A5100667140- Sparse and Compressive Sensing Techniques
- Statistical Methods and Inference
- X-ray Diffraction in Crystallography
- Crystallization and Solubility Studies
- Advanced Thermoelectric Materials and Devices
- Photoacoustic and Ultrasonic Imaging
- Advanced Image Processing Techniques
- Advanced Vision and Imaging
- Blind Source Separation Techniques
- Thermal properties of materials
- Face and Expression Recognition
- Statistical Methods and Bayesian Inference
- Advanced Causal Inference Techniques
- Microwave Imaging and Scattering Analysis
- Bayesian Methods and Mixture Models
- Advanced Image and Video Retrieval Techniques
- Physics of Superconductivity and Magnetism
- Various Chemistry Research Topics
- Photonic and Optical Devices
- Magnetic properties of thin films
- Smart Grid Security and Resilience
- Advanced Optimization Algorithms Research
- Spectroscopy and Chemometric Analyses
- Optical measurement and interference techniques
- Iron-based superconductors research
Hebei University
2025
Nanjing University of Information Science and Technology
2024
Tsinghua University
2021-2024
Shandong Normal University
2022-2024
Rutgers, The State University of New Jersey
2021-2024
Zhejiang University of Technology
2024
Xi'an Jiaotong University
2009-2023
Wuhan University of Science and Technology
2023
Renmin University of China
2023
Southeast University
2022
Blind deconvolution is the problem of recovering a convolutional kernel and an activation signal from their convolution y = a0 * x0. This ill-posed without further constraints or priors. paper studies situation where nonzero entries in are sparsely randomly populated. We normalize to have unit Frobenius norm cast sparse blind as nonconvex optimization over sphere. With this spherical constraint, every spurious local minimum turns out be close some signed shift truncation ground truth, under...
Recovering matrices from compressive and grossly corrupted observations is a fundamental problem in robust statistics, with rich applications computer vision machine learning. In theory, under certain conditions, this can be solved polynomial time via natural convex relaxation, known as principal component pursuit (CPCP). However, many existing provably convergent algorithms for CPCP suffer superlinear per-iteration cost, which severely limits their applicability to large-scale problems....
In this study, a probiotic soy-derived food was developed by novel solid-state fermentation approach using Lactobacillus plantarum B1-6. The parameters were optimized, and changes in anti-nutritional components soy seeds during investigated. Analysis response surface methodology showed the following optimized parameters: addition of 2 g/100 g sucrose, boiling for 5 min, duration 17.61 hr; under condition, resulting lactic acid bacteria (LAB) count 8.12 log cfu/ml. Solid-state significantly...
Blind deconvolution is a ubiquitous problem aiming to recover convolution kernel <sub xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">0</sub> ∈ R <sup xmlns:xlink="http://www.w3.org/1999/xlink">k</sup> and an activation signal x xmlns:xlink="http://www.w3.org/1999/xlink">m</sup> from their y . Unfortunately, this ill-posed in general. This paper focuses on the short sparse blind problem, where (k ≪ m) sparsely randomly supported (llx ll m). variant...
This paper develops new methods to recover the missing entries of a high-rank or even full-rank matrix when intrinsic dimension data is low compared ambient dimension. Specifically, we assume that columns are generated by polynomials acting on low-dimensional variable, and wish under this assumption. We show can identify complete minimum minimizing rank in high dimensional feature space. develop formulation resulting problem using kernel trick together with relaxation objective, propose an...
As science and engineering have become increasingly data-driven, the role of optimization has expanded to touch almost every stage data analysis pipeline, from signal acquisition modeling prediction. The problems encountered in practice are often nonconvex. While challenges vary problem problem, one common source nonconvexity is nonlinearity or measurement model. Nonlinear models exhibit symmetries, creating complicated, nonconvex objective landscapes, with multiple equivalent solutions....
We study the short-and-sparse (SaS) deconvolution problem of recovering a short signal $\boldsymbol{a}_0$ and sparse $\boldsymbol{x}_0$ from their convolution. propose method based on nonconvex optimization, which under certain conditions recovers target signals, up to signed shift symmetry is intrinsic this model. This plays central role in shaping optimization landscape for deconvolution. give regional analysis, characterizes geometrically, union subspaces. Our geometric characterization...
We study the $\textit{Short-and-Sparse (SaS) deconvolution}$ problem of recovering a short signal $\mathbf a_0$ and sparse x_0$ from their convolution. propose method based on nonconvex optimization, which under certain conditions recovers target signals, up to signed shift symmetry is intrinsic this model. This plays central role in shaping optimization landscape for deconvolution. give $\textit{regional analysis}$, characterizes geometrically, union subspaces. Our geometric...
Abstract Modern high-resolution microscopes are commonly used to study specimens that have dense and aperiodic spatial structure. Extracting meaningful information from images obtained such remains a formidable challenge. Fourier analysis is analyze the structure of images. However, transform fundamentally suffers severe phase noise when applied Here, we report development an algorithm based on nonconvex optimization directly uncovers fundamental motifs present in real-space image. Apart...
We investigate the landscape of negative log-likelihood function Gaussian Mixture Models (GMMs) with a general number components in population limit. As objective is non-convex, there can exist multiple spurious local minima that are not globally optimal, even for well-separated mixture models. Our study reveals all share common structure partially identifies cluster centers (i.e., means components) true location mixture. Specifically, each minimum be represented as non-overlapping...
We study the convolutional phase retrieval problem, of recovering an unknown signal x ∈ C <sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">n</sup> from m measurements consisting magnitude its cyclic convolution with a given kernel xmlns:xlink="http://www.w3.org/1999/xlink">m</sup> . This model is motivated by applications such as channel estimation, optics, and underwater acoustic communication, where interest acted on channel/filter, information...
Abstract Semi-supervised (SS) inference has received much attention in recent years. Apart from a moderate-sized labeled data, $\mathcal L$, the SS setting is characterized by an additional, larger sized, unlabeled U$. The of $|\mathcal U\ |\gg |\mathcal L\ |$, makes unique and different standard missing data problems, owing to natural violation so-called ‘positivity’ or ‘overlap’ assumption. However, most literature implicitly assumes L$ U$ be equally distributed, i.e., no selection bias...
We study the convolutional phase retrieval problem, of recovering an unknown signal $\mathbf x \in \mathbb C^n $ from $m$ measurements consisting magnitude its cyclic convolution with a given kernel C^m $. This model is motivated by applications such as channel estimation, optics, and underwater acoustic communication, where interest acted on channel/filter, information difficult or impossible to acquire. show that when a$ random number observations sufficiently large, high probability x$...
A low‐cost and environmental‐friendly direct dye‐based ink‐jet printing system was developed. novel pretreatment method employed, in which the cationic fixing agent, Matexil FC‐ER, applied as colourless ink only on image areas of fabric by printer. It found that this new could more effectively enhance colour strength improve wash fastness (greyscale ≥ 3) when compared with traditional exhaust application. The cross‐staining non‐image also apparently decreased using method. light pretreated...
A low‐cost four‐colour (RBYK) dye‐based ink‐jet printing system for textiles was introduced in this study, which red and blue inks were employed instead of the magenta cyan used half‐tone printing. The basis a colour‐management device developed by determining mapping between XYZ tristimulus values output colours digital RBYK using polynomial transforms. second‐order equation found to give best performance with an average characterisation error under 7 CIELAB units.