Article
Multidisciplinary Sciences
Chin-Wei Lin, Shu-Hsien Liao, Han-Sheng Huang, Li-Min Wang, Jyh-Horng Chen, Chia-Hao Su, Kuen-Lin Chen
Summary: The simulation work demonstrated that multiplane scanning does not significantly improve accuracy in source localization for small animal magnetic particle imaging systems, while the use of a gradient scan method shows potential for enhancing accuracy compared to constant field methods when the source is further away from the sensors.
SCIENTIFIC REPORTS
(2021)
Article
Engineering, Electrical & Electronic
Abderrahim Halimi, Aurora Maccarone, Robert A. Lamb, Gerald S. Buller, Stephen McLaughlin
Summary: This paper introduces a hierarchical Bayesian algorithm for the robust reconstruction of multispectral single-photon Lidar data in challenging environments, providing robust depth and reflectivity estimates using multi-scale information to assist decision making. The weight-based strategy allows the utilization of guide information obtained by state-of-the-art learning based algorithms, with validation showing competitive results in terms of quality of inferences and computational complexity compared to existing algorithms.
IEEE TRANSACTIONS ON COMPUTATIONAL IMAGING
(2021)
Article
Mathematics
Ru Zhao, Jingjing Liu
Summary: This paper proposes a new three-step method based on the fractional-order variational method and data-driven tight frame to solve the problem of multi-modal image fusion for images corrupted by Poisson noise. The fused high-quality images are obtained while removing Poisson noise using the split Bregman algorithm, which has stability and fast convergence. Numerical results show the excellent performance of the proposed method in terms of evaluation metrics and visual quality.
Article
Computer Science, Interdisciplinary Applications
Sumati Mahajan, S. K. Gupta
Summary: The paper introduces an extension of inexact quadratic programming (IQP) by incorporating previously overlooked terms into the formulation, thus expanding its scope of applications and improving computational efficiency.
COMPUTERS & INDUSTRIAL ENGINEERING
(2021)
Article
Mathematics, Applied
Vando A. Adona, Max L. N. Goncalves
Summary: In this paper, an inexact version of the symmetric proximal alternating direction method of multipliers (ADMM) is proposed and analyzed for solving linearly constrained optimization problems. The method allows the first subproblem to be solved inexactly with a relative approximate criterion. Global O(1/k) pointwise and O(1/k) ergodic convergence rates are established for a domain of the acceleration parameters, which is consistent with the largest known rates in the exact case. Numerical experiments demonstrate the practical advantages of the proposed method. This work is the first one to study an inexact version of the symmetric proximal ADMM.
NUMERICAL ALGORITHMS
(2023)
Article
Operations Research & Management Science
Silvia Bonettini, Peter Ochs, Marco Prato, Simone Rebegoldi
Summary: In this paper, a novel abstract descent scheme for minimizing proper and lower semicontinuous functions is introduced. The scheme generalizes properties crucial for the convergence of first-order methods in nonsmooth nonconvex optimization problems. Two inertial-type algorithms, i2Piano and iPila, are proposed, which have the potential to escape local minimizers by leveraging inertial features. Both algorithms enjoy the full convergence guarantees of the abstract descent scheme, making them efficient for general nonsmooth nonconvex optimization.
COMPUTATIONAL OPTIMIZATION AND APPLICATIONS
(2023)
Article
Geography, Physical
Binjie Chen, Yang Ye, Cheng Tong, Jinsong Deng, Ke Wang, Yang Hong
Summary: This article proposes a novel big data based iterative variation mining framework (IVMF) to reconstruct large-scale aerosol optical depth (AOD) data over China from 2000 to 2020. The results demonstrate that the IVMF can effectively and accurately resolve the missing AOD data problem, and have great potential to be generalized to other regions and remote sensing products.
GISCIENCE & REMOTE SENSING
(2022)
Article
Operations Research & Management Science
Xiaoqi Yang, Chenchen Zu
Summary: This paper investigates an inexact quasisubgradient method with extrapolation for solving a quasiconvex optimization problem, establishing convergence results and complexity analysis under certain conditions. Numerical testing demonstrates that using extrapolation is more efficient in terms of iterations needed for reaching an approximate optimal solution.
JOURNAL OF OPTIMIZATION THEORY AND APPLICATIONS
(2022)
Article
Mathematics, Applied
Min-Li Zeng, Zhong Zheng
Summary: This paper studies efficient algorithms for solving nonlinear saddle point problems and proposes an improved algorithmic framework. The effectiveness of the algorithm is demonstrated through theoretical analysis and numerical experiments.
