Article
Computer Science, Artificial Intelligence
Shixian Wen, Amanda Rios, Yunhao Ge, Laurent Itti
Summary: The human brain is a model of adaptive learning, being able to adapt to new situations and learn from experiences. A new biologically plausible deep neural network with task-dependent biasing units has been proposed to address catastrophic forgetting in continual learning of multiple tasks, achieving state-of-the-art performance across different tasks and domains.
IEEE TRANSACTIONS ON NEURAL NETWORKS AND LEARNING SYSTEMS
(2022)
Article
Automation & Control Systems
Mohammed Toum Benchekroun, Smail Zaki, Mohamed Aboussaleh
Summary: The process of cement manufacturing is energy intensive and hard to control, but predictive models can help optimize parameters and improve efficiency.
INTERNATIONAL JOURNAL OF ADVANCED MANUFACTURING TECHNOLOGY
(2023)
Article
Green & Sustainable Science & Technology
Fazal Hussain, Shayan Ali Khan, Rao Arsalan Khushnood, Ameer Hamza, Fazal Rehman
Summary: Nowadays, lightweight aggregate concrete is gaining popularity due to its versatile properties. This research proposes a simplified methodology for the mix designing of lightweight aggregate concrete by incorporating machine learning. Multiple machine learning algorithms were investigated and the Gaussian process of regression model outperformed the others in predicting the compressive strength of lightweight concrete. These simplified modern techniques can make the design of lightweight aggregate concrete easier without extensive experimentation.
Review
Construction & Building Technology
Mohammad Mohtasham Moein, Ashkan Saradar, Komeil Rahmati, Seyed Hosein Ghasemzadeh Mousavinejad, James Bristow, Vartenie Aramali, Moses Karakouzian
Summary: Concrete is widely used in civil engineering and its mechanical properties are important for design and evaluation. Machine learning and deep learning have been applied to predict these properties, offering advantages such as accuracy and responsiveness. This paper reviews successful applications of ML and DL models and provides suggestions for future research.
JOURNAL OF BUILDING ENGINEERING
(2023)
Article
Chemistry, Multidisciplinary
Zara Ahmed, Muhammad Umer Sohail, Asma Javed, Raees Fida Swati
Summary: This research conducted a parametric analysis of a low-bypass turbofan engine using machine learning and deep learning techniques. By investigating the design parameters that affect engine performance, regression models and deep neural networks were developed to accurately predict the performance parameters of the engine.
APPLIED SCIENCES-BASEL
(2023)
Article
Engineering, Environmental
Ma lgorzata Kida, Kamil Pochwat, Sabina Ziembowicz
Summary: The aim of this study was to evaluate the usefulness of machine learning in predicting the migration of pollutants from microplastics. The results showed that artificial neural networks and support vector machines have potential in modeling and predicting leaching of chemical compounds from microplastics.
JOURNAL OF HAZARDOUS MATERIALS
(2024)
Article
Geosciences, Multidisciplinary
Hossein Kheirollahi, Navid Shad Manaman, Ahsan Leisi
Summary: In recent decades, shear wave velocity has been widely used in oil and gas projects, including lithology determination, geomechanical parameter estimation, and reservoir fluid detection. However, most of the existing empirical rock physics models for estimating shear wave velocity are not applicable to carbonate reservoirs due to the complex fracture networks. This study focuses on the effect of lithology in estimating shear wave velocity in a carbonate reservoir. By preprocessing and applying multiple conventional well logs, a synthetic shear wave velocity log is generated and different data-driven predictive models including multiple linear regression, ensemble learning method, and artificial neural networks are built and compared. The feed-forward neural network with optimal design and tuning parameters shows the highest accuracy in predicting shear wave velocity.
JOURNAL OF APPLIED GEOPHYSICS
(2023)
Article
Computer Science, Artificial Intelligence
Sebastian J. Wetzel, Roger G. Melko, Isaac Tamblyn
Summary: Twin neural network regression (TNNR) predicts the differences between target values of different data points, and achieves accurate predictions by ensembling these differences. Semi-supervised training improves the already state of the art TNNR significantly.
MACHINE LEARNING-SCIENCE AND TECHNOLOGY
(2022)
Article
Automation & Control Systems
Daming Wang, Zheng John Shen, Xin Yin, Sai Tang, Xifei Liu, Chao Zhang, Jun Wang, Jose Rodriguez, Margarita Norambuena
Summary: In this article, a new approach called ANN-MPC is proposed as a solution to the increasing complexity and demand of computing resources in power electronic applications. The approach uses an artificial neural network to train an MPC controller and eliminates the need for heavy mathematical computation. Simulation and experimental results demonstrate that the FPGA-based ANN-MPC controller can significantly reduce the resource requirement while offering the same control performance as conventional MPC.
