Article
Computer Science, Information Systems
Ngan Tran, Haihua Chen, Jay Bhuyan, Junhua Ding
Summary: Intrusion detection plays a crucial role in protecting the cyber environment. However, the quality of data used for intrusion detection systems significantly affects their performance. This study shows that mislabeled, inaccurate, duplicated, and overlapped data can lead to poor results. Experimental results reveal that pre-trained models outperform other models, and removing duplicates and overlaps improves their performance.
Article
Biochemical Research Methods
Leqi Tian, Tianwei Yu
Summary: Untargeted metabolomics has wide applications, but the problem of matching uncertainty between data features and known metabolites is not well-addressed. The current approach is to match the mass-to-charge ratio (m/z) of data features to theoretical values of known metabolites. However, the matching uncertainty caused by metabolites with the same molecular composition and the generation of various adduct ions affects the reliability of the results. In this study, an integrated deep learning framework is proposed to consider the matching uncertainty, which can simultaneously evaluate metabolite importance, infer feature-metabolite matching likelihood, and select disease sub-networks.
BRIEFINGS IN BIOINFORMATICS
(2023)
Article
Computer Science, Artificial Intelligence
Adriel Saporta, Xiaotong Gui, Ashwin Agrawal, Anuj Pareek, Steven Q. H. Truong, Chanh D. T. Nguyen, Van-Doan Ngo, Jayne Seekins, Francis G. Blankenberg, Andrew Y. Ng, Matthew P. Lungren, Pranav Rajpurkar
Summary: In this study, seven saliency methods for medical image analyses are evaluated and compared using different neural network architectures. The results show that all seven methods perform worse than a human radiologist benchmark in terms of localizing pathologies. The performance gap between Grad-CAM and the human benchmark is largest for smaller and more complex pathologies, and model confidence is positively correlated with Grad-CAM localization performance. This work highlights the limitations of saliency methods and the need for further improvement in deep learning explainability for medical imaging.
NATURE MACHINE INTELLIGENCE
(2022)
Article
Biology
A. M. Fischer, A. L. Rietveld, P. W. Teunissen, P. C. A. M. Bakker, M. Hoogendoorn
Summary: By analyzing the electrical activity of the uterus, which can lead to preterm birth, a machine learning model has been developed to detect and predict preterm birth. The model achieved comparable results to other machine learning models, and the addition of clinical data did not significantly improve performance.
COMPUTERS IN BIOLOGY AND MEDICINE
(2023)
Article
Engineering, Electrical & Electronic
Xian Sun, Bing Wang, Zhirui Wang, Hao Li, Hengchao Li, Kun Fu
Summary: This article provides a bibliometric analysis of existing works on remote sensing interpretation related to few-shot learning, introduces two categories of few-shot learning methods, and lists three typical remote sensing interpretation applications with corresponding datasets and evaluation criteria. It summarizes the research status and suggests possible research directions for scholars in the field of remote sensing and few-shot learning.
IEEE JOURNAL OF SELECTED TOPICS IN APPLIED EARTH OBSERVATIONS AND REMOTE SENSING
(2021)
Article
Environmental Sciences
Yilin Wang, Dongxu Yin, Liming Lou, Xinying Li, Pengle Cheng, Ying Huang
Summary: Exposed mine gangue hills can cause environmental problems, which can be effectively solved by vegetation restoration. Through the use of UAV aerial photography and deep learning methods, efficient ground cover monitoring and vegetation restoration can be achieved.
Article
Geochemistry & Geophysics
Weihua Wu, Yang Yang, Bangyu Wu, Debo Ma, Zhanxin Tang, Xia Yin
Summary: Seismic fault interpretation plays a crucial role in hydrocarbon reservoir characterization and drilling hazard mitigation. This article proposes a multitask deep learning-based seismic fault detection method called MTL-FaultNet, which incorporates 3-D seismic data reconstruction as an auxiliary task to improve the fault detection performance. The network is equipped with multiscale modules and embedded attention mechanisms to enhance its ability in focusing on fault features and learning stable fault structures. Experimental results on field seismic data demonstrate the reliability and generalization capability of the proposed method.
