiang Huang
Summary: This paper proposes a method based on attention mechanism to address the problem of clutter and occlusion in feature extraction when using convolutional neural networks (CNN). Two attention modules, spatial attention module and channel attention module, are introduced to adjust the weight distribution of feature maps, making the extracted features more discriminative. Furthermore, a scale and mask scheme is presented to filter out redundant features and reduce the disadvantages of various scales. Experimental results demonstrate the effectiveness of the proposed method on four well-known image retrieval datasets.
INTERNATIONAL JOURNAL OF NEURAL SYSTEMS
(2023)
Article
Engineering, Electrical & Electronic
Guanglin Li, Bin Li, Shunquan Tan, Guoping Qiu
Summary: We used deep convolutional neural network architecture and co-occurrence matrix to learn deep co-occurrence features. These features represent the statistics of pixel co-occurrences, overcoming the black box nature of traditional deep representation learning and solving the computational difficulty of the matrix. We proposed a parametric co-occurrence matrix model and developed approaches to decompose the model into linear and nonlinear operations, making it easily implementable and capable of learning arbitrary shape DCOFs.
IEEE TRANSACTIONS ON CIRCUITS AND SYSTEMS FOR VIDEO TECHNOLOGY
(2023)
Article
Environmental Sciences
Zeyu Xu, Cheng Su, Shirou Wang, Xiaocan Zhang
Summary: In this study, a local-global spectral feature (LGSF) extraction and optimization method is proposed for hyperspectral image (HSI) classification. The method transforms the 1D spectral vector into a 2D spectral image and automatically extracts the LGSF by using the local spectral feature extraction module (LSFEM) and the global spectral feature extraction module (GSFEM). The LGSF is further optimized using a loss function inspired by contrastive learning. The proposed method demonstrates its effectiveness in utilizing spectral information and achieving accurate HSI classification.
Article
Computer Science, Artificial Intelligence
Feng Yang, Zheng Ma, Mei Xie
Summary: Principal component analysis (PCA) and kernel principal component analysis (KPCA) algorithms are used to construct a deep learning model called parallel KPCA-PCA network (PK-PCANet) in this article. The proposed model calculates filters for convolution layers using PCA and KPCA algorithms, and fuses the extracted features from PCANet and KPCANet using parallel feature fusion strategy. Compressed sensing algorithm is incorporated in the network to reduce the dimensionality of learned features. Extensive experiments on various visual recognition tasks validate the efficiency of the proposed PK-PCANet method.
COMPUTATIONAL INTELLIGENCE
(2022)
Review
Environmental Sciences
Leiyu Chen, Shaobo Li, Qiang Bai, Jing Yang, Sanlong Jiang, Yanming Miao
Summary: This article summarizes the application of deep learning in image classification, covering the development of CNNs from their predecessors to the latest network architectures, as well as a comprehensive comparison and analysis of various image classification methods.
Article
Automation & Control Systems
Yiming Lei, Haiping Zhu, Junping Zhang, Hongming Shan
Summary: This study proposes a method for medical image classification with ordinal labels, which improves the model's generalization ability by combining convolutional neural networks and differential forests in a meta-learning framework. The key components of the method are the tree-wise weighting network and the grouped feature selection module. Experimental results demonstrate the superior performance of this method over existing state-of-the-art methods on two medical image classification datasets with ordinal labels.
IEEE-CAA JOURNAL OF AUTOMATICA SINICA
(2022)
Article
Environmental Sciences
Yang-Lang Chang, Tan-Hsu Tan, Wei-Hong Lee, Lena Chang, Ying-Nong Chen, Kuo-Chin Fan, Mohammad Alkhaleefah
Summary: The performance of hyperspectral image classification is influenced by spatial and spectral information, but affected by factors like data redundancy and insufficient spatial resolution. While 3D-CNN-based methods perform well, they come with high computational complexity. To address these issues, this study proposes a consolidated C-CNN method that combines 3D-CNN and 2D-CNN. PCA is used to reduce spectral redundancy, and image augmentation techniques are employed to increase training samples and prevent overfitting. The proposed C-CNN model, named C-CNN-Aug, achieves optimal trade-off between accuracy and computational time compared to other methods.
