Article
Computer Science, Artificial Intelligence
Bin Cao, Yuqi Liu, Chenyu Hou, Jing Fan, Baihua Zheng, Jianwei Yin
Summary: This paper proposes ATEC, a solution that can efficiently find a preferable hyperplane by automatically tuning the error cost for between-class samples. ATEC distinguishes itself from all existing parameter tuning strategies by evaluating the effectiveness of error costs and changing them in the right direction if they are not effective.
IEEE TRANSACTIONS ON KNOWLEDGE AND DATA ENGINEERING
(2021)
Article
Computer Science, Artificial Intelligence
M. A. Ganaie, M. Tanveer, Chin-Teng Lin
Summary: In this article, a novel large-scale fuzzy least squares TSVM method is proposed to handle large-scale data. The method avoids the shortcomings of traditional TSVMs in large-scale data processing by using structural risk minimization principle and positive-definite matrices. Experimental results demonstrate the superior performance of the proposed method in large-scale classification problems and its effectiveness in addressing class imbalance.
IEEE TRANSACTIONS ON FUZZY SYSTEMS
(2022)
Article
Computer Science, Artificial Intelligence
M. A. Ganaie, M. Tanveer, Alzheimer's Disease Neuroimaging Initiative
Summary: This paper introduces a novel fuzzy least squares projection twin support vector machines for class imbalance learning, which outperforms baseline models in experiments.
APPLIED SOFT COMPUTING
(2021)
Article
Computer Science, Artificial Intelligence
Imran Razzak, Mohamed Reda Bouadjenek, Raghib Abu Saris, Weiping Ding
Summary: Traditional support vector machines (SVMs) are sensitive to outliers and corrupted data, leading to a deterioration in classification performance. This article proposes an efficient Support Matrix Machine that performs matrix recovery and feature selection simultaneously. It can handle high-dimensional data with corrupted columns and recover an intrinsic matrix of higher rank under incoherence and ambiguity conditions. The objective function combines matrix recovery, low rank, and joint sparsity, and the method leverages structural information and intrinsic data structure. Experimental results show significant improvements in BCI, face recognition, and person identification datasets, especially in the presence of outliers.
IEEE TRANSACTIONS ON NEURAL NETWORKS AND LEARNING SYSTEMS
(2023)
Review
Computer Science, Information Systems
Ahmad S. Tarawneh, Ahmad B. Hassanat, Ghada Awad Altarawneh, Abdullah Almuhaimeed
Summary: Oversampling has been used to address the challenge of learning from imbalanced datasets, but it also has its downsides. The synthesized samples may not truly represent the minority class, resulting in incorrect predictions in real-world applications. This paper analyzes various oversampling methods and proposes a new evaluation system based on comparing hidden majority examples with those generated by oversampling. Experimental results show that all studied oversampling methods produce minority samples that are most likely to be majority. Given the data and methods at hand, oversampling in its current forms and methodologies is deemed unreliable and should be avoided in real-world applications.
Article
Computer Science, Artificial Intelligence
M. A. Ganaie, M. Tanveer, Alzheimer's Disease Neuroimaging Initiative
Summary: In real world problems, the imbalance of data samples presents a challenge for classification models. This paper proposes a K-nearest neighbor based weighted reduced universum twin support vector machine model to address class imbalance issues by incorporating local neighborhood information and utilizing universum data, resulting in improved generalization performance.
KNOWLEDGE-BASED SYSTEMS
(2022)
Article
Computer Science, Artificial Intelligence
Pei-Yi Hao, Jung-Hsien Chiang, Yu-De Chen
Summary: This paper proposes a novel possibilistic classification algorithm using support vector machines (SVMs) to effectively handle uncertain information and improve classification performance. The algorithm aims at finding a maximal-margin fuzzy hyperplane based on possibility theory and solves a fuzzy mathematical optimization problem. The proposed algorithm retains the advantages of fuzzy set theory and SVM theory, and it is more robust for handling outliers. Experimental results demonstrate the satisfactory generalization accuracy and ability to describe inherent vagueness in the given dataset.
