A5074490643- Sparse and Compressive Sensing Techniques
- Knee injuries and reconstruction techniques
- Single-cell and spatial transcriptomics
- Total Knee Arthroplasty Outcomes
- Molecular Biology Techniques and Applications
- Osteoarthritis Treatment and Mechanisms
- Blind Source Separation Techniques
- Genomics and Phylogenetic Studies
- Microwave Imaging and Scattering Analysis
- Cephalopods and Marine Biology
- Gene expression and cancer classification
- Cancer Genomics and Diagnostics
- Advanced Bandit Algorithms Research
- Numerical methods in inverse problems
- Direction-of-Arrival Estimation Techniques
- Reinforcement Learning in Robotics
- Advanced Neuroimaging Techniques and Applications
- Statistical Methods and Inference
- Machine Learning and Algorithms
- Advanced X-ray Imaging Techniques
- Concrete and Cement Materials Research
- Genomic variations and chromosomal abnormalities
- Fluid Dynamics and Heat Transfer
- Neural Networks and Applications
- RNA modifications and cancer
University of Chicago
2021-2025
National Cheng Kung University
2025
Anhui Medical University
2022-2024
First Affiliated Hospital of Anhui Medical University
2022-2024
Princeton University
2018-2024
China University of Mining and Technology
2024
Soochow University
2024
Northwestern Polytechnical University
2011-2024
Second Affiliated Hospital of Guangzhou Medical University
2024
Guangzhou Medical University
2024
Abstract Emerging spatial technologies, including transcriptomics and epigenomics, are becoming powerful tools for profiling of cellular states in the tissue context 1–5 . However, current methods capture only one layer omics information at a time, precluding possibility examining mechanistic relationship across central dogma molecular biology. Here, we present two technologies spatially resolved, genome-wide, joint epigenome transcriptome by cosequencing chromatin accessibility gene...
Recent years have seen a flurry of activities in designing provably efficient nonconvex procedures for solving statistical estimation problems. Due to the highly nature empirical loss, state-of-the-art often require proper regularization (e.g., trimming, regularized cost, projection) order guarantee fast convergence. For vanilla such as gradient descent, however, prior theory either recommends conservative learning rates avoid overshooting, or completely lacks performance guarantees. This...
Chromatin accessibility is essential in regulating gene expression and cellular identity, alterations have been implicated driving cancer initiation, progression metastasis
To study the spatial interactions among cancer and non-cancer cells
Significance Matrix completion finds numerous applications in data science, ranging from information retrieval to medical imaging. While substantial progress has been made designing estimation algorithms, it remains unknown how perform optimal statistical inference on the matrix given obtained estimates—a task at core of modern decision making. We propose procedures debias popular convex and nonconvex estimators derive distributional characterizations for resulting debiased estimators. This...
This paper is concerned with the problem of top-K ranking from pairwise comparisons. Given a collection n items and few comparisons across them, one wishes to identify set K that receive highest ranks. To tackle this problem, we adopt logistic parametric model - Bradley-Terry-Luce model, where each item assigned latent preference score, outcome comparison depends solely on relative scores two involved. Recent works have made significant progress towards characterizing performance (e.g. mean...
Deep learning has achieved tremendous success in recent years. In simple words, deep uses the composition of many nonlinear functions to model complex dependency between input features and labels. While neural networks have a long history, advances significantly improved their empirical performance computer vision, natural language processing other predictive tasks. From statistical scientific perspective, it is ask: What learning? are new characteristics learning, compared with classical...
This paper studies noisy low-rank matrix completion: given partial and entries of a large matrix, the goal is to estimate underlying faithfully efficiently. Arguably one most popular paradigms tackle this problem convex relaxation, which achieves remarkable efficacy in practice. However, theoretical support approach still far from optimal setting, falling short explaining its empirical success. We make progress towards demystifying practical relaxation vis-à-vis random noise. When rank...
Offline reinforcement learning (RL) algorithms seek to learn an optimal policy from a fixed dataset without active data collection. Based on the composition of offline dataset, two main methods are used: imitation which is suitable for expert datasets, and vanilla RL often requires uniform coverage datasets. From practical standpoint, datasets deviate these extremes exact usually unknown. To bridge this gap, we present new framework, called single-policy concentrability, that smoothly...
Abstract Metal halide perovskites are ideal candidates for indoor photovoltaics (IPVs) due to their tunable bandgaps, which allow the active layers be optimized artificial light sources. However, significant non‐radiative carrier recombination under low‐light conditions has limited full potential of perovskite‐based IPVs. To address this challenge, an integration perylene diimide (PDI)‐based sulfobetaines as cathode interlayers (CILs) is proposed and impact varying alkyl chain length (from...
This paper delivers improved theoretical guarantees for the convex programming approach in low-rank matrix estimation, presence of (1) random noise, (2) gross sparse outliers, and (3) missing data. problem, often dubbed as
Spectral methods have emerged as a simple yet surprisingly effective approach for extracting information from massive, noisy and incomplete data.In nutshell, spectral refer to collection of algorithms built upon the eigenvalues (resp.singular values) eigenvectors vectors) some properly designed matrices constructed data.A diverse array applications been found in machine learning, imaging science, financial econometric modeling, signal processing, including recommendation systems, community...
Abstract Developing neural electrodes with high stretchability and stable conductivity is a promising method to explore applications of them in biological medicine electronic skin. However, considering the poor mechanical typical conductive materials, maintaining connection electrode paths under stretching still challenge. Herein, for first time, double‐microcrack coupling strategy highly stretchable proposed. Compared single‐layer microcrack electrodes, design utilizes complement between...
Deep learning has arguably achieved tremendous success in recent years. In simple words, deep uses the composition of many nonlinear functions to model complex dependency between input features and labels. While neural networks have a long history, advances greatly improved their performance computer vision, natural language processing, etc. From statistical scientific perspective, it is ask: What learning? are new characteristics learning, compared with classical methods? theoretical...
Low-rank matrix estimation is a canonical problem that finds numerous applications in signal processing, machine learning and imaging science. A popular approach practice to factorize the into two compact low-rank factors, then optimize these factors directly via simple iterative methods such as gradient descent alternating minimization. Despite nonconvexity, recent literatures have shown heuristics fact achieve linear convergence when initialized properly for growing number of problems...
Many problems in data science can be treated as estimating a low-rank matrix from highly incomplete, sometimes even corrupted, observations. One popular approach is to resort factorization, where the factors are optimized via first-order methods over smooth loss function, such residual sum of squares. While tremendous progresses have been made recent years, natural formulation suffers two sources ill-conditioning, iteration complexity gradient descent scales poorly both with dimension well...
We consider the problem of recovering low-rank matrices from random rank-one measurements, which spans numerous applications including covariance sketching, phase retrieval, quantum state tomography, and learning shallow polynomial neural networks, among others. Our approach is to directly estimate factor by minimizing a nonconvex quadratic loss function via vanilla gradient descent, following tailored spectral initialization. When true rank small, this algorithm guaranteed converge ground...