IEEE TRANSACTIONS ON GEOSCIENCE AND REMOTE SENSING
(2023)
Article
Engineering, Multidisciplinary
Kanan Wahengbam, Kshetrimayum Linthoinganbi Devi, Aheibam Dinamani Singh
Summary: Brain-Media is a discipline that decodes sophisticated human brain activity, such as imagination, memories, colors, textures, patterns, etc. Existing efforts either classify brain signals or map them to an image of the same class, but they ignore the existence of disruptive noises. This research proposes a multimodality time-series and spatial-domain hybrid framework and a unique ResilientNet Generator to robustly classify signals.
IEEE TRANSACTIONS ON NETWORK SCIENCE AND ENGINEERING
(2023)
Article
Geosciences, Multidisciplinary
Ming Sun, Lin Chen, Tim Li, Jing-Jia Luo
Summary: This study developed a deep learning model based on the convolutional neural network to accurately predict the zonal pattern of sea surface temperature anomalies over the equatorial Pacific. The model outperforms current dynamical models in predicting the SSTA zonal pattern 1 year in advance. The physical interpretation shows that the sources of ENSO predictability at different lead times are distinct.
GEOPHYSICAL RESEARCH LETTERS
(2023)
Review
Engineering, Biomedical
Daniel T. Huff, Amy J. Weisman, Robert Jeraj
Summary: This review provides insights into methods for interpreting image models in medical imaging, categorizing them into understanding model structure and function, and understanding model output. It discusses various interpretation techniques, limitations of current methods, and offers recommendations for DL practitioners looking to incorporate model interpretation into their tasks in medical imaging contexts.
PHYSICS IN MEDICINE AND BIOLOGY
(2021)
Article
Radiology, Nuclear Medicine & Medical Imaging
Panagiotis Papadimitroulas, Lennart Brocki, Neo Christopher Chung, Wistan Marchadour, Franck Vermet, Laurent Gaubert, Vasilis Eleftheriadis, Dimitris Plachouris, Dimitris Visvikis, George C. Kagadis, Mathieu Hatt
Summary: Radiomics and Deep Neural Networks (DNNs) have shown promising applications in medical image analysis for clinical practices in terms of classification and prediction. However, most current research is limited by dataset availability and generalizability. The study stresses the importance of multicenter recruitment of large datasets to enhance biomarker variability, establish the potential clinical value of radiomics, and develop robust explainable AI models.
PHYSICA MEDICA-EUROPEAN JOURNAL OF MEDICAL PHYSICS
(2021)
Article
Geochemistry & Geophysics
Dekuan Chang, Guangzhi Zhang, Xueshan Yong, Jianhu Gao, Yihui Wang, Wei Wang
Summary: This article proposes a Fourier domain adaptation (FDA) method to transfer synthesized seismic data to real seismic data, improving the performance of network models trained on synthetic data. By replacing the amplitude spectrum of synthesized seismic data with that of real seismic data, the FDA achieves domain transfer and enhances the generalization ability of network models in seismic structure interpretation.
IEEE GEOSCIENCE AND REMOTE SENSING LETTERS
(2022)
Article
Geochemistry & Geophysics
Wei Han, Jun Li, Sheng Wang, Xinyu Zhang, Yusen Dong, Runyu Fan, Xiaohan Zhang, Lizhe Wang
Summary: Geological remote sensing interpretation is crucial in geological survey and mapping, but machine learning-based methods often yield inferior results. In this study, we propose a deep learning feature-based adaptive multisource data fusion network, which efficiently interprets multiple geological remote sensing elements. Experimental results demonstrate the superiority of our model.