Article
Computer Science, Information Systems
M. Tanveer, S. Sharma, K. Muhammad
Summary: The proposed LS-LSTSVM addresses the shortcomings of TWSVM and LSTSVM by introducing a different Lagrangian function to eliminate the need for calculating inverse matrices, using the kernel trick directly for non-linear cases, and minimizing structural risk. These improvements aim to enhance classification accuracy on datasets, especially for large-scale problems.
ACM TRANSACTIONS ON INTERNET TECHNOLOGY
(2021)
Article
Automation & Control Systems
Guanjin Wang, Kup-Sze Choi, Jeremy Yuen-Chun Teoh, Jie Lu
Summary: This article introduces a new approach called DCOT-LS-SVMs, which is based on least-squares support vector machines and utilizes deep cross-output knowledge transfer. The approach improves the generalizability of LS-SVMs and simplifies the parameter tuning process. Experimental results on UCI datasets and a prostate cancer diagnosis case study demonstrate the effectiveness of the proposed approach.
IEEE TRANSACTIONS ON CYBERNETICS
(2022)
Article
Automation & Control Systems
Barenya Bikash Hazarika, Deepak Gupta
Summary: This study proposes a novel affinity-based fuzzy kernel ridge regression (AFKRR) model to tackle the class imbalance learning (CIL) problem. Experimental results demonstrate the good performance and efficacy of the proposed AFKRR model.
ENGINEERING APPLICATIONS OF ARTIFICIAL INTELLIGENCE
(2023)
Article
Soil Science
Hossein Bayat, Golnaz Ebrahimzadeh, Binayak P. Mohanty
Summary: This research successfully established models for estimating soil thermal conductivity by combining topographical attributes and soil physical properties, and validated and improved them through various methods.
SOIL & TILLAGE RESEARCH
(2021)
Article
Computer Science, Artificial Intelligence
Chia-Feng Juang, Wei-En Ni
Summary: This article proposes a human posture classification system using three-dimensional fuzzy body voxel features and hierarchical fuzzy classifiers. The system successfully classifies standing, bending, sitting, and lying postures. Experimental results show the superiority of the proposed classification method.
IEEE TRANSACTIONS ON FUZZY SYSTEMS
(2022)
Article
Computer Science, Information Systems
Salim Rezvani, Xizhao Wang
Summary: The study introduces a class imbalance learning method using Fuzzy ART and IFTSVM, which effectively addresses classification issues related to class imbalance, noise, outliers, and large-scale datasets.
INFORMATION SCIENCES
(2021)
Article
Thermodynamics
Tiago de Oliveira Nogueira, Gilderlanio Barbosa Alves Palacio, Fabricio Damasceno Braga, Pedro Paulo Nunes Maia, Elineudo Pinho de Moura, Carla Freitas de Andrade, Paulo Alexandre Costa Rocha
Summary: This study combines DFA with SVM and RBFK methods for imbalance level classification in wind turbines, with RBFK showing excellent performance in different rotation speeds.
Article
Geochemistry & Geophysics
Giannis Lantzanakis, Zina Mitraka, Nektarios Chrysoulakis
Summary: Accurate land cover mapping of the Earth's surface using Earth observation data, especially in urban areas, is a challenging task due to the large spectral variability of man-made structures and the mixed pixel phenomenon. Support vector machines (SVMs) are commonly used for classification, but there is no rule of thumb for choosing optimal parameters in classifying satellite imagery, requiring a time-consuming trial-and-error process. Proposed advancements in the C-SVC algorithm aim to improve its performance and reduce manual parameterization.
IEEE TRANSACTIONS ON GEOSCIENCE AND REMOTE SENSING
(2021)
Article
Computer Science, Artificial Intelligence
Arka Ghosh, Sankha Subhra Mullick, Shounak Datta, Swagatam Das, Asit Kr Das, Rammohan Mallipeddi
Summary: This study introduces the DEceit algorithm, which constructs effective universal pixel-restricted perturbations using only black-box feedback from the target network. Through empirical investigations, it is found that perturbing around 10% of the pixels in an image achieves a highly transferable Fooling Rate while maintaining visual quality. Furthermore, DEceit shows success in image-dependent attacks and outperforms several state-of-the-art methods.
