Article
Computer Science, Artificial Intelligence
Yunhe Zhang, Zhoumin Lu, Shiping Wang
Summary: Feature selection is crucial in machine learning, and the proposed unsupervised feature selection scheme based on auto-encoder can solve traditional constrained feature selection problems and adapt to various loss functions and activation functions.
KNOWLEDGE-BASED SYSTEMS
(2021)
Article
Computer Science, Artificial Intelligence
Jinyu Cai, Shiping Wang, Wenzhong Guo
Summary: The paper proposes a deep stacked sparse embedded clustering method that considers both local structure preservation and input sparsity. The deep learning approach jointly learns clustering-oriented features and optimizes cluster label assignments by minimizing both the reconstruction and clustering loss. Comprehensive experiments validate the effectiveness of introducing sparsity and preserving local structure in the proposed method.
EXPERT SYSTEMS WITH APPLICATIONS
(2021)
Article
Computer Science, Hardware & Architecture
Yaonan Guan, Yunwen Xu, Yugeng Xi, Dewei Li
Summary: This paper proposes a method to enhance data feature extraction capabilities by introducing a feedback channel between the Variational Auto Encoder and the Gaussian process. The model addresses common mislabeling issues in unsupervised settings by selecting representative data and constructing a full-data Gaussian process model. By employing a Bayesian framework to guide data point displacement in the latent space, the model effectively discriminates between normal and abnormal data.
JOURNAL OF SUPERCOMPUTING
(2023)
Article
Geochemistry & Geophysics
Huilin Xu, Wei He, Liangpei Zhang, Hongyan Zhang
Summary: This paper proposes an unsupervised deep semantic feature learning network (S3FN) for hyperspectral images (HSIs), which learns spectral-spatial features from a high-level semantic perspective and aligns these features using a contrastive loss function. Experimental results show that S3FN can achieve promising classification results with lower time cost compared to other state-of-the-art unsupervised FE methods.
IEEE TRANSACTIONS ON GEOSCIENCE AND REMOTE SENSING
(2022)
Article
Geochemistry & Geophysics
Ambar Perez-Garcia, Mercedes E. Paoletti, Juan M. Haut, Jose F. Lopez
Summary: This letter presents a new loss function to improve the performance of unsupervised hyperspectral image (HSI) segmentation models. The spectral loss function, Sl, based on the purity of the unmixing endmembers and the spectral similarity of the clusters, is incorporated into a 3-D convolutional autoencoder to validate its performance. The results show that Sl is a breakthrough in unsupervised HSI segmentation, highlighting the importance of spectral signatures.
IEEE GEOSCIENCE AND REMOTE SENSING LETTERS
(2023)
Article
Engineering, Multidisciplinary
Zilong Wang, Young-Jin Cha
Summary: This article proposes an unsupervised deep learning-based approach for structural damage detection, which utilizes a carefully designed deep auto-encoder and one-class support vector machine for extracting damage-sensitive features and detecting future damage. Experimental and numerical studies confirm the high accuracy and stability of the method in detecting structural damage.
STRUCTURAL HEALTH MONITORING-AN INTERNATIONAL JOURNAL
(2021)
Article
Medicine, General & Internal
Shengchen Li, Ke Tian
Summary: This study proposes two methods based on estimating the density of vectors in the latent space to improve the performance of PCG analysis systems, showing that these methods outperform VAE-based methods; in addition, the representation of normal PCG signals in the latent space is investigated, with DBAE introducing Gaussian-like models for normal PCG representation.
FRONTIERS IN MEDICINE
(2021)
Review
Chemistry, Analytical
Ikram Eddahmani, Chi-Hieu Pham, Thibault Napoleon, Isabelle Badoc, Jean-Rassaire Fouefack, Marwa El-Bouz
Summary: In recent years, the rapid development of deep learning approaches has advanced the exploration of underlying factors in data explanation. However, the challenge of extracting this representation with little or no supervision remains. This paper provides a theoretical overview of unsupervised representation learning, focusing on auto-encoding-based approaches and well-known supervised disentanglement metrics. It covers the state-of-the-art methods for unsupervised disentangled representation learning, quantifies disentanglement, and evaluates associated metrics based on modularity, compactness, and informativeness. The Mutual Information Gap score (MIG) is found to meet all three evaluation criteria.
