Article
Computer Science, Artificial Intelligence
Yufei Cui, Yu Mao, Ziquan Liu, Qiao Li, Antoni B. Chan, Xue Liu, Tei-Wei Kuo, Chun Jason Xue
Summary: Nested dropout is a variant of dropout operation that orders network parameters or features based on pre-defined importance. It has been explored in constructing nested nets and learning ordered representation.
IEEE TRANSACTIONS ON PATTERN ANALYSIS AND MACHINE INTELLIGENCE
(2023)
Article
Biology
Sacha Sokoloski, Amir Aschner, Ruben Coen-Cagli
Summary: This paper proposes a response model based on mixture models and exponential families, which can capture the variability and covariability in large-scale neural recordings. Additionally, the model facilitates accurate Bayesian decoding, provides a closed-form expression for the Fisher information, and is compatible with theories of probabilistic population coding.
Article
Engineering, Aerospace
Zuwei Tan, Runze Li, Yufei Zhang
Summary: The inlet is a crucial component of hypersonic vehicles, and the design and optimization of the hypersonic inlet are of great importance. Artificial intelligence techniques, such as variational autoencoder (VAE) and generative adversarial network (GAN), have been used to improve aerodynamic optimization efficiency. This study applies a hybrid multilayer perceptron (MLP) combined with a VAE network to reconstruct and predict the flow field of a two-dimensional multiwedge hypersonic inlet. The results demonstrate that the VAE network accurately reconstructs the overall flow structure of the hypersonic flow field and achieves satisfactory reconstruction accuracy for complex flow structures.
Article
Engineering, Electrical & Electronic
Markus Loschenbrand
Summary: This paper successfully applies a Variational Autoencoder model to probabilistic optimal power flows, showing simplicity and accuracy. It also suggests potential pathways for future research and offers insights for practitioners using similar generative models.
ELECTRIC POWER SYSTEMS RESEARCH
(2021)
Article
Computer Science, Artificial Intelligence
Teemu Sahlstrom, Tanja Tarvainen
Summary: There is a growing interest in using machine learning methods for inverse problems and imaging. However, most of the research has focused on image reconstruction problems, and there are limited studies on the complete solution of the inverse problem. In this study, we explore a machine learning-based approach for the Bayesian inverse problem of photoacoustic tomography. We propose a method for estimating the posterior distribution in photoacoustic tomography using a variational autoencoder, and evaluate it through numerical simulations and comparison with a Bayesian approach for solving the inverse problem.
SIAM JOURNAL ON IMAGING SCIENCES
(2023)
Article
Computer Science, Artificial Intelligence
Shuiqiao Yang, Sunny Verma, Borui Cai, Jiaojiao Jiang, Kun Yu, Fang Chen, Shui Yu
Summary: Recent developments in attributed network clustering have combined graph neural networks and autoencoders for unsupervised learning. However, these techniques suffer from either clustering-unfriendly embedding spaces or limited utilization of attribute information. To address these issues, the proposed model, VCLANC, utilizes deeper information by reconstructing the network structure and node attributes for self-supervised learning. Experimental results on four real-world datasets demonstrate the outstanding performance of VCLANC for attributed network clustering.
KNOWLEDGE-BASED SYSTEMS
(2023)
Article
Chemistry, Multidisciplinary
Shuang Gao, Angpeng Liu, Yuxin Jiang, Yi Liu
Summary: The article focuses on improving combustion efficiency and economic efficiency by measuring and controlling the oxygen content through a soft measurement technique based on flame images. By developing a new generative-based regression model, high-quality flame images can be generated, leading to more accurate estimation results of the oxygen content compared to other methods.
Article
Materials Science, Multidisciplinary
Yinghan Zhao, Patrick Altschuh, Jay Santoki, Lars Griem, Giovanna Tosato, Michael Selzer, Arnd Koeppe, Britta Nestler
Summary: Porous membranes are widely used and understanding their microstructures is crucial for improving their performance. A promising method for quantitative analysis is to generate porous structures at the pore scale and validate them against experimental microstructures, then establish process-structure-property relationships using data-driven algorithms. This study uses a Variational AutoEncoder (VAE) neural network model to characterize the 3D structural information of porous materials and solve the inverse problem of process-structure linkage. Our methods provide a robust and unsupervised way to learn structural descriptors for porous microstructures.
Article
Computer Science, Information Systems
Hyeong Geun Lee, Jee Sic Hur, Yeo Chan Yoon, Soo Kyun Kim
Summary: Traditional 3D facial reconstruction methods use PCA-based 3DMMs, while restoration methods based on GCN have advantages in directly regressing vertex coordinates and colors. This study presents a face restoration approach that improves stability and accuracy by directly regressing vertex coordinates and colors from 2D facial images.