PATTERN RECOGNITION
(2022)
Article
Computer Science, Artificial Intelligence
Souhardya Sengupta, Swagatam Das
Summary: This paper proposes a simple data clustering technique that uses a graph created on nearest neighbors to identify clusters. The algorithm incorporates border detection and outlier detection techniques to construct the graph. The authors also introduce a novel outlier detection technique suitable for their implementation. Comparisons with state-of-the-art clustering techniques and experiments on various aspects of datasets are conducted.
PATTERN RECOGNITION LETTERS
(2022)
Article
Energy & Fuels
Oladayo S. Ajani, Abhishek Kumar, Rammohan Mallipeddi, Swagatam Das, Ponnuthurai Nagaratnam Suganthan
Summary: Energy disaggregation is an effective method to promote energy efficiency, but it faces challenges such as device similarity and measurement errors. In order to develop optimization algorithms for energy disaggregation, standard datasets and evaluation metrics are needed. This paper proposes a dataset with multiple instances and summarizes the performance indicators and baseline results of different optimization algorithms.
Article
Computer Science, Artificial Intelligence
Saptarshi Chakraborty, Swagatam Das
Summary: In this paper, a simple and efficient sparse clustering algorithm called LW-k-means is proposed for high-dimensional data. The algorithm incorporates feature weighting to enable feature selection and has a time complexity similar to traditional algorithms. The strong consistency of the LW-k-means procedure is also established. Experimental results on synthetic and real-life datasets demonstrate that LW-k-means performs competitively in terms of clustering accuracy and computational time compared to existing methods for center-based high-dimensional clustering.
IEEE TRANSACTIONS ON PATTERN ANALYSIS AND MACHINE INTELLIGENCE
(2022)
Article
Computer Science, Information Systems
Ivan Zelinka, Quoc Bao Diep, Vaclav Snasel, Swagatam Das, Giacomo Innocenti, Alberto Tesi, Fabio Schoen, Nikolay V. Kuznetsov
Summary: This paper discusses the alternative applications of pseudo-random processes and deterministic chaos in evolutionary algorithms, and proposes the use of a specific class of deterministic processes based on deterministic chaos to improve the performance of evolutionary algorithms, through comparing the performance of evolutionary algorithms driven by chaotic dynamics and pseudo-random number generators. New research questions are also proposed based on the findings.
INFORMATION SCIENCES
(2022)
Article
Computer Science, Information Systems
Avisek Gupta, Swagatam Das
Summary: This paper proposes an improved approach for model selection in sparse clustering by using the BIC expressions derived from center-based clustering methods. The derived BIC expressions significantly reduce computation costs and enable the comparison and selection of suitable sparse clusterings.
INFORMATION SCIENCES
(2022)
Article
Statistics & Probability
Anish Chakrabarty, Swagatam Das
Summary: By treating the underlying problem as a risk minimization task, this study provides uniform concentration bounds on the kernel k-means clustering objective based on Rademacher complexity, leading to state-of-the-art convergence rates on excess risk and strong consistency of cluster centers.
STATISTICS & PROBABILITY LETTERS
(2022)
Article
Automation & Control Systems
Pourya Shamsolmoali, Masoumeh Zareapoor, Swagatam Das, Salvador Garcia, Eric Granger, Jie Yang
Summary: Image-to-image translation is crucial in generative adversarial networks. Convolutional neural networks have limitations in capturing spatial relationships, making them unsuitable for image translation tasks. Capsule networks are proposed as a remedy, capturing hierarchical spatial relationships. In this paper, a new framework for capsule networks is presented, which can be applied to generator-discriminator architectures without computational overhead. A Gromov-Wasserstein distance is used as a loss function to guide the learned distribution. The proposed method, called generative equivariant network, is evaluated on I2I translation and image generation tasks and shows a principled connection between generative and capsule models.
