Article
Automation & Control Systems
Kshema Shaju, Sherin Babu, Binu Thomas
Summary: This study analyzes the effectiveness of the application of grey theory in feature selection for daily dew point temperature and daily pan-evaporation estimation models. Comparisons and analyses are made between the feature subset identified by grey theory and subsets selected based on different Pearson correlation coefficient slabs. The results show that the models using grey theory-based feature selection demonstrated average or above-average performances.
ENGINEERING APPLICATIONS OF ARTIFICIAL INTELLIGENCE
(2023)
Article
Automation & Control Systems
Federico Zocco, Marco Maggipinto, Gian Antonio Susto, Sean McLoone
Summary: This paper proposes a 'lazy' implementation of Forward Selection Component Analysis (L-FSCA) which is faster than FSCA while having comparable performance. Experimental results show that L-FSCA reduces computation time by 22% to 94% while yielding almost identical performance to FSCA.
ENGINEERING APPLICATIONS OF ARTIFICIAL INTELLIGENCE
(2023)
Article
Engineering, Electrical & Electronic
Nima Nouri, Mohammad Javad Azizipour
Summary: This paper proposes a novel sparse recovery algorithm that utilizes the row, common, and individual sparse elements of a channel in a new way. The algorithm is designed to control the propagation environment of wireless systems using reconfigurable intelligent surface (RIS) technology in millimeter wave (mmWave) communication systems. Experimental results show that the proposed algorithm can substantially reduce pilot overhead while achieving close to the performance bound in terms of mean square error (MSE).
PHYSICAL COMMUNICATION
(2023)
Article
Computer Science, Information Systems
Yuji Saito, Taku Nonomura, Keigo Yamada, Kumi Nakai, Takayuki Nagata, Keisuke Asai, Yasuo Sasaki, Daisuke Tsubakino
Summary: This paper focuses on the sparse sensor placement problem for least-squares estimation and extends the previous novel approach of the sparse sensor selection algorithm. The study shows that the method used when the number of sensors is less than the number of state variables is mathematically the same as the previously proposed QR method, while a new algorithm is developed for cases where the number of sensors is greater than the number of state variables. Furthermore, the effectiveness of the proposed algorithm is demonstrated through comparisons with other algorithms using real datasets.
Article
Computer Science, Artificial Intelligence
Zahra Atashgahi, Ghada Sokar, Tim van der Lee, Elena Mocanu, Decebal Constantin Mocanu, Raymond Veldhuis, Mykola Pechenizkiy
Summary: A novel unsupervised feature selection method QuickSelection is proposed in this paper, which introduces the concept of neuron strength in sparse neural networks and combines it with sparsely connected denoising autoencoders to derive the importance of all input features. The method achieves the best trade-off of classification and clustering accuracy, running time, and maximum memory usage on benchmark datasets, with considerable speed increase, memory reduction, and the least amount of energy consumption compared to other state-of-the-art autoencoder-based feature selection methods.
Article
Computer Science, Information Systems
Heping Song, Xiaolong Ren, Yuping Lai, Hongying Meng
Summary: The paper introduces a new sparse analysis recovery algorithm ICDE for signals following the cosparse analysis model, which alternates between cosupport detection and analysis pursuit to estimate the signal.
Article
Engineering, Multidisciplinary
Prashant Shekhar, Abani Patra
Summary: This article proposes a feature-driven multiscale reproducing kernel Hilbert space (RKHS) with a weighted multiscale structure in the associated kernel. A practical forward-backward algorithm is provided for generating approximations in this space, which greedily constructs a set of basis functions with a multiscale structure for sparse efficient representation of data and efficient predictions. The article includes a detailed analysis of the algorithm, including recommendations for selecting algorithmic hyperparameters and estimating probabilistic rates of convergence at individual scales. The performance of the approach is analyzed on various simulations and real datasets, demonstrating its efficiency in terms of model quality and data reduction.
