Article
Computer Science, Artificial Intelligence
Jiapeng Wang, Jiaxiang Luo
Summary: This paper proposes a new method for optimizing the kernel and penalty parameters of SVM classifiers. The method introduces a new distance measure in the feature space and presents a fast parameter optimization approach that significantly reduces training time while maintaining competitive model accuracy.
APPLIED SOFT COMPUTING
(2022)
Article
Quantum Science & Technology
N. Innan, M. A. Z. Khan, B. Panda, M. Bennai
Summary: We propose a new model, quantum variational kernel SVM (QVK-SVM), in quantum machine learning (QML), which combines the quantum kernel and quantum variational algorithms to improve accuracy. extensive experiments on the Iris dataset demonstrate that QVK-SVM outperforms existing models regarding accuracy, loss, and confusion matrix indicators.
QUANTUM INFORMATION PROCESSING
(2023)
Article
Computer Science, Information Systems
B. Sakthi Karthi Durai, J. Benadict Raja
Summary: The early detection of retinal abnormalities like diabetic retinopathy (DR) can be achieved using computerized analysis of retinal fundus images. This study presents an automated process that employs an optimized SVM classifier and a new feature extraction method for more accurate and efficient detection of DR. The proposed technique is validated using a standard dataset and achieves high sensitivity, specificity, and accuracy.
MULTIMEDIA TOOLS AND APPLICATIONS
(2023)
Article
Quantum Science & Technology
Fei Wang, Kunlun Xie, Lin Han, Menghui Han, Zeshi Wang
Summary: This paper introduces support vector machine and its problems, proposes an improved quantum genetic algorithm, and applies it to SVM parameter optimization. Experimental results show that the improved algorithm has better performance compared to other algorithms.
QUANTUM INFORMATION PROCESSING
(2023)
Article
Computer Science, Artificial Intelligence
Lang Yu, Shengjie Li, Siyi Liu
Summary: A novel three-term conjugate-like sequential minimal optimization algorithm (TCSMO) is proposed for SVM classification and regression tasks, based on the conjugate sequential minimal optimization algorithm (CSMO). The TCSMO algorithm slightly increases the arithmetic operations but significantly reduces the number of iterations required for convergence, resulting in shorter training time for SVM. Numerical experiments demonstrate the improved performance of the TCSMO algorithm in both classification and regression tasks.
PATTERN RECOGNITION
(2023)
Article
Computer Science, Artificial Intelligence
Liuyuan Chen, Kanglei Zhou, Junchang Jing, Haiju Fan, Juntao Li
Summary: This work proposes a fast regularization parameter tuning algorithm for the twin multi-class support vector machine. By adopting a novel sample data set partition strategy and utilizing linear equations and block matrix theory, the regularization parameters are continuously updated, and the relationship between the Lagrangian multipliers and the regularization parameters is proven. Finally, different events are defined to seek for the starting event for the next iteration, and the effectiveness of the proposed method is validated through experiments.
EXPERT SYSTEMS WITH APPLICATIONS
(2022)
Article
Computer Science, Artificial Intelligence
Wu Huiyong, Jin Shuchun, Jin Zhu
Summary: This paper analyzes the working state of the aircraft's air circulation system at high altitude and proposes a simulation analysis method based on simulated annealing-grasshopper optimization algorithm and support vector machine. By conducting system simulation and fault injection analysis, the results show that this method can effectively simulate temperature changes and distinguish fault types of the aircraft, and the convergence speed of the system is accelerated to avoid local optimal problems.
Article
Computer Science, Artificial Intelligence
Chen Ding, Tian-Yi Bao, He-Liang Huang
Summary: The study proposes a quantum-inspired classical algorithm for LS-SVM, utilizing an improved sampling technique for classification. The theoretical analysis indicates that the algorithm can achieve classification with logarithmic runtime for low-rank, low-condition number, and high-dimensional data matrices.