IEEE TRANSACTIONS ON CYBERNETICS
(2023)
Review
Computer Science, Artificial Intelligence
Kanchan Rajwar, Kusum Deep, Swagatam Das
Summary: As industrialization progresses, solving optimization problems becomes more challenging. More than 500 new metaheuristic algorithms (MAs) have been developed, with over 350 of them emerging in the last decade. This study tracks approximately 540 MAs and provides statistical information. The proliferation of MAs has led to the issue of significant similarities between algorithms with different names. The study categorizes MAs based on the number of control parameters, which is a new taxonomy. Real-world applications of MAs are demonstrated and limitations and open challenges are identified.
ARTIFICIAL INTELLIGENCE REVIEW
(2023)
Article
Automation & Control Systems
Saptarshi Chakraborty, Debolina Paul, Swagatam Das
Summary: This article introduces a center-based clustering method that incorporates an entropy incentive term to learn feature importance efficiently. A scalable block-coordinate descent algorithm with closed-form updates is used to minimize the objective function. The merits of this method are showcased through detailed experimental analysis.
IEEE TRANSACTIONS ON CYBERNETICS
(2023)
Article
Computer Science, Artificial Intelligence
Kushal Bose, Swagatam Das
Summary: Graph Neural Networks (GNNs) are powerful for learning on graph-structured data, but they often suffer from over-smoothing, where node features become indistinguishable. This paper identifies the recursive and higher-to-lower order aggregation as the primary causes of over-smoothing and proposes a novel non-recursive aggregation strategy using randomized path exploration. Extensive comparative studies on benchmark datasets demonstrate the efficacy of the proposed method in semi-supervised and fully-supervised learning tasks.
IEEE TRANSACTIONS ON EMERGING TOPICS IN COMPUTATIONAL INTELLIGENCE
(2023)
Article
Computer Science, Artificial Intelligence
Debolina Paul, Saptarshi Chakraborty, Swagatam Das
Summary: Principal component analysis (PCA) is widely used for data visualization, denoising, and dimensionality reduction, but it is sensitive to outliers and may fail to detect low-dimensional structures. This article proposes a PCA method called MoMPCA based on the Median of Means (MoM) principle, which is computationally efficient and achieves optimal convergence rates. The method is robust to outliers and does not make assumptions about them. The efficacy of the proposed method is demonstrated through simulations and real data applications.
IEEE TRANSACTIONS ON NEURAL NETWORKS AND LEARNING SYSTEMS
(2023)
Review
Computer Science, Information Systems
Arkaprabha Basu, Sandip Paul, Sreeya Ghosh, Swagatam Das, Bhabatosh Chanda, Chakravarthy Bhagvati, Vaclav Snasel
Summary: Digitized methodologies have revolutionized various fields, including the restoration of buildings with historical significance. This interdisciplinary field attracts computer scientists who use computerized tools to reconstruct the values of these structures. The wear of time has endangered significant historical values, but this survey explores the use of 3D reconstruction, image inpainting, IoT-based methods, genetic algorithms, and image processing to restore cultural heritage. Machine Learning, Deep Learning, and Computer Vision-based methods are discussed, offering insights into faster, cheaper, and more beneficial techniques for image reconstruction in the near future.
Article
Computer Science, Artificial Intelligence
Debolina Paul, Saptarshi Chakraborty, Swagatam Das, Jason Xu
Summary: Kernel k-means clustering is a powerful tool for unsupervised learning of non-linearly separable data. In this paper, we propose a novel algorithm called Kernel Power k-Means, which leverages a general family of means to combat sub-optimal local solutions in the kernel and multi-kernel settings. Our algorithm uses majorization-minimization to solve the non-convex problem and implicitly performs annealing in kernel feature space. We rigorously analyze its convergence properties and demonstrate its efficacy through various experiments.
IEEE TRANSACTIONS ON PATTERN ANALYSIS AND MACHINE INTELLIGENCE
(2023)
Article
Computer Science, Information Systems
Arjun Ghosh, Nanda Dulal Jana, Swagatam Das, Rammohan Mallipeddi
Summary: This study proposes a two-phase evolutionary framework, TPEvo-CNN, for automatically designing CNN models for medical image classification. The framework utilizes differential evolution to determine the number of layers of the CNN architecture and genetic algorithm to fine-tune the hyperparameters. Experimental results demonstrate the superiority of the proposed framework in medical image classification tasks compared to existing CNN models.