Article
Computer Science, Artificial Intelligence
Sayed Kamaledin Ghiasi Shirazi, Reza Safabakhsh
COMPUTER VISION AND IMAGE UNDERSTANDING
(2009)
Article
Computer Science, Artificial Intelligence
Kamaledin Ghiasi-Shirazi
NEURAL PROCESSING LETTERS
(2019)
Article
Computer Science, Artificial Intelligence
Mohammad Reza Mohammadnia-Qaraei, Reza Monsefi, Kamaledin Ghiasi-Shirazi
PATTERN RECOGNITION LETTERS
(2018)
Article
Engineering, Biomedical
Noushin Eftekhari, Hamid-Reza Pourreza, Mojtaba Masoudi, Kamaledin Ghiasi-Shirazi, Ehsan Saeedi
BIOMEDICAL ENGINEERING ONLINE
(2019)
Article
Computer Science, Artificial Intelligence
Kamaledin Ghiasi-Shirazi
NEURAL PROCESSING LETTERS
(2019)
Article
Computer Science, Artificial Intelligence
Naeem Paeedeh, Kamaledin Ghiasi-Shirazi
Summary: The paper extends NLMS and APA algorithms to multi-layer neural networks, treating neural networks as a set of adaptive filters. A more robust algorithm is proposed that predicts and corrects adverse consequences during training, making it easier to tune and reducing reliance on other techniques.
Article
Computer Science, Artificial Intelligence
Saeid Abbaasi, Kamaledin Ghiasi-Shirazi, Ahad Harati
Summary: Capsule networks are deep neural networks that use a part-to-whole association and transformation matrices to represent classes. In this paper, a new multi-prototype capsule network architecture is proposed along with a soft competitive learning algorithm. The results show that this approach improves the classification accuracy compared to the original capsule networks and high-dimensional capsules.
NEURAL PROCESSING LETTERS
(2023)
Article
Oncology
M. E. Ravari, Sh. Nasseri, M. Mohammadi, M. Behmadi, S. K. Ghiasi-Shirazi, M. Momennezhad
Summary: The study demonstrates that a deep-learning model can accurately and precisely predict the three-dimensional dose distribution for patients with breast cancer. This approach can be extended to other cancer treatment planning systems.
Article
Computer Science, Artificial Intelligence
Ramin Zarei-Sabzevar, Kamaledin Ghiasi-Shirazi, Ahad Harati
Summary: This article introduces the application of winner-take-all network based on minimum Euclidean distance in prototype-based learning and proposes +/- ED-WTA method for model construction and training. Experiments show that this method can construct highly interpretable prototypes and be used to explain the functionality of deep neural networks, as well as detect outliers and adversarial examples.
IEEE TRANSACTIONS ON NEURAL NETWORKS AND LEARNING SYSTEMS
(2023)
Article
Computer Science, Hardware & Architecture
Kamaledin Ghiasi-Shirazi, Mahdi Mohseni, Majid Darvishan, Reza Yousefzadeh
COMPUTER STANDARDS & INTERFACES
(2017)
Proceedings Paper
Computer Science, Theory & Methods
Amir Ahooye Atashin, Kamaledin Ghiasi-Shirazi, Ahad Harati
2016 6TH INTERNATIONAL CONFERENCE ON COMPUTER AND KNOWLEDGE ENGINEERING (ICCKE)
(2016)
Article
Automation & Control Systems
Kamaledin Ghiasi-Shirazi, Reza Safabakhsh, Mostafa Shamsi
JOURNAL OF MACHINE LEARNING RESEARCH
(2010)
Article
Computer Science, Information Systems
Xia Liang, Jie Guo, Peide Liu
Summary: This paper investigates a novel consensus model based on social networks to manage manipulative and overconfident behaviors in large-scale group decision-making. By proposing a novel clustering model and improved methods, the consensus reaching is effectively facilitated. The feedback mechanism and management approach are employed to handle decision makers' behaviors. Simulation experiments and comparative analysis demonstrate the effectiveness of the model.