Article
Geochemistry & Geophysics
Shuyu Zhang, Meng Xu, Jun Zhou, Sen Jia
Summary: This article proposes an unsupervised multiscale and diverse feature learning (UMsDFL) method for the classification of hyperspectral images (HSIs). The method utilizes convolutional neural networks (CNNs) to consider the spatial-spectral features and employs superpixel segmentation for feature learning. The proposed method achieves high classification accuracy on real-world HSI datasets.
IEEE TRANSACTIONS ON GEOSCIENCE AND REMOTE SENSING
(2022)
Article
Geochemistry & Geophysics
Yubao Sun, Ying Yang, Qingshan Liu, Mohan Kankanhalli
Summary: Hyperspectral compressive imaging utilizes compressive sensing theory to achieve coded aperture snapshot measurement without temporal scanning, focusing on reconstructing underlying hyperspectral images. An unsupervised spatial-spectral network is proposed for HSI reconstruction, optimizing network parameters to adapt to different imaging settings, resulting in better reconstruction results than current state-of-the-art methods.
IEEE TRANSACTIONS ON GEOSCIENCE AND REMOTE SENSING
(2022)
Article
Geochemistry & Geophysics
Yifan Sun, Bing Liu, Xuchu Yu, Anzhu Yu, Kuiliang Gao, Lei Ding
Summary: In recent years, significant development has been achieved in deep-learning-based hyperspectral image (HSI) classification methods. This article proposes a framework called SMF-UL, which learns spectral variation knowledge through unsupervised learning on mass unlabeled HSI data to obtain a robust capability of feature extraction with generalization. Experimental results show that SMF-UL achieves competitive classification performance and demonstrates flexibility and superiority compared to advanced methods.
IEEE TRANSACTIONS ON GEOSCIENCE AND REMOTE SENSING
(2022)
Article
Computer Science, Artificial Intelligence
Jie Lei, Meiqi Li, Weiying Xie, Yunsong Li, Xiuping Jia
Summary: A novel unsupervised hyperspectral change detection (UHCD) framework is proposed in this paper, designed for hyperspectral images with high dimensions and low availability. The framework possesses distinctive properties including unsupervised spectral mapping, enhanced feature quality, and spatial attribute optimization.
Article
Engineering, Multidisciplinary
Shuai Teng, Zongchao Liu, Wenjun Luo, Gongfa Chen, Li Cheng
Summary: This study presents a novel approach for bridge anomaly detection using unsupervised convolutional auto-encoder. The method reconstructs real-time vibration signals and calculates damage indicators based on the reconstruction results to detect abnormal signals. Experimental results show that the proposed approach achieves high accuracy in bridge anomaly detection and outperforms other methods in terms of accuracy. The method is also confirmed to be applicable to different types of bridges.
STRUCTURAL HEALTH MONITORING-AN INTERNATIONAL JOURNAL
(2023)
Article
Computer Science, Information Systems
Farshad Gheytasi, S. Hadi Yaghoubyan, Zahra Rezaei, Karamollah Bagherifard, Hamid Parvin
Summary: This study proposes a novel method based on extended deep learning and autoencoder ensembles, using the eagle strategy to address complexities in spectral clustering and data training gaps. Experimental results demonstrate that the proposed method outperforms previous algorithms on multiple metrics.
MULTIMEDIA TOOLS AND APPLICATIONS
(2023)
Article
Environmental Sciences
Chao Yao, Lingfeng Zheng, Longchao Feng, Fan Yang, Zehua Guo, Miao Ma
Summary: Background: The importance of superpixel-based methods in hyperspectral image processing. Objective: Addressing the issue of increased intra-class disparity caused by superpixel-based methods. Innovation: Introducing the unsupervised DR method Collaborative superpixelwise Auto-Encoder (ColAE). Results: ColAE achieves promising performance compared to existing methods on three hyperspectral datasets.
Article
Chemistry, Physical
Klytaimnistra Katsara, Konstantina Psatha, George Kenanakis, Michalis Aivaliotis, Vassilis M. Papadakis
Summary: Raman spectroscopy is a precise analysis technique that can be used for non-destructive and label-free cell characterization. This study proposes a fast characterization and differentiation methodology of lymphoma cell subtypes based on Raman spectroscopy, which is fast, accurate, and requires minimal sample preparation.