NUMERICAL ALGORITHMS
(2023)
Article
Plant Sciences
Wentao Liu, Chenglin Wang, De Yan, Weilin Chen, Lufeng Luo
Summary: This study explores a method for estimating grape feature parameters based on point cloud information: segment the grape point cloud using filtering and region growing algorithm, and register the complete grape point cloud model using an improved iterative closest point algorithm. The grape bunch surface was reconstructed using the Poisson algorithm after estimating model phenotypic size characteristics. Comparative analysis with existing methods shows that the proposed algorithm provides the closest estimation results to the measured parameters.
FRONTIERS IN PLANT SCIENCE
(2022)
Article
Health Care Sciences & Services
Ali Karamoozian, Mohammad Reza Baneshi, Abbas Bahrampour
Summary: This study proposed two models using continuous gamma distribution and discrete hyper-Poisson distribution as frailty random variables, with Weibull distribution and generalized modified Weibull distribution as baseline distributions. Through simulation and analysis of gastric cancer patient data, it was found that the model with modified Weibull and hyper-Poisson distributions fits better for practical studies compared to the model with Weibull and Gamma distributions.
STATISTICAL METHODS IN MEDICAL RESEARCH
(2021)
Article
Multidisciplinary Sciences
Evelyn Cueva, Alexander Meaney, Samuli Siltanen, Matthias J. Ehrhardt
Summary: This work discusses synergistic multi-spectral CT reconstruction that combines information from all energy channels to enhance reconstruction of each individual channel. By fusing available data to obtain a polyenergetic image and using directional total variation as prior information, improvements in image quality and computational speed are observed. The study also analyzes the use of directional total variation in variational regularization and iterative regularization processes.
PHILOSOPHICAL TRANSACTIONS OF THE ROYAL SOCIETY A-MATHEMATICAL PHYSICAL AND ENGINEERING SCIENCES
(2021)
Article
Mathematics, Applied
Tatiana A. Bubba, Luca Ratti
Summary: Statistical inverse learning theory, at the intersection of inverse problems and statistical learning, has gained increasing attention lately. This study extends the convergence rates of a class of convex regularizers and applies it to a special class of non-tight Banach frames. The theoretical results are validated and applied in the context of x-ray tomography.
Article
Computer Science, Artificial Intelligence
Willem Diepeveen, Jan Lellmann
Summary: The proposed method uses a Riemannian semismooth Newton method for solving nonsmooth optimization problems on manifolds, achieving superlinear convergence by applying it to a recent extension of Fenchel duality theory on Riemannian manifolds. Numerical experiments confirm the superlinear convergence on manifolds with positive and negative curvature, regardless of the sign of the curvature.
SIAM JOURNAL ON IMAGING SCIENCES
(2021)
Article
Biology
Carsten Jentsch, Eun Ryung Lee, Enno Mammen
Summary: The Poisson reduced-rank models are discussed for low-dimensional summaries of high-dimensional Poisson vectors, allowing inference on individual locations in a low-dimensional space. Consistent estimation of locations can be achieved using Poisson maximum likelihood estimation under weak dependence conditions. This method is applied to political text data to make statistical inferences on the multi-dimensional evolution of party positions.
Article
Mathematics, Applied
S. Bonettini, A. Benfenati, V. Ruggiero
SIAM JOURNAL ON OPTIMIZATION
(2016)
Article
Engineering, Electrical & Electronic
A. Benfenati, E. Chouzenoux, J-c Pesquet
Article
Computer Science, Artificial Intelligence
Alessandro Benfenati, Francesco Bonacci, Tarik Bourouina, Hugues Talbot
Summary: The paper introduces a method to estimate the 3D coordinates of spherical particles' centers by processing scanned volume images and utilizing Total Variation functional and regularized weighted Least Squares fit. Experiments show the significance of image denoising for particle tracking procedures.
JOURNAL OF MATHEMATICAL IMAGING AND VISION
(2021)
Article
Mathematics
Giacomo Aletti, Alessandro Benfenati, Giovanni Naldi
Summary: Networks and graphs are powerful tools for studying the spread of infection in human and animal populations, especially in analyzing directly transmitted infectious diseases in heterogeneous populations. This paper presents a multi-group version of the epidemiological SEIR model, taking into account heterogeneity in contact weights between different groups and proposing a simple control algorithm to optimize connection weights and minimize economic and social costs. Numerical simulations are provided to support the findings.
Article
Mathematics
G. Aletti, A. Benfenati, G. Naldi
Summary: We prove the existence and uniqueness of the solution to a novel feature-preserving nonlinear nonlocal diffusion equation for signal denoising in the one-dimensional case. The equation is based on a new diffusivity coefficient that utilizes a nonlocal automatically detected parameter related to the local bounded variation and oscillating pattern of the noisy input signal.