IEEE TRANSACTIONS ON NEURAL NETWORKS AND LEARNING SYSTEMS
(2022)
Article
Computer Science, Information Systems
Guoquan Li, Linxi Yang, Zhiyou Wu, Changzhi Wu
Summary: Proximal support vector machine (PSVM) is a variant of support vector machine (SVM) which aims to generate a pair of non-parallel hyperplanes for classification. Introducing l(0)-norm regularization in PSVM enables simultaneous selection of important features and removal of redundant features for classification. The proposed method utilizes a continuous nonconvex function and difference of convex functions algorithms (DCA) to solve the optimization problem efficiently.
INFORMATION SCIENCES
(2021)
Article
Computer Science, Artificial Intelligence
Wojciech Dudzik, Jakub Nalepa, Michal Kawulok
Summary: This paper addresses the optimization problem of SVMs for binary classification of difficult datasets, introducing an evolutionary technique and a co-evolutionary scheme. Experimental results show that the proposed algorithm outperforms popular supervised learners and other techniques for optimizing SVMs.
KNOWLEDGE-BASED SYSTEMS
(2021)
Article
Automation & Control Systems
Jian Zhou, Yingui Qiu, Shuangli Zhu, Danial Jahed Armaghani, Chuanqi Li, Hoang Nguyen, Saffet Yagiz
Summary: This research aims to optimize the hyper-parameters of the support vector machine technique through the use of three optimization algorithms, namely gray wolf optimization, whale optimization algorithm, and moth flame optimization, for predicting the advance rate of a tunnel boring machine in hard rock conditions. The results indicate that the moth flame optimization algorithm can better capture the hyperparameters of the SVM model in predicting TBM AR.
ENGINEERING APPLICATIONS OF ARTIFICIAL INTELLIGENCE
(2021)
Article
Computer Science, Artificial Intelligence
Shifei Ding, Zichen Zhang, Yuting Sun, Songhui Shi
Summary: The parameters of MBSVM are currently determined mainly by experience or artificially specified by the grid method, which can lead to insufficient parameter selection. To address this issue, a dynamic quantum particle swarm optimization algorithm (DQPSO) is proposed to optimize the MBSVM parameters. Experimental results show that the DQPSO algorithm outperforms the classical QPSO algorithm in terms of optimization performance and convergence speed, and the proposed DQPSO-MBSVM algorithm effectively improves the classification performance of MBSVM.
Article
Computer Science, Artificial Intelligence
Tong Gao, Hao Chen
Summary: In this study, a multicycle disassembly-based decomposition algorithm (MCD-DA) is proposed to efficiently solve the training problem of multiclass support vector machine (SVM). MCD-DA constructs a graph model to re-express the constraints in multiclass SVM, partitions the complex feasible region into simple sub-feasible regions, and designs multiple cycle-based disassembly strategies to update the working variables analytically. Experimental results demonstrate that MCD-DA outperforms typical optimization algorithms for more sample cases.
PATTERN RECOGNITION
(2023)
Article
Computer Science, Artificial Intelligence
Rosita Guido, Maria Carmela Groccia, Domenico Conforti
Summary: Hyperparameter tuning is essential for improving model performance in machine learning. This research focuses on classifying imbalanced data using cost-sensitive support vector machines and proposes a multi-objective approach to optimize the model's hyperparameters. The algorithm is presented in a basic version and an improved version utilizing genetic algorithms and decision trees. Experimental results demonstrate the importance of using appropriate evaluation measures for assessing the classification performance of imbalanced data classification models.
Article
Computer Science, Artificial Intelligence
Bagesh Kumar, Ayush Sinha, Sourin Chakrabarti, O. P. Vyas
Summary: In this paper, a fast training method for OCSSVM is proposed, which enhances its scalability without compromising precision significantly. Experimental results show that the proposed method achieves the best tradeoff between training time and accuracy, providing similar accuracies to regular OCSSVM and better scalability compared to existing state-of-the-art one-class classifiers.
KNOWLEDGE-BASED SYSTEMS
(2021)