COMPUTER METHODS IN APPLIED MECHANICS AND ENGINEERING
(2022)
Article
Engineering, Multidisciplinary
Prashant Shekhar, Abani Patra
Summary: The study introduces a feature-driven RKHS with a practical forward-backward algorithm for generating approximations that enable sparse efficient representation of data and efficient predictions. The analysis includes recommendations for selecting algorithmic hyperparameters and estimating probabilistic rates of convergence at different scales.
COMPUTER METHODS IN APPLIED MECHANICS AND ENGINEERING
(2022)
Article
Forestry
Mi Luo, Yifu Wang, Yunhong Xie, Lai Zhou, Jingjing Qiao, Siyu Qiu, Yujun Sun
Summary: Increasing the number of explanatory variables in quantitative remote sensing of forest aboveground biomass may lead to information redundancy and dimensional disaster. Feature selection, particularly using the CatBoost algorithm, improves the accuracy of AGB estimates. Different combinations of feature selection methods and machine learning algorithms can significantly impact the performance of AGB estimation models.
Article
Computer Science, Artificial Intelligence
Jian-Sheng Wu, Meng-Xiao Song, Weidong Min, Jian-Huang Lai, Wei-Shi Zheng
Summary: The study introduces a novel unsupervised feature selection framework JAMEL, which aims to preserve the manifold structure among data by iteratively and adaptively learning lower-dimensional embeddings. The results show the effectiveness and efficiency of the approach in various tasks such as k-means, spectral clustering and nearest neighbor classification.
PATTERN RECOGNITION
(2021)
Article
Automation & Control Systems
Amir Moslemi
Summary: Curse of dimensionality is a major challenge in data mining, pattern recognition, computer vision, and machine learning. Feature selection and feature extraction are two main approaches to address this challenge. This survey provides a comprehensive overview of state-of-the-art feature selection techniques, categorized into five domains. It can be helpful for researchers to gain a deep understanding of feature selection techniques and choose a suitable domain for future study.
ENGINEERING APPLICATIONS OF ARTIFICIAL INTELLIGENCE
(2023)
Article
Computer Science, Artificial Intelligence
Yonghao Li, Liang Hu, Wanfu Gao
Summary: In recent years, joint feature selection and multi-label learning have been widely studied. However, existing multi-label feature selection methods face three challenges: neglecting feature redundancy, using low-quality graphs to capture local label correlations, and considering only either local or global label correlations. To address these challenges, we propose a method that preserves global and dynamic local label correlations by preserving the graph structure. We also introduce regularization terms to select low redundant features. Experimental results demonstrate the superiority of our method in classification tasks.
PATTERN RECOGNITION
(2023)
Article
Computer Science, Artificial Intelligence
Rong Wang, Jintang Bian, Feiping Nie, Xuelong Li
Summary: Feature selection is crucial for dealing with high-dimensional data in machine learning and data mining tasks. However, most existing methods overlook the fuzziness in the data, resulting in sub-optimal results. To address this, we propose a novel unsupervised feature selection method that simultaneously conducts fuzziness learning and sparse learning, selecting discriminative features.
IEEE TRANSACTIONS ON KNOWLEDGE AND DATA ENGINEERING
(2022)
Article
Engineering, Electrical & Electronic
Ben Adcock, Simone Brugiapaglia, Matthew King-Roskamp
Summary: The study focuses on the application of the sparse in levels model in compressive imaging, and proposes new stable and robust uniform recovery guarantees, expanding the current research scope available under standard sparsity.
IEEE TRANSACTIONS ON SIGNAL PROCESSING
(2021)
Article
Automation & Control Systems
Alex Devonport, Forest Yang, Laurent El Ghaoui, Murat Arcak
Summary: This article presents algorithms for estimating the forward reachable set of a dynamical system using a finite collection of independent and identically distributed samples. The accuracy bounds of these algorithms are proven under the PAC framework. The algorithms can also be applied to estimating the support of a random variable, which is useful for detecting novelties and outliers in datasets.
IEEE TRANSACTIONS ON AUTOMATIC CONTROL
(2023)