BULLETIN OF THE AUSTRALIAN MATHEMATICAL SOCIETY
(2022)
Article
Mathematics, Applied
Alessandro Benfenati, Giacomo Borghi, Lorenzo Pareschi
Summary: This work introduces a new class of gradient-free global optimization methods based on a binary interaction dynamics governed by a Boltzmann type equation. Convergence to the global minimizer is guaranteed for a large class of functions under appropriate parameter constraints. The resulting Fokker-Planck partial differential equations generalize the current class of consensus based optimization methods.
APPLIED MATHEMATICS AND OPTIMIZATION
(2022)
Article
Computer Science, Artificial Intelligence
Alessandro Benfenati, Alessio Marta
Summary: This paper proposes a geometric framework for studying deep neural networks, viewing them as sequences of mappings between manifolds using singular Riemannian geometry. The authors present an application of this framework, specifically focusing on constructing equivalence classes of input points that are mapped to the same output by the network. This approach has practical implications in generating synthetic data and understanding the sensitivity of classifiers to input perturbations.
Article
Computer Science, Artificial Intelligence
Alessandro Benfenati, Alessio Marta
Summary: This study investigates a sequence of maps between manifolds, with the last manifold equipped with a Riemannian metric. The structures induced through pullbacks on the other manifolds of the sequence and on related quotients are explored. It is shown that the pullbacks of the final Riemannian metric induce a degenerate Riemannian metric on any manifold of the sequence, resulting in a pseudometric space structure. The Kolmogorov quotient of this pseudometric space is proven to be a smooth manifold, serving as the base space of a specific vertical bundle. The theoretical properties of the maps in this sequence are examined, particularly focusing on maps between manifolds that implement neural networks of practical interest. Applications of the introduced geometric framework are presented.
Article
Mathematics, Applied
A. Benfenati, A. Catozzi, V Ruggiero
Summary: The blind deconvolution problem is challenging in scientific imaging fields such as microscopy, medicine, and astronomy. Deep learning techniques have gained interest for their impressive performance in reconstruction using the Deep Image Prior framework. In this paper, the authors propose a modified approach that considers the noise and regularization functions specific to microscopy images and achieves promising results.
Article
Imaging Science & Photographic Technology
Alessandro Benfenati
Summary: This study employs Deep Learning techniques to address issues in microscopy imaging, particularly related to auto-fluorescence and noise. The proposed architecture, based on U-Nets, successfully removes fluorescence and acts as a denoiser for different types of noise. The performance of this approach is evaluated on real microscopy images and applied for particle recognition.
JOURNAL OF IMAGING
(2022)
Article
Imaging Science & Photographic Technology
Giacomo Aletti, Alessandro Benfenati, Giovanni Naldi
Summary: A new semi-supervised method for multilabel segmentation of hyperspectral images was proposed, combining linear discriminant analysis, a similarity index, and a random walk-based model. User-marked regions were used to project the original feature space to a lower dimension, maximizing class separation while automatically retaining informative features and reducing computational burden. The method involved a combinatorial Dirichlet problem with weighted graph nodes representing projected pixels and probabilities assigned to each pixel indicating subregion likelihood.
JOURNAL OF IMAGING
(2021)
Article
Imaging Science & Photographic Technology
Giacomo Aletti, Alessandro Benfenati, Giovanni Naldi
Summary: This study proposes a new multi-label image segmentation method that combines a random walk model with direct label assignment using a suitable color distance calculation. The approach is semi-automatic and involves user interaction for initialization. By computing color distances, the segmentation process is optimized to improve segmentation quality.
JOURNAL OF IMAGING
(2021)
Proceedings Paper
Computer Science, Interdisciplinary Applications
A. Benfenati, P. Causing, M. G. Lupieri, G. Naldi
9TH INTERNATIONAL CONFERENCE ON NEW COMPUTATIONAL METHODS FOR INVERSE PROBLEMS, NCMIP 2019
(2020)
Proceedings Paper
Computer Science, Artificial Intelligence
Elodie Puybareau, Edwin Carlinet, Alessandro Benfenati, Hugues Talbot
MATHEMATICAL MORPHOLOGY AND ITS APPLICATIONS TO SIGNAL AND IMAGE PROCESSING, ISMM 2019
(2019)
Proceedings Paper
Acoustics
Alessandro Benfenati, Emilie Chouzenoux, Jean-Christophe Pesquet
2018 IEEE INTERNATIONAL CONFERENCE ON ACOUSTICS, SPEECH AND SIGNAL PROCESSING (ICASSP)
(2018)