Article
Computer Science, Interdisciplinary Applications
Kei Hirose, Kanta Miura, Atori Koie
Summary: This article proposes a cluster-based LDA method that improves prediction accuracy through hierarchical clustering and cross-validation, while addressing computational efficiency.
COMPUTATIONAL STATISTICS & DATA ANALYSIS
(2023)
Article
Medicine, General & Internal
Ju-Yang Jung, Hyun-Young Lee, Eunyoung Lee, Hyoun-Ah Kim, Dukyong Yoon, Chang-Hee Suh
Summary: This study identified three clusters of patients with systemic lupus erythematosus (SLE) based on their initial laboratory findings. These clusters had different laboratory characteristics, but did not differ significantly in terms of damage and mortality rates.
JOURNAL OF CLINICAL MEDICINE
(2022)
Article
Computer Science, Interdisciplinary Applications
Sanjeev Bhatta, Ji Dang
Summary: This paper presents a novel quantum convolutional neural network (QCNN) approach for detecting damage to reinforced concrete (RC) buildings from images after an earthquake. The QCNN model is developed and trained using RC building damaged images collected from past earthquakes, and its performance is evaluated based on real-world RC building damaged images from the recent earthquake in Turkey in February 2023. Furthermore, the seismic damage detection accuracy obtained from the QCNN model is compared with various CNN architecture results.
COMPUTER-AIDED CIVIL AND INFRASTRUCTURE ENGINEERING
(2023)
Article
Automation & Control Systems
Shuyuan Lin, Xing Wang, Guobao Xiao, Yan Yan, Hanzi Wang
Summary: In this article, a novel hierarchical representation via message propagation (HRMP) method is proposed for robust model fitting, which combines consensus analysis and preference analysis to estimate parameters of multiple model instances from data corrupted by outliers. The method can accurately estimate the number and parameters of multiple model instances while handling multistructural data contaminated with a large number of outliers, outperforming state-of-the-art model fitting methods in terms of fitting accuracy and speed.
IEEE TRANSACTIONS ON INDUSTRIAL ELECTRONICS
(2021)
Article
Engineering, Multidisciplinary
Siddharth Singh Chouhan, Uday Pratap Singh, Utkarsh Sharma, Sanjeev Jain
Summary: In this study, a computer vision methodology was proposed to automate disease diagnosis system, utilizing a Hybrid Neural Network integrated with Superpixel clustering for image segmentation. The results showed superior performance in distinguishing disease regions for the two plants evaluated separately.
Article
Computer Science, Artificial Intelligence
Quan Wang, Fei Wang, Fuji Ren, Zhongheng Li, Feiping Nie
Summary: This paper presents a novel clustering optimization method called Un-LDA(CD), which uses a coordinate descent algorithm instead of the K-means algorithm to achieve better performance on complex clustering cases.
IEEE TRANSACTIONS ON KNOWLEDGE AND DATA ENGINEERING
(2023)
Article
Ecology
Sarkhan Badirli, Christine Johanna Picard, George Mohler, Frannie Richert, Zeynep Akata, Murat Dundar
Summary: Machine learning can be used to create an accurate and efficient method for classifying insect species, including both described and undescribed species. A deep hierarchical Bayesian model is proposed, which can classify samples based on the taxonomic hierarchy of insects. The combination of image and DNA data in the model leads to significant improvement in classification accuracy.
METHODS IN ECOLOGY AND EVOLUTION
(2023)
Article
Ecology
Kim Bjerge, Quentin Geissmann, Jamie Alison, Hjalte M. R. Mann, Toke T. Hoye, Mads Dyrmann, Henrik Karstoft
Summary: Cameras and computer vision have opened up new research opportunities in the study of insects, particularly in agriculture, epidemiology, evolution, ecology, and biodiversity monitoring. This study presents an algorithm that uses a simple taxonomy to hierarchically classify insects from images, improving accuracy at higher taxonomic ranks. The approach incorporates multitask learning, anomaly detection, and state-of-the-art convolutional neural networks to achieve reliable and automated classification of live insects.
ECOLOGICAL INFORMATICS
(2023)
Article
Computer Science, Artificial Intelligence
Yuxin Fang, Xinggang Wang, Rui Wu, Wenyu Liu
Summary: Recent studies show that hierarchical Vision Transformer (ViT) with a macro architecture of interleaved non-overlapped window-based self-attention & shifted-window operation can outperform convolutional neural networks (CNNs) in various visual recognition tasks. Self-attention is not the only choice for hierarchical ViT to achieve strong performance.
IEEE TRANSACTIONS ON PATTERN ANALYSIS AND MACHINE INTELLIGENCE
(2023)
Article
Computer Science, Artificial Intelligence
Ruizheng Su, Xiongwen Pang, Hui Wang
Summary: Federated learning is used to train machine learning models based on decentralised and private data to address statistical heterogeneity. Personalised federated learning methods have been explored to enhance feature learning effectiveness. The proposed method PCCFED utilizes parameter decoupling and meta-transfer to improve the performance of personalised models.
