Article
Geochemistry & Geophysics
Jian He, Huailiang Li, Xianguo Tuo, Xiaotao Wen, Liyuan Feng
Summary: This article introduces a wavelet-weighted online dictionary learning (WWODL) denoising strategy for noisy microseismic recordings. An adaptive parameters estimation approach is developed for tunable Q-factor wavelet transform (TQWT) to provide accurate subband information for online dictionary learning (ODL). Results show that WWODL can effectively suppress noise and minimally impact the first arrival signal.
IEEE TRANSACTIONS ON GEOSCIENCE AND REMOTE SENSING
(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
Statistics & Probability
Li-Pang Chen, Grace Y. Yi
Summary: This paper investigates the analysis of error-prone data under graphical models. The asymptotic bias of the naive analysis without considering measurement error is studied, and a de-noising estimation procedure is developed to account for the impact of measurement error.
ELECTRONIC JOURNAL OF STATISTICS
(2022)
Article
Mathematics, Applied
Xiao Li, Shixiang Chen, Zengde Deng, Qing Qu, Zhihui Zhu, Anthony Man-Cho So
Summary: This study focuses on nonsmooth optimization problems over the Stiefel manifold and the convergence properties of using Riemannian subgradient-type methods to solve these problems, providing new convergence guarantees such as iteration complexity and local linear convergence, which contributes to the optimization methods in this field.
SIAM JOURNAL ON OPTIMIZATION
(2021)
Article
Computer Science, Information Systems
Zhiyong Che, Bo Liu, Yanshan Xiao, Luyue Lin
Summary: Transfer learning is a popular method in machine learning that transfers knowledge from a source task to a target task. This paper proposes a dictionary-based transfer learning method (U-DTL) with Universum data to improve the classifier's performance.
INFORMATION SCIENCES
(2022)
Article
Automation & Control Systems
Amr Alanwar, Anne Koch, Frank Allgoewer, Karl Henrik Johansson
Summary: This paper discusses how to compute reachable sets directly from noisy data without a given system model. Several reachability algorithms are presented for different types of systems generating the data. For linear systems, an algorithm based on matrix zonotopes is proposed, which computes over-approximated reachable sets. Constrained matrix zonotopes are introduced to provide less conservative reachable sets at the cost of increased computational expenses and incorporate prior knowledge about the unknown system model. The approach is also extended to polynomial and nonlinear systems with theoretical guarantees of proper over-approximation.
IEEE TRANSACTIONS ON AUTOMATIC CONTROL
(2023)
Article
Geosciences, Multidisciplinary
Yangqin Guo, Si Guo, Ke Guo, Huailai Zhou
Summary: A novel method combining MCA, DL, and deep noise reduction is proposed to attenuate random and coherent noise in seismic data while preserving effective signals. This method consists of three steps that effectively suppress undesired noise, maximize the preservation of geological bodies and structures, and improve the signal-to-noise ratio. The algorithm provides new ideas for data processing to enhance the quality and S/N of seismic data.
INTERNATIONAL JOURNAL OF EARTH SCIENCES
(2021)
Article
Computer Science, Information Systems
Jian Chen, Lan Du, Yuchen Guo
Summary: This paper introduces a novel approach for small sample classification, the Label Constrained Convolutional Factor Analysis (LCCFA) model, which uses dictionary atoms as small convolution kernels to learn shared features of observed data and improve classification performance. By projecting weight vectors onto class labels, it increases inter-class separability, and efficient parameter estimation is achieved through the VB algorithm.
INFORMATION SCIENCES
(2021)
Article
Geochemistry & Geophysics
Feng Wang, Bo Yang, Yuqing Wang, Ming Wang
Summary: This paper proposes an unsupervised denoising method based on model-based deep learning, which combines domain knowledge and a data-driven approach. This method reduces the dependency on labeled data and explores insights into the denoising system. Experimental results demonstrate that the proposed method achieves competitive performance compared to conventional methods.
IEEE TRANSACTIONS ON GEOSCIENCE AND REMOTE SENSING
(2022)
Article
Computer Science, Information Systems
Junbao Zhuo, Shuhui Wang, Qingming Huang
Summary: This paper addresses the problem of domain adaptation under noisy environments. It models the uncertainty in the predictions of a convolutional neural network classifier and adjusts the classification loss accordingly to reduce the influence of noise. It also introduces a novel regularizer, UncertaintyRank, which makes the uncertainty more sensitive to noisy labels and helps to eliminate the adverse effects of noisy representations while estimating the domain discrepancy.
IEEE TRANSACTIONS ON MULTIMEDIA
(2023)
Article
Mathematics, Applied
Richard Archibald, Hoang Tran
Summary: There has been an increasing interest in developing and applying dictionary learning (DL) to process massive datasets. This paper presents a new DL approach for compressing and denoising massive datasets in real time, where the data are streamed through in a preset order. The proposed approach incrementally builds up the dictionary and applies denoising directly to the snapshots before sparse coding, achieving competitive performance in data reconstruction and denoising.
DISCRETE AND CONTINUOUS DYNAMICAL SYSTEMS-SERIES S
(2022)
Article
Computer Science, Artificial Intelligence
Zhongyi Han, Xian-Jin Gui, Haoliang Sun, Yilong Yin, Shuo Li
Summary: In this paper, a noise-robust domain adaptation method is proposed to address the issue of corrupted source domain examples in multiple noisy environments. By utilizing offline curriculum learning, gradually decreasing noisy distribution distance, estimating open-set noise degree, robust parameter learning, and domain-invariant feature learning, these components are seamlessly transformed into an adversarial network for efficient joint optimization, leading to significant improvements in transfer tasks.
IEEE TRANSACTIONS ON PATTERN ANALYSIS AND MACHINE INTELLIGENCE
(2023)
Article
Engineering, Multidisciplinary
Xuebin Qin, Yutong Shen, Jiachen Hu, Mingqiao Li, Peijiao Yang, Chenchen Ji, Xinlong Zhu
Summary: This paper introduces the application of compressed sensing theory based on dictionary learning in electrical capacitance tomography (ECT) and establishes a nonlinear mapping between observed capacitance and sparse space using the K-SVD algorithm. By modeling and simulating the two-phase flow distribution in a cylindrical pipe, three variations without sparse constraint based on Landweber, Tikhonov, and Newton-Raphson algorithms were used to rapidly reconstruct a 2-D image.
CMES-COMPUTER MODELING IN ENGINEERING & SCIENCES
(2022)
Article
Computer Science, Artificial Intelligence
Jing Dong, Kai Wu, Chang Liu, Xue Mei, Wenwu Wang
Summary: Dictionary learning is widely used in signal and image processing. The Discriminative Convolutional Analysis Dictionary Learning (DCADL) algorithm has shown promising results in image classification with low computational complexity. However, it is limited by the lack of constraints on dictionary structures. To address this, this study introduces an adaptively ordinal locality preserving (AOLP) term to DCADL, improving classification performance by preserving distance ranking and training a linear classifier. The proposed model is solved using a new optimization method. Experimental results demonstrate the effectiveness of the proposed algorithm.
Article
Computer Science, Artificial Intelligence
Haishun Du, Yonghao Zhang, Luogang Ma, Fan Zhang
Summary: The SDADL method proposes a structured discriminant analysis dictionary learning approach to improve pattern classification by associating class-specific analysis sub-dictionaries. It introduces classification error term, discriminant analysis sparse code error term, and structured discriminant term to optimize the dictionary learning process, along with designing an efficient iterative algorithm for optimization.
KNOWLEDGE-BASED SYSTEMS
(2021)