Article
Biochemical Research Methods
Pegah Mavaie, Lawrence Holder, Michael K. Skinner
Summary: A hybrid learning approach that combines deep learning and non-deep learning methods is proposed in this study. The approach trains a deep network to extract features and then uses these features for classification. Results show that this hybrid approach outperforms standalone deep and non-deep learning methods, especially on low-volume, high-dimensional datasets. The robustness of the approach is further validated on diverse datasets from different domains.
BMC BIOINFORMATICS
(2023)
Article
Computer Science, Artificial Intelligence
Danilo Giordano, Eliana Pastor, Flavio Giobergia, Tania Cerquitelli, Elena Baralis, Marco Mellia, Alessandra Neri, Davide Tricarico
Summary: In the current automotive scenario, cars are equipped with thousands of sensors collecting data that can now be transferred to the cloud for advanced analytics like predictive maintenance. This study dissects a data-driven prognostic pipeline applied in the automotive field, aiming to predict failures of the High-Pressure Fuel System. The research highlights the importance of accurate model selection to identify a robust model suitable for deployment.
EXPERT SYSTEMS WITH APPLICATIONS
(2021)
Article
Computer Science, Artificial Intelligence
Tianxun Zhou, Keng-Hwee Chiam
Summary: Knowledge distillation is a technique to compress a large neural network into a smaller one while maintaining its performance. Existing methods mainly focus on classification tasks and require access to the training data. To tackle knowledge distillation for regression tasks without original data, a new method is proposed to generate synthetic data using adversarial training. Experimental results show that the proposed strategy enables the student model to learn better and closely emulate the teacher model's performance.
EXPERT SYSTEMS WITH APPLICATIONS
(2023)
Article
Mathematics
Stefan Rohrmanstorfer, Mikhail Komarov, Felix Moedritscher
Summary: With the increasing amount of image data, automatic search and analysis of image information has become necessary. By studying and implementing different methods, features are successfully extracted from fashion data using convolutional neural networks and TensorFlow to build image classification models.
Article
Remote Sensing
Shuai Liu, Mulan Gao
Summary: By combining SJCR and MFELM, a decision fusion framework is constructed to classify hyperspectral images with limited training samples, effectively addressing the classification problem.
INTERNATIONAL JOURNAL OF REMOTE SENSING
(2022)
Article
Computer Science, Artificial Intelligence
Sathishkumar Karupusamy, J. Refonaa, Sakthidasan Sankaran, Priyanka Dahiya, Mohd Anul Haq, Anil Kumar
Summary: The development of IoT systems has created higher demands for processing and storage environment, as well as raised issues such as energy consumption and data compression. This article proposes an approach that utilizes data mining to optimize energy usage and compress data, validated through the analysis of driving behavior.
Article
Computer Science, Information Systems
Zhen Liu, Chuxin Chen, Qianli Ma
Summary: This paper proposes a Category-Aware Optimal Transport Neural Network (CAOT-NN) for classification with incomplete data, which combines the imputation of missing values and classification learning with category information optimization. Extensive experiments on real-world and synthetic datasets demonstrate the superiority of CAOT-NN over existing methods in classification performance.
INFORMATION SCIENCES
(2023)
Article
Green & Sustainable Science & Technology
G. Revathy, Saleh A. Alghamdi, Sultan M. Alahmari, Saud R. Yonbawi, Anil Kumar, Mohd Anul Haq
Summary: Effective data science approaches are used to classify sentiments and emotions of a person on social media, with the use of neural-fuzzy and optimization algorithms. The technical contribution of this article is the double feed forward neural network, which outperforms existing algorithms in terms of classification parameters.
SUSTAINABLE ENERGY TECHNOLOGIES AND ASSESSMENTS
(2022)
Article
Chemistry, Analytical
Qi Wang, Daniel Smythe, Jun Cao, Zhilin Hu, Karl J. Proctor, Andrew P. Owens, Yifan Zhao
Summary: This study investigates the cognitive load in driving environment using EEG recordings and demonstrates the feasibility of using EEG as an indicator of changes in cognitive load. The results show that high accuracy classification of driving conditions can be achieved through statistical features and multi-frequency band analysis. The findings of this study are significant for improving the Human-Machine Interface of vehicles and enhancing safety.
Article
Multidisciplinary Sciences
Adam Joseph Ronald Pond, Seongwon Hwang, Berta Verd, Benjamin Steventon
Summary: Machine learning approaches are increasingly common in various research fields due to the large amount of data available for training models. Recent research shows that high accuracy models can be trained even with small data sets.
Article
Computer Science, Information Systems
Muhammad Usama, Dong Eui Chang
Summary: This paper addresses the problem of learning fair data representations in the vendor-user setting. It proposes splitting the latent space into sensitive and non-sensitive latent variables using maximum mean discrepancy (MMD) to induce statistical independence. Experimental results show that the proposed method achieves better or comparable performance at the utility task while simultaneously achieving sub-group and group fairness.
Article
Chemistry, Multidisciplinary
Jie Li, Huailian Tan, Wentao Huang
Summary: This paper introduces a PC-based data exploration method that dynamically adjusts classifier settings to identify new patterns during the exploration process. It also proposes a PC-based visualization framework that allows analysts to explore an exploring space simultaneously.
APPLIED SCIENCES-BASEL
(2022)
Article
Medicine, Legal
Chihyun Park, Joon-bae Lee, Wooyong Park, Dong-kye Lee
Summary: Using a practical GC-MS dataset of approximately 4000 suspected arson cases, researchers developed three machine learning models and evaluated their performances. The models were trained to classify fire residue into six categories: no fire accelerants detected or one of the accelerants being gasoline, kerosene, diesel, solvents, or candles. The classification accuracies of the random forest, support vector machine, and convolutional neural network models were 0.88, 0.88, and 0.92, respectively. By analyzing the feature importance of the random forest model, potential chemical fingerprints of fire accelerants were discovered.
FORENSIC SCIENCE INTERNATIONAL
(2023)
Article
Computer Science, Interdisciplinary Applications
Shi-Zhi Chen, De-Cheng Feng
Summary: This paper discusses a machine learning approach based on multifidelity data to improve the accuracy of prediction models in structural behavior. The feasibility of this method is validated through a case study, and the impact of different factors on model performance is thoroughly investigated.
COMPUTER-AIDED CIVIL AND INFRASTRUCTURE ENGINEERING
(2022)
Article
Engineering, Environmental
Yanlong Liu, Wenli Yao, Fenghui Qin, Lei Zhou, Yian Zheng
Summary: This study developed and compared four machine learning-based classifiers and used two large-scale blended plastic datasets to identify and classify microplastics (MPs). The 1D CNN model achieved the highest overall accuracy of 96.43% on a small dataset and outperformed other models. It was found that the RF model was the most robust with less spectral data, while the 2D CNN and RF models could be evaluated for plastic identification with fewer spectral data. An open-source MP spectroscopic analysis tool was developed for quick and accurate analysis of existing MP samples.
ENVIRONMENTAL SCIENCE & TECHNOLOGY
(2023)