IEEE TRANSACTIONS ON INDUSTRIAL ELECTRONICS
(2022)
Article
Multidisciplinary Sciences
Martin Schrimpf, Idan Asher Blank, Greta Tuckute, Carina Kauf, Eghbal A. Hosseini, Nancy Kanwisher, Joshua B. Tenenbaum, Evelina Fedorenko
Summary: A recent approach in neuroscience has connected computation, brain function, and behavior to provide new insights into cognitive and neural mechanisms. Powerful transformer models in language processing can predict neural responses and correlate with model accuracy on next-word prediction task.
PROCEEDINGS OF THE NATIONAL ACADEMY OF SCIENCES OF THE UNITED STATES OF AMERICA
(2021)
Article
Automation & Control Systems
Mohammed Toum Benchekroun, Smail Zaki, Mohamed Aboussaleh
Summary: Occupational health and safety is of utmost importance in the cement industry. The cyclone preheaters in cement kiln plants are prone to coatings and blockages, which disrupt the flow of hot kiln feed and exhaust gases. Our research aims to use a digital transformation tool to monitor temperature and pressure in real time, predict failures, and prevent preheater cyclone blockages. This new technology will enhance occupational safety, improve industrial process efficiency, and increase productivity.
INTERNATIONAL JOURNAL OF ADVANCED MANUFACTURING TECHNOLOGY
(2023)
Article
Computer Science, Information Systems
Radwa Marzouk, Ala Saleh Alluhaidan, Sahar A. El Rahman
Summary: This study aimed to develop a personalized diabetes monitoring system based on machine learning techniques and a web-based platform. It utilized patients' QR cards to connect them with doctors and provide real-time health information. The study evaluated different machine learning algorithms and found that the Artificial Neural Network model had the highest prediction accuracy.
Article
Engineering, Environmental
N. P. Bakas, A. Langousis, M. A. Nicolaou, S. A. Chatzichristofis
Summary: We propose a numerical scheme for computing Artificial Neural Network (ANN) weights based on the Universal Approximation Theorem. This algorithm is fast, adheres to the underlying theory, and yields low errors in regression and classification of complex data sets, as well as in handwritten digit recognition. It is also capable of approximating highly nonlinear functions and numerically solving partial differential equations, with low errors and good generalization efficiency.
STOCHASTIC ENVIRONMENTAL RESEARCH AND RISK ASSESSMENT
(2023)
Article
Biochemistry & Molecular Biology
Evgeny A. Kozyrev, Evgeny A. Ermakov, Anastasiia S. Boiko, Irina A. Mednova, Elena G. Kornetova, Nikolay A. Bokhan, Svetlana A. Ivanova
Summary: This study utilizes machine learning and artificial intelligence technologies to analyze multi-domain data in precision psychiatry, developing five predictive models for the binary classification of schizophrenia patients and healthy individuals. Results show that the deep neural network algorithm has slightly higher sensitivity and specificity than other algorithms, and combining all variables into one classifier exceeds the effectiveness of individual variables, indicating the need for multiple biomarkers to diagnose schizophrenia.
Article
Computer Science, Information Systems
Zhan Shi, Chongjun Fan
Summary: This paper mainly studies the research and analysis of machine short text sentiment classification based on Bayesian network and deep neural network algorithm. It first introduces Bayesian network and deep neural network algorithms, and analyzes the comments of various social software such as Twitter, Weibo. The results show that the triplet dependency features have advantages in text representation and Bayesian and deep neural network show good advantages in short text emotion classification.
Article
Computer Science, Information Systems
You Zhang, Jin Wang, Xuejie Zhang
Summary: A multiview attention model was proposed for learning sentence representation, using multiple view vectors to map attentions from different perspectives and a fusion gate to combine them, improving the performance of previously proposed attention models.
INFORMATION SCIENCES
(2021)
Article
Computer Science, Artificial Intelligence
You Zhang, Jin Wang, Xuejie Zhang
Summary: The proposed interactive attributes attention model considers user and product information in customer reviews, and utilizes a bilinear interaction and multiloss objective function to align attribute features with text representations for improved sentiment polarity classification performance.
KNOWLEDGE-BASED SYSTEMS
(2021)
Article
Computer Science, Information Systems
Housheng Xie, Wei Lin, Shuying Lin, Jin Wang, Liang-Chih Yu
Summary: Dimensional sentiment analysis focuses on representing affective states as continuous numerical values on multiple dimensions, taking into account the relationships between different dimensions for better prediction accuracy. The proposed multi-dimensional relation model incorporates these relationships into deep neural networks, outperforming traditional models that treat each dimension independently. Internal mode, which integrates dimension relations before prediction, showed better performance than the external mode, and a combination of both modes achieved the best results.