INFORMATION SCIENCES
(2024)
Article
Computer Science, Information Systems
Xiang Li, Haiwang Guo, Xinyang Deng, Wen Jiang
Summary: This paper proposes a method based on class gradient networks for generating high-quality adversarial samples. By introducing a high-level class gradient matrix and combining classification loss and perturbation loss, the method demonstrates superiority in the transferability of adversarial samples on targeted attacks.
INFORMATION SCIENCES
(2024)
Article
Computer Science, Information Systems
Lingyun Lu, Bang Wang, Zizhuo Zhang, Shenghao Liu
Summary: Many recommendation algorithms only rely on implicit feedbacks due to privacy concerns. However, the encoding of interaction types is often ignored. This paper proposes a relation-aware neural model that classifies implicit feedbacks by encoding edges, thereby enhancing recommendation performance.
INFORMATION SCIENCES
(2024)
Article
Computer Science, Information Systems
Jaehong Yu, Hyungrok Do
Summary: This study discusses unsupervised anomaly detection using one-class classification, which determines whether a new instance belongs to the target class by constructing a decision boundary. The proposed method uses a proximity-based density description and a regularized reconstruction algorithm to overcome the limitations of existing one-class classification methods. Experimental results demonstrate the superior performance of the proposed algorithm.
INFORMATION SCIENCES
(2024)
Article
Computer Science, Information Systems
Hui Tu, Shifei Ding, Xiao Xu, Haiwei Hou, Chao Li, Ling Ding
Summary: Border-Peeling algorithm is a density-based clustering algorithm, but its complexity and issues on unbalanced datasets restrict its application. This paper proposes a non-iterative border-peeling clustering algorithm, which improves the clustering performance by distinguishing and associating core points and border points.
INFORMATION SCIENCES
(2024)
Article
Computer Science, Information Systems
Long Tang, Pan Zhao, Zhigeng Pan, Xingxing Duan, Panos M. Pardalos
Summary: In this work, a two-stage denoising framework (TSDF) is proposed for zero-shot learning (ZSL) to address the issue of noisy labels. The framework includes a tailored loss function to remove suspected noisy-label instances and a ramp-style loss function to reduce the negative impact of remaining noisy labels. In addition, a dynamic screening strategy (DSS) is developed to efficiently handle the nonconvexity of the ramp-style loss.
INFORMATION SCIENCES
(2024)
Article
Computer Science, Information Systems
Raghunathan Krishankumar, Sundararajan Dhruva, Kattur S. Ravichandran, Samarjit Kar
Summary: Health 4.0 is gaining global attention for better healthcare through digital technologies. This study proposes a new decision-making framework for selecting viable blockchain service providers in the Internet of Medical Things (IoMT). The framework addresses the limitations in previous studies and demonstrates its applicability in the Indian healthcare sector. The results show the top ranking BSPs, the importance of various criteria, and the effectiveness of the developed model.
INFORMATION SCIENCES
(2024)
Article
Computer Science, Information Systems
Tao Tan, Hong Xie, Liang Feng
Summary: This paper proposes a heterogeneous update idea and designs HetUp Q-learning algorithm to enlarge the normalized gap by overestimating the Q-value corresponding to the optimal action and underestimating the Q-value corresponding to the other actions. To address the limitation, a softmax strategy is applied to estimate the optimal action, resulting in HetUpSoft Q-learning and HetUpSoft DQN. Extensive experimental results show significant improvements over SOTA baselines.