Article
Agronomy
Dimitrios Fanourakis, Vassilis M. Papadakis, Evangelos Psyllakis, Vasileios A. Tzanakakis, Panayiotis A. Nektarios
Summary: Long storage periods decrease the vase life of cut flowers. The duration of storage linearly reduces vase life, and this rate of decrease varies among different cultivars. Storage duration does not affect stomatal functioning in leaves, non-leaf tissue transpiration, or the relative contribution of each organ to whole-cut flower transpiration. Variation in oxidative state is the main factor determining the cultivar differences in vase life response to cold storage.
Article
Chemistry, Physical
Nectarios Vidakis, Markos Petousis, Panagiotis Mangelis, Emmanuel Maravelakis, Nikolaos Mountakis, Vassilis Papadakis, Maria Neonaki, Georgia Thomadaki
Summary: Polycarbonate-based nanocomposites were successfully developed through a material extrusion additive manufacturing process. The addition of AlN nanoparticles reinforced most of the mechanical properties, with the nanocomposite containing 2 wt.% filler concentration exhibiting the best mechanical performance. This research expands the application of 3D printing technology and provides a new approach for developing nanocomposite materials with multifunctional properties for industrial applications.
Article
Chemistry, Multidisciplinary
Markos Petousis, Nectarios Vidakis, Nikolaos Mountakis, Vassilis Papadakis, Lazaros Tzounis
Summary: This study reports the effects of adding aluminum oxide nanoparticles as reinforcing agents in Polyamide 12 and Polylactic acid in fused filament fabrication. The nanoparticles demonstrated positive reinforcement mechanism and achieved maximum performance improvement at specific filler loadings.
Article
Engineering, Biomedical
Nectarios Vidakis, Markos Petousis, Nikolaos Mountakis, Vassilis Papadakis, Amalia Moutsopoulou
Summary: The efficacy of three modeling methods (full factorial, Taguchi, Box-Behnken) on the performance of MEX 3D printed nanocomposites was investigated in order to reduce experimental effort. The results showed that a mixture of 3% CNF, 270°C nozzle temperature, and 80°C bed temperature led to a 24% increase in tensile strength in the medical-grade PA12 reinforced with CNF. The Taguchi and Box-Behnken methods required only 7.4% and 11.8% of the experimental effort of the full factorial method, respectively.
JOURNAL OF THE MECHANICAL BEHAVIOR OF BIOMEDICAL MATERIALS
(2023)
Article
Chemistry, Multidisciplinary
Nectarios Vidakis, Panagiotis Mangelis, Markos Petousis, Nikolaos Mountakis, Vassilis Papadakis, Amalia Moutsopoulou, Dimitris Tsikritzis
Summary: Acrylonitrile Butadiene Styrene (ABS) nanocomposites with Titanium Nitride (TiN) nanoparticles were developed using Material Extrusion (MEX) Additive Manufacturing (AM) and Fused Filament Fabrication (FFF) methods. Mechanical tests and morphological characterization were conducted to investigate the effect of TiN nanoparticles on the mechanical performance and surface characteristics of the nanocomposites. The inclusion of 6 wt. % of TiN nanoparticles significantly improved the mechanical properties of the ABS/TiN composites, including flexural modulus of elasticity and toughness.
Article
Chemistry, Multidisciplinary
Markos Petousis, Nikolaos Michailidis, Vassilis M. Papadakis, Apostolos Korlos, Nikolaos Mountakis, Apostolos Argyros, Evgenia Dimitriou, Chrysa Charou, Amalia Moutsopoulou, Nectarios Vidakis
Summary: The aim of this research was to investigate the thermomechanical properties of new nanocomposites in additive manufacturing (AM). Material extrusion (MEX) 3D printing was used to fabricate ABS nanocomposites with silicon nitride nano-inclusions. Standard tests were conducted to evaluate the influence of Si3N4 nanofiller content on the mechanical and thermal response of the printed samples. The results showed that silicon nitride nanoparticles significantly enhanced the mechanical properties of the polymer matrix.
Article
Biochemistry & Molecular Biology
Vassilis M. Papadakis, Christina Cheimonidi, Maria Panagopoulou, Makrina Karaglani, Paraskevi Apalaki, Klytaimnistra Katsara, George Kenanakis, Theodosis Theodosiou, Theodoros C. Constantinidis, Kalliopi Stratigi, Ekaterini Chatzaki
Summary: In this study, Raman spectroscopy was used to analyze the biomolecular characteristics of cell-free DNA (ccfDNA) in the blood of individuals with different health conditions. The study found distinct spectral patterns in ccfDNA from breast cancer patients undergoing neoadjuvant therapy, as well as some differences in the biomolecular fingerprints of ccfDNA from healthy, prediabetic, and diabetic males. The analysis also confirmed that ccfDNA mirrors its cellular origin. Overall, the study highlights the potential of Raman spectroscopy as a new approach for liquid biopsy diagnostics.
