Article
Nanoscience & Nanotechnology
Nan Wang, Yongnan Chen, Gang Wu, Qinyang Zhao, Zhen Zhang, Lixia Zhu, Jinheng Luo
Summary: This study reveals the different contributions of geometrically necessary dislocation (GND) and statistically stored dislocation (SSD) to work hardening in dual-phase steel. By introducing high-density GND through pre-tensile loading-unloading-reloading (LUR) and high-density SSD through monotonic pre-tensile, it is found that the steel with high GND exhibits higher yield stress and stronger strain hardening ability compared to the high-SSD steel, even with almost the same total dislocation density.
MATERIALS SCIENCE AND ENGINEERING A-STRUCTURAL MATERIALS PROPERTIES MICROSTRUCTURE AND PROCESSING
(2022)
Article
Materials Science, Multidisciplinary
Mohammed Mendas, Stephane Benayoun, Mohamed Hadj Miloud, Ibrahim Zidane
Summary: This study extends the analysis of the indentation size effect (ISE) to lamellar cast irons, demonstrating that the tensile model and the concept of geometrically necessary dislocations (GNDs) can be used to predict the ISE of the pearlitic matrix within these materials. The summation of stresses associated with GNDs and statistically stored dislocations (SSDs) is shown to be more adequate in the prediction of ISE compared to considering only one work-hardening stress.
JOURNAL OF MATERIALS RESEARCH AND TECHNOLOGY-JMR&T
(2021)
Article
Computer Science, Interdisciplinary Applications
Vishnu Preetham Revelli, Gauri Sharma, S. Kiruthika Devi
Summary: This project aims to extract text from braille text images and provide translated English text and audio output using a customized CNN model. The CNN model demonstrates robustness in image recognition and classification tasks, making it valuable for addressing challenges faced by visually impaired individuals.
ADVANCES IN ENGINEERING SOFTWARE
(2022)
Article
Computer Science, Artificial Intelligence
Sergey N. Pozdnyakov, Michele Ceriotti
Summary: Graph neural networks (GNN) are popular in machine learning and have been successful in predicting properties of molecules and materials. However, first-order GNNs are known to be incomplete, leading to the design of more complex schemes. The construction of graph representations for molecules adds a geometric dimension, with the most common approach being to consider atoms as vertices and connect them with bonds. This approach, known as distance graph NNs (dGNN), has shown excellent resolving power in chemical ML. However, the authors present a counterexample that proves dGNNs are not complete even for fully-connected graphs induced by 3D atom clouds.
MACHINE LEARNING-SCIENCE AND TECHNOLOGY
(2022)
Article
Computer Science, Artificial Intelligence
Nemanja Milosevic, Milos Rackovic
Summary: Convolutional neural networks have become indispensable in various machine learning applications, particularly in image classification, but research on their robustness and susceptibility to adversarial attacks is crucial. A new classification method based on missing features has been proposed, showing improved robustness compared to traditional models, although the enhancement in validation accuracy may come at the cost of losing important knowledge. Proposed solutions are being validated against the CIFAR-10 image classification dataset.
NEURAL COMPUTING & APPLICATIONS
(2021)
Article
Engineering, Marine
K. Janas, I. A. Milne, J. R. Whelan
Summary: This study demonstrates a novel application of CNN for identifying mooring line failure of turret-moored FPSO. The CNN successfully distinguished turret responses associated with intact and broken mooring, with classification accuracy improved through additional hidden layers and retraining, particularly in minimal offset response conditions. The CNN offers an effective and lower-cost alternative to existing mooring failure detection approaches for the offshore industry.
Article
Biochemistry & Molecular Biology
Heiko Dunkel, Henning Wehrmann, Lars R. Jensen, Andreas W. Kuss, Stefan Simm
Summary: Non-coding RNA (ncRNA) classes play important roles in cell regulation and identification of diagnostic and therapeutic biomarkers. Researchers utilized machine learning models, including different neural network architectures, to improve the classification and prediction of ncRNAs.