IET COMPUTER VISION
(2023)
Article
Environmental Sciences
Giovanni Scardino, Giovanni Scicchitano, Marco Chirivi, Pedro J. M. Costa, Antonio Luparelli, Giuseppe Mastronuzzi
Summary: Coastal monitoring is an important field that has been developed using different approaches, including video monitoring coupled with machine learning and computer vision techniques. LEUCOTEA is an innovative system that uses geophysical surveys, convolutional neural network, and optical flow techniques to assess tide and storm parameters through video records.
Article
Computer Science, Artificial Intelligence
Fa Zhu, Junbin Gao, Jian Yang, Ning Ye
Summary: Linear Discriminant Analysis (LDA) assumes samples from the same class are independently and identically distributed, which may lead to failure when there are multiple clusters within a class. This paper proposes a neighborhood linear discriminant analysis (nLDA) that defines scatter matrices based on a neighborhood of reverse nearest neighbors, eliminating the need for the i.i.d. assumption. Experimental results show that nLDA outperforms previous discriminators in terms of performance.
PATTERN RECOGNITION
(2022)
Article
Chemistry, Analytical
Idoia Ruiz, Joan Serrat
Summary: Recent studies have made significant progress in novelty detection, but current methods can only determine that a new sample is unknown. This work leverages hierarchical taxonomies to provide informative outputs for samples of novel classes and introduces a novel loss function for hierarchical novelty detection. The proposed method shows promising results in traffic sign recognition.
Article
Computer Science, Information Systems
Grover E. Castro Guzman, Andre Fujita
Summary: This paper discusses methods for analyzing time series data, introduces the application of convolution-based linear discriminant analysis (cLDA) in classifying time series data, and points out that cLDA performs better than other methods in practical data sets.
INFORMATION SCIENCES
(2021)
Article
Computer Science, Artificial Intelligence
Gyoung S. Na
Summary: This paper proposes a new hierarchical approach to learning rate adaptation in gradient methods, called learning rate optimization (LRO). LRO optimizes the learning rate based on the alternating direction method of multipliers (ADMM), without requiring second-order information and probabilistic models, and without any additional hyperparameters. In experiments, SGD and Adam with LRO outperformed other methods in image classification tasks.
Article
Energy & Fuels
Bingwei Chen, Yu'na Kan, Shengcheng Zhai, Changtong Mei, Caoxing Huang, Qiang Yong
Summary: This study investigated the relationship between variations in chromatic values of biomass during different pretreatments and substrate digestibility. The results showed that luminance and total color difference were affected by the pretreatment conditions and chemical composition. Microscopic analysis and BET analysis confirmed that the severity of pretreatments increased the porosity of the cell wall. Moreover, the increment in total color difference was positively correlated with enzymatic hydrolysis efficiency, providing a rough yet high-throughput indicator for predicting saccharification efficiency of the giant reed.
BIOMASS CONVERSION AND BIOREFINERY
(2023)
Article
Forestry
Sung-Wook Hwang, Junji Sugiyama
Summary: This paper discusses image partitioning strategies that preserve spatial information of wood cross-sectional images, achieving better recognition performance than traditional methods. Radial partitioning is particularly effective for radial-porous species and can provide auxiliary correlation coefficients for multi-feature datasets. The contribution of image partitioning strategies is limited to species recognition and is not significant at the genus level.
JOURNAL OF WOOD SCIENCE
(2021)
Article
Forestry
Sung-Wook Hwang, Hiroshi Isoda, Takafumi Nakagawa, Junji Sugiyama
Summary: This study investigated the flexural anisotropy of wood and discovered different deformation mechanisms of various wood grain patterns. Rift-sawn softwood demonstrated excellent flexibility, largely due to the tapered shape of the tracheid walls.
JOURNAL OF WOOD SCIENCE
(2021)
Review
Biochemical Research Methods
Sung-Wook Hwang, Junji Sugiyama
Summary: The advancements in computer vision and machine learning have revolutionized scientific disciplines and created a new research field in wood science known as computer vision-based wood identification. Research has reviewed mainstream studies using machine learning procedures to familiarize wood scientists with this area and help them choose appropriate techniques in wood science.
Article
Materials Science, Paper & Wood
Shingo Kiyoto, Junji Sugiyama
Summary: The study revealed that the outer layer and inner material of birch cork mainly consist of suberin, while the inner layer and compound middle lamella are composed of lignin, cellulose, and other polysaccharides. Cellulose microfibrils in the inner layer bear tensile loads. Additionally, in the wet state, water and cellulose transfer tensile stress.
Article
Multidisciplinary Sciences
Sung-Wook Hwang, Un Taek Hwang, Kyeyoung Jo, Taekyeong Lee, Jinseok Park, Jong-Chan Kim, Hyo Won Kwak, In-Gyu Choi, Hwanmyeong Yeo
Summary: This study established prediction models for the non-destructive evaluation of lignin-derived hydrochars' carbonization characteristics using near-infrared spectra, accurately predicting carbon content, oxygen/carbon, and hydrogen/carbon ratios with high coefficients of determination and low root mean square errors. The models showed better prediction performance compared to ordinary least squares regression models.