IEEE TRANSACTIONS ON GEOSCIENCE AND REMOTE SENSING
(2022)
Article
Ecology
Endre Moen, Rune Vabo, Szymon Smolinski, Come Denechaud, Nils Olav Handegard, Ketil Malde
Summary: This study developed a machine learning framework for fish age prediction using images of otoliths. The models based on convolutional neural networks achieved an average accuracy of 72.7% and performed well in predicting the age of one- and two-year-old individuals. The best models were EfficientNet B4 and EfficientNet B6 using images taken with low exposure times.
ECOLOGICAL INFORMATICS
(2023)
Article
Computer Science, Interdisciplinary Applications
Gert-Jan Both, Subham Choudhury, Pierre Sens, Remy Kusters
Summary: DeepMoD is a deep learning based model discovery algorithm that uses sparse regression to discover the underlying partial differential equation in a library of possible functions and their derivatives. It is robust to noise, applicable to small data sets, and does not require a training set. The algorithm has shown promising results in benchmark tests on physical problems and successfully discovered the advection-diffusion equation in noisy experimental time-series data.
JOURNAL OF COMPUTATIONAL PHYSICS
(2021)
Article
Engineering, Electrical & Electronic
Jian Zhou, Lianyu Zheng, Yiwei Wang, Cheng Wang, Robert X. Gao
Summary: This article presents an automatic modeling framework based on reinforcement learning and neural architecture search for fault diagnosis in machinery. The results show that the method successfully searches high-accuracy diagnostic models within a short time.
IEEE TRANSACTIONS ON INSTRUMENTATION AND MEASUREMENT
(2022)
Article
Engineering, Industrial
Yilin Li, Jinjiang Wang, Zuguang Huang, Robert X. Gao
Summary: This paper introduces a new physics-informed meta-learning framework for tool wear prediction under varying wear rates, improving prediction accuracy by enhancing modeling strategy and constraining optimization process with a loss term informed by physics.
JOURNAL OF MANUFACTURING SYSTEMS
(2022)
Review
Engineering, Industrial
Shenghan Guo, Mohit Agarwal, Clayton Cooper, Qi Tian, Robert X. Gao, Weihong Guo Grace, Y. B. Guo
Summary: Machine learning has proven to be an effective alternative to physical models in quality prediction and process optimization of metal additive manufacturing. However, the interpretability of machine learning outcomes within the complex thermodynamics of additive manufacturing has been a challenge. Physics-informed machine learning (PIML) addresses this challenge by integrating data-driven methods with physical domain knowledge.
JOURNAL OF MANUFACTURING SYSTEMS
(2022)
Article
Engineering, Multidisciplinary
Dimitris K. Iakovidis, Melanie Ooi, Ye Chow Kuang, Serge Demidenko, Alexandr Shestakov, Vladimir Sinitsin, Manus Henry, Andrea Sciacchitano, Stefano Discetti, Silvano Donati, Michele Norgia, Andreas Menychtas, Ilias Maglogiannis, Selina C. Wriessnegger, Luis Alberto Barradas Chacon, George Dimas, Dimitris Filos, Anthony H. Aletras, Johannes Toger, Feng Dong, Shangjie Ren, Andreas Uhl, Jacek Paziewski, Jianghui Geng, Francesco Fioranelli, Ram M. Narayanan, Carlos Fernandez, Christoph Stiller, Konstantina Malamousi, Spyros Kamnis, Konstantinos Delibasis, Dong Wang, Jianjing Zhang, Robert X. Gao
Summary: Signal processing plays a crucial role in sensor-enabled systems and has various applications. The advancement in artificial intelligence and machine learning has shifted research focus towards intelligent, data-driven signal processing. This roadmap provides a critical overview of current methods and applications, aiming to identify future challenges and research opportunities for next generation measurement systems.
MEASUREMENT SCIENCE AND TECHNOLOGY
(2022)
Article
Engineering, Industrial
Peng Wang, Joseph Kershaw, Matthew Russell, Jianjing Zhang, Yuming Zhang, Robert X. Gao
Summary: This paper presents a data-driven process characterization and online adaptive control framework for robotic arc welding to automatically achieve desired weld pool condition. Pool width is characterized through a pixel-level image segmentation network and used for determining parameter adjustment for robotic execution.