INTERNATIONAL JOURNAL OF MOLECULAR SCIENCES
(2023)
Article
Agronomy
Astrid Tempelaere, Bart De Ketelaere, Jiaqi He, Ioannis Kalfas, Michiel Pieters, Wouter Saeys, Remi Van Belleghem, Leen Van Doorselaer, Pieter Verboven, Bart M. Nicolai
Summary: This article introduces recent developments in artificial intelligence for extracting information on postharvest disorders from complex image data obtained by modern imaging systems. Machine vision inspection using RGB imaging and advanced techniques such as spectral cameras, X-ray, and MRI is increasingly used in the postharvest industry to nondestructively analyze disorders in horticultural products. However, challenges in the design of deep learning models, such as the need for large quantities of labeled data, model explainability, and generalizability, need to be addressed.
POSTHARVEST BIOLOGY AND TECHNOLOGY
(2023)
Review
Chemistry, Analytical
Bruno Debus, Hadi Parastar, Peter Harrington, Dmitry Kirsanov
Summary: Recent extensive research in Deep Learning has led to the development of machine learning algorithms dedicated to solving complex tasks, drawing attention from various fields including analytical chemistry. These powerful algorithms can extract qualitative and quantitative information from complex chemical measurements.
TRAC-TRENDS IN ANALYTICAL CHEMISTRY
(2021)
Article
Mathematical & Computational Biology
Donald E. Brown, Suchetha Sharma, James A. Jablonski, Arthur Weltman
Summary: This study investigates the use of neural network techniques to predict patient health conditions with CPET data and finds that these techniques provide higher levels of accuracy compared to traditional flowchart methods.
Article
Computer Science, Information Systems
Manuel Torres, Rafael Alvarez, Miguel Cazorla
Summary: Cybercriminals constantly develop new techniques to evade security measures, resulting in rapid evolution of malware. Detecting malware across multiple systems is challenging due to unique characteristics of each computing environment. Traditional signature-based malware detection has been replaced by modern approaches, such as machine learning and behavior-based threat detection. Researchers use these techniques to extract features from various data sources and feed them to models for accurate prediction.
Article
Computer Science, Artificial Intelligence
Sanjeev Kumar, Kajal Panda
Summary: This paper proposes a novel malware detection and classification architecture based on image visualization using fine-tuned convolutional neural networks. The methodology involves using a pre-trained VGG16 model as a feature extractor and different feature selection methods to construct a feature map. The MLP classifier achieves the best accuracy in detecting malware.
APPLIED SOFT COMPUTING
(2023)
Article
Computer Science, Information Systems
Dawid Grzelak, Krzysztof Podlaski, Grzegorz Wiatrowski
Summary: An approach using deep learning technique to generalize OCR task for recognizing Polish letters is presented. By extending the dataset with Polish diacritics A and C, the study shows that convolutional neural network can properly recognize shadows and noises of Polish characters and distinguish between similar letters A and Ą. A neural network trained without Polish characters, however, fails to treat letters A and Ą correctly.
JOURNAL OF KING SAUD UNIVERSITY-COMPUTER AND INFORMATION SCIENCES
(2021)
Article
Computer Science, Interdisciplinary Applications
Nicolas Bertin, Fei Zhou
Summary: Discrete dislocation dynamics (DDD) is a computational method used to study plasticity by connecting the motion of dislocation lines to the macroscopic response of crystalline materials. The computational cost of DDD simulations has been a limitation, but a new DDDGNN framework uses a graph neural network (GNN) model to replace the expensive time-integration of dislocation motion, showing stability and accuracy in reproducing DDD simulation responses. This approach opens opportunities to accelerate DDD simulations and incorporate complex dislocation motion behaviors.
JOURNAL OF COMPUTATIONAL PHYSICS
(2023)
Article
Computer Science, Artificial Intelligence
Felipe Buitrago, Luis Fernando Castillo Ossa, Jeferson Arango-Lopez
Summary: Surface defects in industrial refrigerator manufacturing processes can cause production losses and compromise product quality. Visual quality inspection is currently a subjective process that requires expert intervention, limiting efficiency and leading to errors. This paper proposes a novel approach using CNN and deflectometry for automatic surface defect detection, which shows promise for quality control in refrigerator manufacturing and other industries. The method combines the accuracy of CNNs in image classification and the sensitivity of deflectometry to detect subtle surface variations.