Article
Computer Science, Artificial Intelligence
Koji Hashimoto, Hong-Ye Hu, Yi-Zhuang You
Summary: The neural ordinary differential equation (neural ODE) is a novel machine learning architecture applied to holographic QCD, where it is able to train consistent bulk geometry and automatically discover emergent black hole horizons. This model provides a trustworthy approach to calculating holographic Wilson loops with consistent temperature dependence.
MACHINE LEARNING-SCIENCE AND TECHNOLOGY
(2021)
Article
Physics, Multidisciplinary
Carlo Manzo
Summary: This article introduces a simple approach - AnDi-ELM, which combines extreme learning machine and feature engineering to tackle the tasks of the AnDi challenge. The method offers satisfactory performance, simple implementation, and fast training time.
JOURNAL OF PHYSICS A-MATHEMATICAL AND THEORETICAL
(2021)
Article
Automation & Control Systems
Javier Vales-Alonso, Francisco Javier Gonzalez-Castano, Pablo Lopez-Matencio, Felipe Gil-Castineira
Summary: This study addresses the challenge of measuring the unpredictability of boxers during training sessions by combining novel techniques and using unlabeled data. The system achieves high accuracy and has the potential to improve sports training and evaluation.
IEEE TRANSACTIONS ON SYSTEMS MAN CYBERNETICS-SYSTEMS
(2023)
Article
Automation & Control Systems
Qiuyang Zhao, Yu Zhao, Lijun Bao
Summary: In this article, a primary-auxiliary coupled neural network (PANet) is proposed for 3-D holographic particle field characterization, which achieves excellent performance in solving imaging artifact, noise, and blur.
IEEE TRANSACTIONS ON INDUSTRIAL INFORMATICS
(2022)
Review
Optics
Andrey Romanov, Maxim A. Yurkin
Summary: The field of light-scattering characterization of single particles has grown rapidly in the past 30 years due to advancements in measurement and simulation capabilities. Various methods have been developed to characterize particles with high geometric resolution, but development has been fragmented and specific to experimental setups. Existing methods are categorized into model-driven, model-free, and data-driven methods, with a focus on algorithms and experimental aspects.
LASER & PHOTONICS REVIEWS
(2021)
Article
Chemistry, Analytical
Aime Lay-Ekuakille, John Djungha Okitadiowo, Moise Avoci Ugwiri, Sabino Maggi, Rita Masciale, Giuseppe Passarella
Summary: The article discusses the importance of monitoring the flow of water in open channels to prevent floods, and proposes the use of video imaging technology and particle tracking algorithms to improve accuracy.
Article
Physics, Applied
Bin Liu, Liujun Xu, Jiping Huang
Summary: A new mechanism for engineering and manipulating thermal metamaterials is proposed in this work, relaxing previous constraints and using a machine learning-based approach to solve the inverse design problem. The neural network is trained and validated using the finite-element method, showing potential for application in thermal transparency and other related problems.
JOURNAL OF APPLIED PHYSICS
(2021)
Article
Engineering, Electrical & Electronic
Eugeny Chubchev, Kirill Tomyshev, Igor Nechepurenko, Alexander Dorofeenko, Oleg Butov
Summary: In this study, machine learning methods were employed for the first time to process spectral data of a plasmonic fiber sensor based on a tilted fiber Bragg grating, achieving a high resolution.
JOURNAL OF LIGHTWAVE TECHNOLOGY
(2022)
Article
Mechanics
Rongrong Xie, Matteo Marsili
Summary: In this study, a generic ensemble of deep belief networks (DBN) is explored, with the distribution of energy levels of hidden states serving as the parameter. It is demonstrated that statistical dependence can only propagate from the visible to deep layers if each layer is tuned close to the critical point during learning. Accordingly, efficiently trained learning machines exhibit a broad distribution of energy levels. The analysis of DBNs and restricted Boltzmann machines on different datasets supports these findings.
JOURNAL OF STATISTICAL MECHANICS-THEORY AND EXPERIMENT
(2022)
Article
Optics
Mathias Geisler, Jacob Larsen, Kai Dirscherl, Soren Alkaersig Jensen
Summary: This study employs a neural network approach to classify aerosolized spherical particles based on both constituent material and size. The neural network shows good accuracy in sizing and classification when tested on the same particle types it was trained on. Additional wavelengths and detectors can significantly improve the accuracy of size and classification, achieving an overall accuracy of > 95%. However, materials with optical properties far from the training data or near the boundaries of chosen categories are more prone to misclassification.
JOURNAL OF QUANTITATIVE SPECTROSCOPY & RADIATIVE TRANSFER
(2022)
Article
Physics, Applied
Taegeun Song, Nojoon Myoung, Hunpyo Lee, Hee Chul Park
Summary: This study proposes a method to recognize nanobubbles in graphene by analyzing electronic properties based on a machine learning approach. The machine learning algorithm efficiently classifies the density of states spectra by the height and width of the nanobubbles, even in cases with a substantial magnitude of noise. The machine-learning-based analysis of electronic properties may introduce a changeover in the probing of nanobubbles from image-based detection to electrical-measurement-based recognition.
APPLIED PHYSICS LETTERS
(2021)
Article
Computer Science, Artificial Intelligence
M. Abedi, M. Z. Naser
Summary: The study introduces a rapid, automated, and intelligent approach that leverages machine learning to identify vulnerable bridges to fire hazard, backed by a comprehensive database comprising actual observations. This method can assist engineers and government officials in swiftly assessing fire-vulnerable bridges with a high accuracy of 89.6%.
APPLIED SOFT COMPUTING
(2021)
Article
Soil Science
Felipe B. de Santana, Rebecca. L. Hall, Victoria Lowe, Margaret A. Browne, Eric C. Grunsky, Mairead M. Fitzsimons, Vincent Gallagher, Karen Daly
Summary: Soil spectroscopy has been used to predict soil particle size in an area of approximately 35,716 km² in Ireland, and a systematic approach using MIR spectroscopy and chemometrics was developed. Spectral control charts were used to identify abnormal spectra, and the accuracy of predicted values was assessed using a validation set. This approach enables the creation of regional and national scale soil maps with confidence.
Article
Chemistry, Analytical
Qiannan Duan, Zhaoyi Xu, Shourong Zheng, Jiayuan Chen, Yunjin Feng, Luo Run, Jianchao Lee
Summary: This study successfully detected mixed pollutants using holographic spectrum and convolutional neural network, demonstrating the potential of machine learning applications in the field of chemistry.
ANALYTICA CHIMICA ACTA
(2021)
Review
Chemistry, Physical
Xinhua Liu, Lisheng Zhang, Hanqing Yu, Jianan Wang, Junfu Li, Kai Yang, Yunlong Zhao, Huizhi Wang, Billy Wu, Nigel P. Brandon, Shichun Yang
Summary: This study demonstrates a method to evaluate the overall lifecycle of lithium-ion batteries (LIBs) and discusses the bridging role of characterization techniques and modeling. Key parameters extracted from characterization can be used as digital inputs for modeling. Furthermore, advanced computational techniques can enhance the understanding and control of the battery lifecycle. The introduction of digital twins techniques enables real-time monitoring and control, as well as intelligent manufacturing.
ADVANCED ENERGY MATERIALS
(2022)
