Article
Computer Science, Information Systems
Yanhui Guo, Qian Yu, Yang Gao, Xudong Liu, Chenglong Li
Summary: This paper proposes an unsupervised hyperspectral image classification framework, which achieves competitive performance on widely used hyperspectral datasets through the adoption of multi-scale spatial features, deep autoencoder with max-min distance embedding, and k-means clustering.
INTERNET OF THINGS
(2023)
Article
Chemistry, Analytical
Changwan Ko, Jaeseung Baek, Behnam Tavakkol, Young-Seon Jeong
Summary: Cluster validity indices (CVIs) are critical measures for evaluating optimal cluster number in clustering problems. Most CVIs are designed for certain data objects that have no uncertainty. In this study, new CVIs are proposed for uncertain clusters with arbitrary shapes, sub-clusters, and noise. By transforming uncertain data into kernel spaces, the proposed CVI accurately measures the compactness and separability of a cluster for arbitrary cluster shapes and is robust to noise and outliers.
Article
Computer Science, Artificial Intelligence
Anjali Patel, Subhankar Jana, Juthika Mahanta
Summary: Due to the potential harm to public health and the environment, medical waste treatment has become a critical concern, especially in developing nations. This study provides a hybrid multi-criteria decision-making technique for assessing medical waste treatment techniques, considering social, environmental, economic, and technical factors. The proposed technique, IF-EM-SWARA-TOPSIS, utilizes intuitionistic fuzzy sets and incorporates subjective and objective weights for criteria evaluation and alternative ranking.
APPLIED SOFT COMPUTING
(2023)
Article
Mathematics
Dalila Azzam-Laouir
Summary: This paper establishes the existence and uniqueness result of a right continuous bounded variation solution for a perturbed differential inclusion governed by time-dependent maximal monotone operators.
TOPOLOGICAL METHODS IN NONLINEAR ANALYSIS
(2022)
Article
Computer Science, Artificial Intelligence
Letao Qu, Bohyun Wang, Joon S. Lim
Summary: Distance measures of fuzzy sets play a key role in feature selection and redundant feature identification, with subnormal and non-convex fuzzy sets offering more precise definitions. Weighted fuzzy membership functions help prevent combinatorial explosion of fuzzy rules, resulting in higher accuracies in training and testing compared to other methods.
JOURNAL OF INTELLIGENT & FUZZY SYSTEMS
(2021)
Article
Education, Special
D. Shashi Kala, M. P. Indra Gandhi
Summary: This paper presents an offline handwritten Tamil character recognition system that focuses on recognizing characters regardless of their writing style. The system employs feature extraction and feature selection techniques, using the 8 directional chain code method and centroid based Chessboard distance measure to generate a feature vector that is invariant to changes in size, translation, and rotation. The system is implemented using a Neural Network algorithm in matlab to achieve better recognition rate. Experiments on a set of 12 vowel Tamil characters show the performance of the NN classifier using confusion matrix, ROC, and AUC curve.
INTERNATIONAL JOURNAL OF EARLY CHILDHOOD SPECIAL EDUCATION
(2022)
Proceedings Paper
Business
Fabian Roesel, Stephan A. Fahrenkog-Petersen, Han van der Aa, Matthias Weidlich
Summary: To enable process analysis based on event logs while protecting privacy, logs can be anonymized using feature learning-based distance measures. This approach helps to better preserve the utility of the original log by incorporating semantic information of activities. Existing syntactic measures may neglect semantics, leading to merged events of unrelated activities in anonymization. Experiments with real-world data show that using feature learning-based measures yields logs closer to the original log and higher utility for process analysis.
BUSINESS PROCESS MANAGEMENT WORKSHOPS, BPM 2021
(2022)
Proceedings Paper
Automation & Control Systems
Ning Ouyang, Junyan Wang, Peng Jiang, Xiaodong Cai
Summary: Generating accurate and effective short text summaries is challenging. This study proposes a text summarization generation model called GMELC, which aims to enhance the local correlation in generated summaries. The model introduces the residual concept and a scaled I-2 normalization method to improve the dependencies of words in phrases and reduce unnecessary computation, resulting in higher recall and better readability of the generated summaries.
2022 IEEE 6TH ADVANCED INFORMATION TECHNOLOGY, ELECTRONIC AND AUTOMATION CONTROL CONFERENCE (IAEAC)
(2022)
Article
Biotechnology & Applied Microbiology
Dongho Kwak, Thomas Combriat, Alexander Refsum Jensenius, Petter Angell Olsen
Summary: This paper describes an innovative experimental setup that uses audio technology principles to subject adherent cells to rhythmic vertical vibrations. The setup's performance is evaluated using a new approach that combines three-axis acceleration measurements and particle tracking velocimetry. The findings show that mechanical stimuli have a significant impact on the size and orientation of F-actin filaments, as well as the accumulation of cells in the G1 phase of the cell cycle.
BIOENGINEERING-BASEL
(2023)
Article
Computer Science, Information Systems
Gang Sun, Hancheng Yu, Xiangtao Jiang, Mingkui Feng
Summary: This article introduces an edge detection method based on a convolutional neural network, which utilizes a non-maximum suppression layer to obtain sharp boundaries and proposes an ODS F-measure loss function for training. Additionally, an adaptive multi-level feature pyramid network and a pyramid context module are introduced to improve feature fusion and extraction.
Review
Biochemical Research Methods
Xiaoqing Ru, Xiucai Ye, Tetsuya Sakurai, Quan Zou
Summary: Learning to rank algorithms have been gradually applied to bioinformatics over the past decades, showing significant advantages in various research tasks. This paper analyzes the characteristics and strengths of LTR algorithms compared to other types of algorithms in bioinformatics, discussing ways to better utilize them and addressing current open problems.
BRIEFINGS IN BIOINFORMATICS
(2021)
Article
Biochemical Research Methods
Xiaoqing Ru, Xiucai Ye, Tetsuya Sakurai, Quan Zou
Summary: Motivation: Predicting drug-target binding affinity is crucial in drug discovery and repurposing. However, existing methods face challenges like focusing on one application scenario and not considering the priority order of proteins related to each target drug. Results: The proposed NerLTR-DTA method utilizes neighbor relationship, similarity, and sharing to predict affinity values and priority order, achieving excellent performance in multiple scenarios and outperforming state-of-the-art methods on commonly used datasets. This comprehensive tool can accurately rank drug-protein associations, contributing to new drug discoveries and repurposing efforts.
Article
Biochemical Research Methods
Xiaoqing Ru, Quan Zou, Chen Lin
Summary: Motivated by the need for high accuracy and interpretability, this study explores various feature selection and dimensionality reduction techniques to optimize drug-target affinity prediction models. Experimental results demonstrate that regression tree-based feature selection is most effective in constructing models with good performance and robustness. Moreover, the study identifies a high-quality feature subset and highlights the breakthrough impact of the top 20D features on prediction. This research emphasizes the importance of feature optimization in constructing high-performance and interpretable models for drug-target affinity prediction.