INTERNATIONAL JOURNAL OF MOLECULAR SCIENCES
(2023)
Article
Chemistry, Physical
Nikolaos Rafael Vrithias, Klytaimnistra Katsara, Lampros Papoutsakis, Vassilis M. Papadakis, Zacharias Viskadourakis, Ioannis N. Remediakis, George Kenanakis
Summary: This study reports on the fabrication of high-density polyethylene sponges decorated with Mn-doped ZnO nanostructures using three-dimensional printing technology. Mn-doped ZnO nanostructures were grown at mild temperatures, and their characterization and photocatalytic properties were investigated. The results indicate that Mn-doped ZnO/HDPE nanostructures show better photocatalytic activity at higher doping levels, making them promising candidates for real environmental applications.
Article
Polymer Science
Nectarios Vidakis, Amalia Moutsopoulou, Markos Petousis, Nikolaos Michailidis, Chrysa Charou, Nikolaos Mountakis, Apostolos Argyros, Vassilis Papadakis, Evgenia Dimitriou
Summary: This paper investigates tungsten carbide (WC) as a reinforcement material in the widely used material extrusion (MEX) additive manufacturing (AM) process. The study demonstrates that WC can enhance and stabilize commonly used polymeric matrices in MEX 3D printing. The mechanical properties, structure, and thermomechanical properties of hybrid polymer/ceramic nanocomposites made with different filler loadings were fully characterized. The results show significant improvements in tensile strength, flexural strength, and microhardness, indicating the potential use of WC nanocomposites in wear-related applications.
Article
Materials Science, Composites
Amalia Moutsopoulou, Markos Petousis, Nikolaos Michailidis, Nikolaos Mountakis, Apostolos Argyros, Vassilis Papadakis, Mariza Spiridaki, Chrysa Charou, Ioannis Ntintakis, Nectarios Vidakis
Summary: In this study, innovative nanocomposite materials for material extrusion (MEX) 3D printing were developed using a polypropylene (PP) polymer with tungsten carbide (WC) nanopowder. The mechanical characteristics, thermal stability, and processability of the nanocomposite were extensively examined. The results showed that the filler significantly enhanced the mechanical characteristics of the matrix polymer without reducing its thermal stability or processability.
JOURNAL OF COMPOSITES SCIENCE
(2023)
Article
Materials Science, Multidisciplinary
Nectarios Vidakis, Markos Petousis, Nikolaos Michailidis, Nektarios Nasikas, Vassilis Papadakis, Apostolos Argyros, Nikolaos Mountakis, Chrysa Charou, Amalia Moutsopoulou
Summary: In this study, titanium carbide was evaluated as a reinforcing additive in polypropylene thermoplastic. The effects of changing the weight percentage of titanium carbide on the characteristics of the nanocomposite material were investigated. The addition of titanium carbide significantly improved the mechanical properties of the polypropylene material.
MATERIALS TODAY COMMUNICATIONS
(2023)
Article
Engineering, Biomedical
Nectarios Vidakis, Markos Petousis, Nikolaos Mountakis, Apostolos Korlos, Vassilis Papadakis, Amalia Moutsopoulou
Summary: In this study, polyamide 12 nanocomposites with enhanced mechanical response and antibacterial performance were developed using 3D printing. This research demonstrates the potential of 3D printing in expanding the applications of nanocomposite materials.
JOURNAL OF FUNCTIONAL BIOMATERIALS
(2022)
Article
Materials Science, Multidisciplinary
Markos Petousis, Nectarios Vidakis, Nikolaos Mountakis, Vassilis Papadakis, Sotiria Kanellopoulou, Aikaterini Gaganatsiou, Nikolaos Stefanoudakis, John Kechagias
Summary: This study presents an effective process for developing multifunctional nanocomposites for material extrusion 3D printing in industrial environments. Nanocomposites with binary inclusions were prepared and investigated. Specimens were built using a thermomechanical process according to international standards, and various tests were conducted. The thermal properties, morphological characteristics, and antibacterial performance of the nanocomposites were evaluated. The results showed improved mechanical properties and antibacterial performance compared to pure PLA material.