CIRP ANNALS-MANUFACTURING TECHNOLOGY
(2022)
Editorial Material
Engineering, Industrial
Ihab Ragai, Robert X. Gao, Livan Fratini
JOURNAL OF MANUFACTURING SYSTEMS
(2022)
Article
Engineering, Industrial
Jinjiang Wang, Yilin Li, Robert X. Gao, Fengli Zhang
Summary: This paper reviews the latest development of hybrid physics-based data-driven models in smart manufacturing and summarizes the principles and characteristics of three types of these models. It discusses the application of these models in product design, operation and maintenance, and intelligent decision-making.
JOURNAL OF MANUFACTURING SYSTEMS
(2022)
Editorial Material
Engineering, Industrial
Kunpeng Zhu, Yongjie Jessica Zhang, Robert Gao, Markus Bambach, Erman Tekkaya
JOURNAL OF MATERIALS PROCESSING TECHNOLOGY
(2022)
Article
Engineering, Mechanical
Jianjing Zhang, Chuanping Liu, Robert X. Gao
Summary: This study introduces a physics-guided Gaussian process (PGGP) that combines physical knowledge with data learning to address the limitations of data-driven methods that require a large amount of training data and the ability to capture system behavior.
MECHANICAL SYSTEMS AND SIGNAL PROCESSING
(2022)
Article
Automation & Control Systems
Jinjiang Wang, Peilun Fu, Shuaihang Ji, Yilin Li, Robert X. Gao
Summary: This article proposes a light weight multisensory fusion model for induction motor data fusion and diagnosis. By introducing inverted residual block and network architecture search technology, the training speed and prediction speed of the diagnostic model are accelerated, and fault patterns can be accurately judged in a shorter prediction time.
IEEE-ASME TRANSACTIONS ON MECHATRONICS
(2022)
Article
Engineering, Industrial
Clayton Cooper, Jianjing Zhang, Joshua Huang, Jennifer Bennett, Jian Cao, Robert X. Gao
Summary: This research explores a data-driven predictive model with Shapley additive explanation (SHAP) to improve the accuracy and interpretability of predicting mechanical properties in directed energy deposition (DED) processes. The results demonstrate that by interpreting input features and reducing model complexity, both the accuracy and effectiveness of predictive models for Inconel 718 (IN718) tensile strength can be improved.
JOURNAL OF MATERIALS PROCESSING TECHNOLOGY
(2023)
Article
Engineering, Industrial
Clayton Cooper, Jianjing Zhang, Robert X. Gao
Summary: This study presents an efficient and error-aware PINN ensembling technique for error homogenization in solving manufacturing problems. The method constrains the training process of PINN to ensure physical consistency and combines the outputs of multiple PINNs to improve the accuracy of PDE solutions. Experimental results demonstrate the effectiveness of the developed method in reducing PINN errors and improving consistency.
JOURNAL OF MANUFACTURING SYSTEMS
(2023)
Article
Engineering, Electrical & Electronic
Clayton Cooper, Jianjing Zhang, Liwen Hu, Yuebin Guo, Robert X. Gao
Summary: This paper introduces a ridgelet transform-based method for machined surface characterization, which improves the accuracy of surface characterization by predicting and quantifying the uncertainty of surface roughness using texture-aware features and machine learning algorithms.
IEEE TRANSACTIONS ON INSTRUMENTATION AND MEASUREMENT
(2022)
Article
Engineering, Electrical & Electronic
Jinjiang Wang, Jiazheng Sun, Weifeng Ge, Fengli Zhang, Robert X. Gao
Summary: This article proposes a new virtual sensing method for online fault diagnosis of heat exchangers, which quantifies fouling thickness and tube leakage in real-time by incorporating equipment failure mechanism and inference analytics with high-precision in-process data measurement.
IEEE TRANSACTIONS ON INSTRUMENTATION AND MEASUREMENT
(2022)