INFORMATION SCIENCES
(2024)
Article
Computer Science, Information Systems
Chao Yang, Xianzhi Wang, Lina Yao, Guodong Long, Guandong Xu
Summary: This paper proposes a dynamic transformer-based architecture called Dyformer for multivariate time series classification. Dyformer captures multi-scale features through hierarchical pooling and adaptive learning strategies, and improves model performance by introducing feature-map-wise attention mechanisms and a joint loss function.
INFORMATION SCIENCES
(2024)
Article
Computer Science, Information Systems
Xiguang Li, Baolu Feng, Yunhe Sun, Ammar Hawbani, Saeed Hammod Alsamhi, Liang Zhao
Summary: This paper proposes an enhanced scatter search strategy, using opposition-based learning, to solve the problem of automated test case generation based on path coverage (ATCG-PC). The proposed ESSENT algorithm selects the path with the lowest path entropy among the uncovered paths as the target path and generates new test cases to cover the target path by modifying the dimensions of existing test cases. Experimental results show that the ESSENT algorithm outperforms other state-of-the-art algorithms, achieving maximum path coverage with fewer test cases.
INFORMATION SCIENCES
(2024)
Article
Computer Science, Information Systems
Shirin Dabbaghi Varnosfaderani, Piotr Kasprzak, Aytaj Badirova, Ralph Krimmel, Christof Pohl, Ramin Yahyapour
Summary: Linking digital accounts belonging to the same user is crucial for security, user satisfaction, and next-generation service development. However, research on account linkage is mainly focused on social networks, and there is a lack of studies in other domains. To address this, we propose SmartSSO, a framework that automates the account linkage process by analyzing user routines and behavior during login processes. Our experiments on a large dataset show that SmartSSO achieves over 98% accuracy in hit-precision.
INFORMATION SCIENCES
(2024)
Article
Computer Science, Information Systems
Renchao Wu, Jianjun He, Xin Li, Zuguo Chen
Summary: This paper proposes a memetic algorithm with fuzzy-based population control (MA-FPC) to solve the joint order batching and picker routing problem (JOBPRP). The algorithm incorporates batch exchange crossover and a two-level local improvement procedure. Experimental results show that MA-FPC outperforms existing algorithms in terms of solution quality.
INFORMATION SCIENCES
(2024)
Article
Computer Science, Information Systems
Guoxiang Zhong, Fagui Liu, Jun Jiang, Bin Wang, C. L. Philip Chen
Summary: In this study, we propose the AMFormer framework to address the problem of mixed normal and anomaly samples in deep unsupervised time-series anomaly detection. By refining the one-class representation and introducing the masked operation mechanism and cost sensitive learning theory, our approach significantly improves anomaly detection performance.
INFORMATION SCIENCES
(2024)
Article
Computer Science, Information Systems
Jin Zhou, Kang Zhou, Gexiang Zhang, Ferrante Neri, Wangyang Shen, Weiping Jin
Summary: In this paper, the authors focus on the issue of multi-objective optimisation problems with redundant variables and indefinite objective functions (MOPRVIF) in practical problem-solving. They propose a dual data-driven method for solving this problem, which consists of eliminating redundant variables, constructing objective functions, selecting evolution operators, and using a multi-objective evolutionary algorithm. The experiments conducted on two different problem domains demonstrate the effectiveness, practicality, and scalability of the proposed method.
INFORMATION SCIENCES
(2024)
Article
Computer Science, Information Systems
Georgios Charizanos, Haydar Demirhan, Duygu Icen
Summary: This article proposes a new fuzzy logistic regression framework that addresses the problems of separation and imbalance while maintaining the interpretability of classical logistic regression. By fuzzifying binary variables and classifying subjects based on a fuzzy threshold, the framework demonstrates superior performance on imbalanced datasets.
INFORMATION SCIENCES
(2024)