Article
Marine & Freshwater Biology
Arina L. Maltseva, Marina A. Varfolomeeva, Anna V. Kursheva, Inna P. Morgunova, Roman V. Ayanka, Elizaveta R. Gafarova, Polina A. Pavlova, Egor A. Repkin, Arseniy A. Lobov, Elena A. Golikova, Natalia A. Mikhailova, Paul E. Renaud, Andrei I. Granovitch
Summary: Coastal marine ecosystems in the Arctic are vulnerable to human activities such as oil and gas exploration and urban development. This study assessed the effects of anthropogenic and natural factors on a particular gastropod species, and found that urban pollution had a significant impact on reproduction and shell shape.
ESTUARINE COASTAL AND SHELF SCIENCE
(2022)
Article
Engineering, Electrical & Electronic
Zhi-Yong Wang, Hing Cheung So, Abdelhak M. Zoubir
Summary: The correntropy criterion or Welsch function, widely used to resist outliers, has been recently applied to robust matrix recovery. However, it affects all observations, including uncontaminated data. Additionally, its implicit regularizer cannot achieve sparsity, which is desirable in many practical scenarios. To address these issues, a novel M-estimator called hybrid ordinary-Welsch (HOW) function is developed, which only affects the outlier-contaminated data and generates a regularizer that guarantees sparsity. Experimental results demonstrate that the proposed approach outperforms state-of-the-art methods in terms of recovery accuracy and runtime.
IEEE TRANSACTIONS ON SIGNAL PROCESSING
(2023)
Article
Computer Science, Information Systems
Yan Zhang, Haoqing Xu
Summary: The novel approach of demixed sparse principal component analysis (dSPCA) proposed in this research greatly improves the interpretability of multivariate data and addresses some issues existing in traditional SPCA. By optimizing the loss function to demix dependencies between various task parameters and feature responses, and accelerating the algorithm optimization, significant progress has been made in separating neural activity into different task parameters and visualizing the demixed components.
Article
Acoustics
Pei Yi Siow, Zhi Chao Ong, Shin Yee Khoo, Kok-Sing Lim
Summary: This paper proposes an integration strategy of a modal-based method into a hybrid machine learning-based method for damage detection and localization of beam-like structures. By combining unsupervised and supervised methods, the cold-start and manual labeling issues are addressed, and high accuracy results are achieved.
JOURNAL OF VIBRATION AND CONTROL
(2023)
Article
Quantum Science & Technology
Max Hunter Gordon, M. Cerezo, Lukasz Cincio, Patrick J. Coles
Summary: Principal component analysis (PCA) is a dimensionality reduction method that involves diagonalizing the covariance matrix of a dataset. Recently, quantum algorithms for PCA based on diagonalizing a density matrix have been proposed. However, a concrete protocol for encoding the covariance matrix as a density matrix has been lacking. In this study, we address this gap by providing a simple means for preparing the covariance matrix for arbitrary quantum datasets or centered classical datasets. We also propose a method for uncentered classical datasets, which we interpret as PCA on a symmetrized dataset. We demonstrate the effectiveness of our method through numerical experiments on the MNIST handwritten digit dataset and molecular ground-state datasets.
Article
Computer Science, Interdisciplinary Applications
Jeffrey R. Armstrong, J. Quinn Campbell, Anthony J. Petrella
Summary: This study compared two methods for quantifying differences in geometric shapes of human lumbar vertebra using statistical shape modeling, and found similar results between the methods for specificity and generality. The Hybrid SSM showed less compactness compared to the Cartesian SSM.
COMPUTER METHODS AND PROGRAMS IN BIOMEDICINE
(2021)
Article
Engineering, Chemical
Xiangwei Zhang, Chunquan Li, Jialin Liang, Shanshan Yang, Fang Yuan, Shuilin Zheng, Jiabao Yi, Zhiming Sun
Summary: In this study, three catalytic membranes composed of 2D/2D hybrid g-C3N4/kaolinite with different iron species were prepared. The evaluation results showed that the γ-FeOOH nanosheets-modified sample had a higher water flux, while the Fe single atom-modified sample achieved improved bisphenol A removal efficiency. The Fe2O3 quantum dots-modified sample exhibited the highest rate constant. Overall, this study provided a systematic and quantitative assessment of 2D/2D hybrid g-C3N4/layered clay-based catalytic membranes for wastewater treatment.
JOURNAL OF MEMBRANE SCIENCE
(2023)
Article
Food Science & Technology
Anca Monica Brata, Daniel Chiciudean, Vlad Dumitru Brata, Dorin Popa, Gabriela O. Chiciudean, Iulia C. Muresan
Summary: This study conducted in two counties of Romania explores consumers' perception towards wine consumption and the decision process of choosing a certain type of wine. The findings indicate that consumers prioritize intrinsic cues and their own experience when selecting wine, with experienced individuals and those with higher income valuing notoriety more.