Article
Engineering, Multidisciplinary
Vahid Keshavarzzadeh, Robert M. Kirby, Akil Narayan
Summary: Inverse problems are common in engineering simulations, and Bayesian inference is a predominant approach to infer unknown parameters. This paper presents a variational inference method that incorporates observation data and the gradient information of the forward map to invert unknown latent parameters. The method utilizes a trained neural network to generate samples for statistical calculations. The effectiveness of the method is demonstrated through examples, and future research directions are discussed.
COMPUTER METHODS IN APPLIED MECHANICS AND ENGINEERING
(2022)
Article
Computer Science, Information Systems
Kostas Touloupas, Paul Peter Sotiriadis
Summary: This work proposes a deep representation learning method to build continuous-valued representations of individual integrated circuit devices, which are used to solve continuous sizing problems and demonstrate efficiency in real-world applications.
Article
Computer Science, Artificial Intelligence
Ali Aldhubri, Yu Lasheng, Farida Mohsen, Majjed Al-Qatf
Summary: Probabilistic matrix factorization (PMF) is a popular method for addressing the sparsity problem in collaborative filtering, but is vulnerable to overfitting. Bayesian PMF suggests an alternative approach, but comes with a high computational cost.
APPLIED INTELLIGENCE
(2021)
Article
Engineering, Biomedical
Kai Qiao, Jian Chen, Linyuan Wang, Chi Zhang, Li Tong, Bin Yan
Summary: This study proposes an alternating encoding and decoding method to improve the quality of reconstructing natural images from fMRI. By utilizing shared semi-supervised learning, we achieve mutual promotion between fMRI voxels and images, leading to better reconstruction results.
BIOMEDICAL SIGNAL PROCESSING AND CONTROL
(2022)
Article
Computer Science, Artificial Intelligence
Benjamin Gutierrez-Becker, Ignacio Sarasua, Christian Wachinger
Summary: The article introduces the use of deep neural networks for analyzing anatomical shapes, focusing on learning a low-dimensional shape representation specific to the task at hand rather than relying on manually engineered representations. The framework proposed is modular, with several computing blocks performing fundamental shape processing tasks on unordered point clouds, providing invariance to similarity transformations.
MEDICAL IMAGE ANALYSIS
(2021)
Article
Materials Science, Multidisciplinary
Yiping Sun, Zhaoyu Li, Jiadui Chen, Xuefeng Zhao, Meng Tao
Summary: This paper investigates a VAE model-based topology optimization method for optimizing the cavity structure of anechoic coatings. The finite-element method is used to calculate the sound absorption coefficient, and the VAE model is trained to learn the key features of an anechoic coating. The method efficiently generates new anechoic coatings with specific sound absorption properties.
MATERIALS TODAY COMMUNICATIONS
(2022)
Article
Neurosciences
Junxing Shi, Haiguang Wen, Yizhen Zhang, Kuan Han, Zhongming Liu
HUMAN BRAIN MAPPING
(2018)
Article
Neurosciences
Lauren K. Lynch, Kun-Han Lu, Haiguang Wen, Yizhen Zhang, Andrew J. Saykin, Zhongming Liu
HUMAN BRAIN MAPPING
(2018)
Article
Multidisciplinary Sciences
Yizhen Zhang, Gang Chen, Haiguang Wen, Kun-Han Lu, Zhongming Liu
SCIENTIFIC REPORTS
(2017)
Article
Multidisciplinary Sciences
Yizhen Zhang, Kuan Han, Robert Worth, Zhongming Liu
NATURE COMMUNICATIONS
(2020)
Article
Neurosciences
Jung-Hoon Kim, Yizhen Zhang, Kuan Han, Zheyu Wen, Minkyu Choi, Zhongming Liu
Summary: By training a variational autoencoder with rsfMRI data, researchers have been able to untangle the underlying sources of brain cortical activity and connectivity, representing spatiotemporal characteristics and driving changes in cortical networks. The resultant latent variables can be used as a reliable feature for accurate subject identification, even with limited data available. This demonstrates the value of VAE for unsupervised representation learning in resting state fMRI activity.
Article
Engineering, Biomedical
Yizhen Zhang, Jung-Hoon Kim, David Brang, Zhongming Liu
Summary: Movies, audio stories, and virtual reality are increasingly used as stimuli for functional brain imaging, setting up more ecologically relevant conditions and inducing highly reproducible brain responses as compared to traditional experimental reductionism.
CURRENT OPINION IN BIOMEDICAL ENGINEERING
(2021)
Article
Neurosciences
Haiguang Wen, Junxing Shi, Yizhen Zhang, Kun-Han Lu, Jiayue Cao, Zhongming Liu
Article
Neurosciences
Jose Sanchez-Bornot, Roberto C. Sotero, J. A. Scott Kelso, Ozguer Simsek, Damien Coyle
Summary: This study proposes a multi-penalized state-space model for analyzing unobserved dynamics, using a data-driven regularization method. Novel algorithms are developed to solve the model, and a cross-validation method is introduced to evaluate regularization parameters. The effectiveness of this method is validated through simulations and real data analysis, enabling a more accurate exploration of cognitive brain functions.