2. 您的位置
3. 首页
4. 会议论文
5. A Novel Geometric Multiscale Approach to Structured Dictionary Learning on High Dimensional Data

A Novel Geometric Multiscale Approach to Structured Dictionary Learning on High Dimensional Data

来源 :第八届工业与应用数学国际大会 | 被引量 : 0次 | 上传用户：ronalito

【摘 要】

：

　　Many real data sets have high ambient dimensions,but they are often intrinsically low-dimensional.We exploit such an assumption for efficiently representing

【作 者】

：

Chen Guangliang 

【机 构】

：

San Jose State Univ.

【出 处】

：

第八届工业与应用数学国际大会

【发表日期】

：

2015年期

下载到本地 , 更方便阅读

下载此文 赞助VIP

声明 : 本文档内容版权归属内容提供方 , 如果您对本文有版权争议 , 可与客服联系进行内容授权或下架

论文部分内容阅读

　　Many real data sets have high ambient dimensions,but they are often intrinsically low-dimensional.We exploit such an assumption for efficiently representing high-dimensional data by using an adaptive,sparsifying dictionary.

其他文献

Central-Upwind Schemes for Systems of Balance Laws

　　I will present second-order well-balanced central-upwind schemes for systems of hyperbolic balance laws and their applications to the Euler equations of gas

会议

A New Hydrostatic Reconstruction Scheme Motivated by the Wet-dry Front

　　We consider well-balanced,positivity preserving numerical schemes for the shallow water equations.It has recently been noted by Berthon et al.and Morales et

会议

Timescale Separation and Forecasting with Dynamics-adapted Kernels

　　We discuss kernel methods for extracting intrinsic timescales of dynamical systems and nonparametric forecasting.These so-called cone kernels utilize the ti

会议

Statistical Learning for Model Reduction with ATLAS

　　We discuss ATLAS,a statistical learning framework for model reduction of high-dimensional dynamical systems with few intrinsic degrees of freedom.The algori

会议

Uncertainty Quantification(and Sensitivity!)in Fluid Dynamics and Control

　　Fluid systems are characterized by high-dimensional,nonlinear dynamics,although they often evolve on low-dimensional attractors.The behavior of these attrac

会议

Data-driven methods for quantifying uncertainty of multiscale dynamical systems

　　A major challenge in contemporary applied science is to design efficient models for predicting dynamical behavior resulting from complex interaction of mult

会议

Data Analysis Tools for Large-scale Computer Vision and Multi-media Information Retrieval

　　The rapid growth of the amounts of produced and consumed visual and multi-media information challenges existing tools and practices used in the organization

会议

Geometric Understanding of Data in 3D and Higher

　　Rapid development of data acquisition technology stimulates research on developing new computational tools for analyzing and processing data to make more ef

会议

Matrix Method for Persistence Modules on Commutative Ladders of Finite Type

　　Motivated by applications,a recent paper extends the ideas of persistent homology to the commutative ladder quivers.We discuss an algorithm to compute indec

会议

Towards Optimal Control of Gliders for Velocity Field Assimilation

　　Autonomous vehicles,or gliders,offer flexible platforms for the collection of physical oceanography data,and are being deployed in a number of projects of s

会议