Article
Engineering, Biomedical
Chengjun Huang, Maoqi Chen, Yingchun Zhang, Sheng Li, Cliff S. Klein, Ping Zhou
Summary: This study presents a novel method to estimate a muscle's innervation zone (IZ) location from monopolar high density surface electromyography (EMG) signals. The method is based on the linear relationship between the 2nd principal component coefficients derived from PCA and the time delay of different channels. The developed method achieved comparable or superior performance to existing methods in both simulation and experimental evaluations.
IEEE TRANSACTIONS ON NEURAL SYSTEMS AND REHABILITATION ENGINEERING
(2023)
Article
Engineering, Civil
Pei Yi Siow, Zhi Chao Ong, Shin Yee Khoo, Kok-Sing Lim, Bee Teng Chew
Summary: Machine learning-based structural health monitoring methods have been extensively studied in recent years due to the availability of advanced information and sensing technology. These methods have excellent pattern recognition capabilities for complex problems. However, the main challenge is the requirement of pre-collected historical data for model training. To address this "cold-start" problem, a two-stage hybrid modal-machine learning damage detection scheme is proposed, which includes an unsupervised detection stage and a further identification stage. Modal-based methods are used to detect damage based on changes in global properties, while unsupervised learning is used to detect damage presence. The performance of this scheme in alleviating the cold-start issue is highlighted, showing that it can identify single and multiple damages accurately even with limited training samples.
SMART STRUCTURES AND SYSTEMS
(2023)
Article
Automation & Control Systems
Dusan Kojis, Masato Yasui
Summary: Overfitting refers to exaggerated prediction outcomes caused by spurious correlations, leading to misleadingly optimistic models. We investigate systematic overfitting occurring when modeling changes in bandwidths, positions, and line shapes using popular matrix-decomposition techniques in spectra with strongly overlapped bands, as these parameters amplify variance in steep spectral sections detached from any chemical or physical interpretations. Our analysis of profile shape geometry demonstrates how indirect assessment of lateral spectral changes distorts original spectral information and creates unnecessary complexity, resulting in misleading interpretations. We propose an alternative strategy to remedy this situation and aim to improve the stability of spectral modeling in the future.
CHEMOMETRICS AND INTELLIGENT LABORATORY SYSTEMS
(2023)
Article
Chemistry, Physical
Hongzhao Zhou, Wuyun Xiao, Haixia Liu, Tao Sun, Zhiyuan Li, Dongxi Wang
Summary: This paper analyzes the influencing factors of dataset construction on PCA-based n/? discrimination. Four datasets were constructed based on the normalization factor and the start point. After dimensionality reduction, the distributions of the datasets were examined and figure of merits were calculated. Results show that the tail-peak dataset has the best performance and outperforms charge comparison method.
RADIATION PHYSICS AND CHEMISTRY
(2023)
Article
Computer Science, Hardware & Architecture
Yiming Wen, Weize Yu
Summary: The paper proposes a novel hardware attack based on principal component analysis to efficiently break a leakage power analysis-resistant cryptographic circuit. The attack can remove all added false keys used in the LPA-resistant CC to expose the secret key, and only 2000 plaintexts are needed for success.
INTEGRATION-THE VLSI JOURNAL
(2021)
Article
Engineering, Electrical & Electronic
Cristian Rusu, Lorenzo Rosasco
Summary: Orthogonal projections are a common technique for dimensionality reduction in machine learning applications. This study focuses on approximating orthogonal matrices to achieve fast and accurate numerical computation. By solving an optimization problem over a set of structured matrices called extended orthogonal Givens transformations, which includes Givens rotations as a special case, an approximation is obtained. An efficient greedy algorithm is proposed to solve the problem, striking a balance between approximation accuracy and computation speed. The approach is relevant to spectral methods and is applied to PCA.
Review
Biochemical Research Methods
Sanghun Lee, Georg Hahn, Julian Hecker, Sharon M. Lutz, Kristina Mullin, Winston Hide, Lars Bertram, Dawn L. Demeo, Rudolph E. Tanzi, Christoph Lange, Dmitry Prokopenko
Summary: Genetic similarity matrices are commonly used to assess population substructure, and in this study, the performance of three different matrices was evaluated. The results showed that rare single-nucleotide variations (SNVs) provided the best clustering performance, with the unweighted Jaccard matrix being the most preferable for rare variants. The visual inspection and clustering metrics confirmed its superiority compared to other methods and different allele frequency cutoffs. The application of these matrices to Alzheimer's disease data sets and different populations revealed the role of rare variants in local and global population substructure.
BRIEFINGS IN BIOINFORMATICS
(2023)