SCIENTIFIC REPORTS
(2021)
Article
Chemistry, Applied
YunJin Kim, Junsik Bang, Jungkyu Kim, June-Ho Choi, Sung-Wook Hwang, Hwanmyeong Yeo, In-Gyu Choi, Hyoung-Joon Jin, Hyo Won Kwak
Summary: Nanocellulose is gaining interest as an eco-friendly water treatment material due to its large specific surface area and abundant hydroxyl functional groups, but its dispersion state hampers practical usage. A regenerated cellulose hydrogel with cationic functional groups was developed to address the low performance of nanocellulose in removing anionic pollutants. The PEI surface cationization process improved mechanical rigidity and showed excellent Cr(VI) removal capacity, maintaining high efficiency after multiple reuses.
CARBOHYDRATE POLYMERS
(2022)
Article
Materials Science, Paper & Wood
Ryo Nagamine, Kayoko Kobayashi, Ryosuke Kusumi, Masahisa Wada
Summary: This study is the first to compare the biodegradability of cellulose fibers with different structures in natural waters. The results showed that cellulose fibers are easily degraded into fine fragments, but it is difficult to completely decompose them into water and carbon dioxide. Additionally, the degradation behavior of cellulose fibers was influenced by different water environments.
Article
Engineering, Environmental
Yuna Kan, Shengcheng Zhai, Bingwei Chen, Mingzhu Pan, Xiaodong Fan, Weiqi Leng
Summary: The study investigated the effect of using decayed lignocellulosic biomass filler on the preparation and properties of polyurethane foam. The results showed that the addition of decayed lignocellulosic biomass improved the physical properties of the foam, including low density, low thermal conductivity, and high water absorption. These properties were found to be dependent on the amount of filler added.
JOURNAL OF POLYMERS AND THE ENVIRONMENT
(2022)
Article
Materials Science, Paper & Wood
Yangyang Zhang, Qinfeng He, Kayoko Kobayashi, Ryosuke Kusumi, Masahisa Wada
Summary: Hydrogels prepared from enzymatically synthesized dextran and carboxymethyl cellulose showed different properties depending on the CMC and dextran content. They exhibited excellent adsorption capabilities for heavy metal ions and dyes, and demonstrated potential for multiple reuses.
Article
Agricultural Engineering
Jiahui Wei, Hao Ren, Huamin Zhai, Shengcheng Zhai
Summary: Understanding the morphological and chemical changes of the cell wall is crucial for optimizing the processing conditions of lignocellulose deconstruction. In this study, subcritical water autohydrolysis of naked oat stems was conducted, and the depolymerization behaviors and microstructural changes were analyzed. The results provide valuable insights into the dynamic depolymerization mechanism of lignocellulosic composite cell walls under subcritical water autohydrolysis conditions.
INDUSTRIAL CROPS AND PRODUCTS
(2022)
Article
Forestry
Hairi Cipta, Kayoko Kobayashi, Shuoye Chen, Junji Sugiyama
Summary: This study analyzed the wood grain of Cinnamomum camphora using X-ray computed tomography, PIV, and 2D-FFT. The findings revealed periodic changes in wood grain orientation, with a wide range of grain angle variation and minor deviation in vessel inclining direction.
JOURNAL OF WOOD SCIENCE
(2022)
Article
Agricultural Engineering
Shengcheng Zhai, Yu ' na Kan, Siqi Lv, Bingwei Chen, Enhui Sun, Mingzhu Pan
Summary: This study investigates the liquefaction behavior and properties of brown-rotted wood, focusing on the effect of chemical composition and reaction time. The results show that brown-rotted pine with high lignin content has higher liquefaction efficiency. The reaction time also affects the yield and hydroxyl number of bio-polyol and phenolic products. The study concludes that the liquefaction of brown-rotted wood is feasible and has potential application advantages.
INDUSTRIAL CROPS AND PRODUCTS
(2023)
Article
Chemistry, Physical
Tomoki Ito, Kazuho Daicho, Shuji Fujisawa, Tsuguyuki Saito, Kayoko Kobayashi
Summary: Atomic-scale dent structures were found on the surfaces of cellulose nanofibers, constituting a significant portion of the total length and leading to kinking and fragmentation of the nanofibers.
NANOSCALE HORIZONS
(2022)
Article
Forestry
Sung-Wook Hwang, Taekyeong Lee, Hyunbin Kim, Hyunwoo Chung, Jong Gyu Choi, Hwanmyeong Yeo
Summary: This paper discusses feature-based techniques for wood knot classification, comparing the performance of texture and local feature descriptors and determining that texture descriptors are more suitable for wood knot classification. Additionally, it confirms that artificial neural network models are better suited for wood knot classification tasks.