INFORMATION SCIENCES
(2021)
Article
Computer Science, Artificial Intelligence
Jin Wang, You Zhang, Liang-Chih Yu, Xuejie Zhang
Summary: Sentiment embeddings can distinguish words with similar contexts but opposite sentiment, yet traditional methods may ignore the variation of word meaning in different contexts. Dealing with out-of-vocabulary or informal-writing sentiment words is a challenge. The proposed training model effectively addresses these issues and outperforms previous methods.
KNOWLEDGE-BASED SYSTEMS
(2022)
Article
Computer Science, Artificial Intelligence
Jun Kong, Jin Wang, Xuejie Zhang
Summary: In this paper, a hierarchical BERT model called HAdaBERT is proposed to address the limitation of applying pretrained language models to document classification. The model utilizes a local encoder and a global encoder to encode the documents, and employs an adaptive fine-tuning strategy to improve performance.
KNOWLEDGE-BASED SYSTEMS
(2022)
Article
Computer Science, Artificial Intelligence
Yanjun Qian, Jin Wang, Dawei Li, Xuejie Zhang
Summary: Existing sentiment analysis models rely on evident emotive words, but implicit emotion communication without explicit phrases is common in various cultures. To address the issue of losing syntactic information when emotional words are removed, this paper proposes an interactive capsule network approach. By utilizing interactive attention and capsule network with dynamic routing, the model can learn the insightful relationship between former and latter contexts, achieving better performance in sentiment analysis tasks.
APPLIED INTELLIGENCE
(2023)
Article
Biochemical Research Methods
Jin Wang, Liang-Chih Yu, Xuejie Zhang
Summary: Analysis of health-related texts is important for detecting adverse drug reactions (ADR). However, the imbalanced data distributions pose a challenge for ADR detection. This paper proposes a weighted variant of conditional random field (CRF) that can alleviate the data distribution imbalance and improve the performance of sequence labeling tasks.
PLOS COMPUTATIONAL BIOLOGY
(2022)
Article
Computer Science, Artificial Intelligence
Lung-Hao Lee, Jian-Hong Li, Liang-Chih Yu
Summary: This study builds a Chinese valence-arousal resource for sentiment analysis and evaluates it using different categories of classifiers. It provides more fine-grained sentiment analysis through annotation on multiple levels of text. The performance is compared with a similar evaluation of an English sentiment resource.
ACM TRANSACTIONS ON ASIAN AND LOW-RESOURCE LANGUAGE INFORMATION PROCESSING
(2022)
Article
Computer Science, Information Systems
Li Yuan, Jin Wang, Liang-Chih Yu, Xuejie Zhang
Summary: This paper proposes a hierarchical template-transformer model for personalized fine-grained sentiment controllable generation. The model utilizes a hierarchical structure and part of speech (POS) templates to generate aspect-level sentiment controllable reviews with personalized information.
INFORMATION PROCESSING & MANAGEMENT
(2022)
Article
Automation & Control Systems
Ruijun Chen, Jin Wang, Liang-Chih Yu, Xuejie Zhang
Summary: Artificial emotional intelligence (AEI) is crucial for intelligent systems, and representing emotional states as continuous sentiment intensity enables more precise sentiment application. However, existing variational autoencoder methods face challenges in distinguishing emotional information from semantic information. To address this, we propose a decoupled VAE with interactive attention that extracts and maps sentiment intensities into the latent space using sentiment decouplers and embeddings.
ENGINEERING APPLICATIONS OF ARTIFICIAL INTELLIGENCE
(2023)
Proceedings Paper
Computer Science, Artificial Intelligence
Mengjiao Xie, Jin Wang, Xuejie Zhang
Summary: This paper proposes a framework called TriLab for triple relation extraction, based on label embedding and using a joint extraction mode. Experimental results show that the TriLab framework outperformed previous methods in triple extraction.
2021 INTERNATIONAL CONFERENCE ON ASIAN LANGUAGE PROCESSING (IALP)
(2021)
Proceedings Paper
Computer Science, Artificial Intelligence
Fulai Nan, Jin Wang, Xuejie Zhang
Summary: Machine comprehension of text is important in natural language processing, but data imbalance in extractive MRC tasks can cause issues in training and testing. Introducing focal loss and mirror distillation can help address these problems and improve model performance.
2021 INTERNATIONAL CONFERENCE ON ASIAN LANGUAGE PROCESSING (IALP)
(2021)
Article
Computer Science, Information Systems
Lung-Hao Lee, Yuh-Shyang Wang, Chao-Yi Chen, Liang-Chih Yu
Summary: The study introduces an Ensemble Multi-Channel Neural Networks (EMC-NN) model for scientific language editing evaluation, which outperforms previous competition winners and the recent BERT transformers in improving the quality of scientific writing through a combination of multi-channel word embedding representation, deep learning models, and majority voting ensemble.