课题基金 / 基金详情

Diffusion Multiscale Analysis

Diffusion Multiscale Analysis
扩散多尺度分析
批准号:
0512050
负责人:
Mauro Maggioni
金额:
$13.65万
依托单位:
依托单位国家:
美国
项目类别:
Standard Grant
财政年份:
2005
资助国家:
美国
项目状态:
已结题
起止时间:
2005-07-01 至 2006-10-31

项目摘要

项目成果

Mauro Maggioni的其他基金

相似基金

相关文献

中文摘要
翻译
在这个提议中,研究者和他的合作者解决了几个问题,这些问题来自于集合的多尺度几何的数学分析,以及函数空间的多尺度分解,这些问题来自于流形、图和其他相当一般的度量空间上的扩散半群的作用。虽然这些多尺度几何在微分几何、偏微分方程以及图论的许多分支(在计算机科学中的应用)中是部分隐式的(和经典的),但直到最近,研究者和他的合作者才引入了一个非常普遍的、有效的、连贯的和统一的构造。多尺度函数空间分解,镜像这些多尺度扩散几何结构也被构造,通过引入特殊的小波函数。这是一个深远的,长期以来一直在寻求,推广小波分析,无论是数学和计算。研究者和他的合作者已经证明,有效地计算这些多尺度分解的算法是存在的,它推广了快速小波变换和快速多极方法,产生了保证高精度的快速多尺度算法。研究人员将研究双正交扩散多尺度分解的构造,粗糙集上的多尺度函数近似,数据集的多尺度扩散分析及其与几何测度理论,多尺度马尔可夫链,偏微分方程数值分析,学习理论,高光谱成像和文档语料库分析的关系。研究人员希望这种新颖的多尺度构造在所有这些学科中产生影响,本提案强调了多尺度分析的几个方面的跨学科性质,以及思想、工具、结构对纯数学和应用数学的广泛适用性,以及对计算机科学、物理学、工程学、天文学和统计学等其他学科的广泛适用性。这些新的多尺度技术的引入揭示了新的和有趣的多尺度几何结构的图形和集合,以及有效的计算工具来发现它们。应用范围非常广泛,包括大型复杂网络的分析和组织(例如计算机网络,生物调控网络等),用于信息提取的文档语料库,高光谱图像(用于医学,目标识别等应用),和大型数据集。它还可以应用于学习和人工智能的新算法的开发,用于复杂任务的自动化。该研究员的目的是加强他现有的合作,并建立新的,与其他机构,在美国和国外,在几个学科,特别是计算机科学,天文学,生物学和医学。他将继续与开发下一代仪器的公司合作,用于高光谱成像。他将继续积极参与多学科和跨学科的会议,研讨会和研究活动,并有效地沟通和传播的想法和技术,以多学科的观众,使他的工作,包括论文和相应的算法的计算机代码,易于电子访问。
英文摘要
In this proposal, the investigator and his collaborators address several questions arising from the mathematical analysis of multiscale geometries of sets, and multiscale decomposition of function spaces, that arise from the action of a diffusion semigroup on a manifold, a graph and other rather general metric spaces. While these multiscale geometries are partly implicit (and classical) in differential geometry, in partial differential equations, as well as in many branches in graph theory (with applications to problems in computer science), only recently a very general, yet efficient, coherent and unifying construction has been introduced by the investigator and his collaborators. Multiscale function space decompositions that mirror these multiscale diffusion geometries are also constructed, through the introduction of special wavelet functions. This is a far-reaching, and long sought, generalization of wavelet analysis, both mathematically and computationally. The investigator and his collaborators have shown that algorithms for efficiently computing these multiscale decompositions exist, which generalize the fast wavelet transform and Fast Multipole Methods, yielding fast multiscale algorithms guaranteeing high-precision. The investigator will study the construction of biorthogonal diffusion multiscale decompositions, multiscale function approximation on rough sets, multiscale diffusion analysis of data sets and its relationships with geometric measure theory, multiscale Markov chains, numerical analysis of PDEs, learning theory, hyperspectral imaging and document corpora analysis.The investigator expects this novel multiscale construction to have impact in all these disciplines, in a way similar to the impact wavelet analysis had on low-dimensional signal processing and numerical analysis.The present proposal stresses the inter-disciplinary nature of several aspects of multiscale analysis, and the vast applicability of the ideas, tools, constructions, to pure and applied mathematics, and to other disciplines such as computer science, physics, engineering, astronomy and statistics, among others. The introduction of these novel multiscale techniques reveals new and interesting multiscale geometric structures of graphs and sets, together with effective computational tools to discover them. The range of applications is very wide, and includes the analysis and organization of large and complex networks (e.g. computer networks, biological regulatory networks etc...), document corpora for information extraction, hyperspectral imagery (for applications to medicine, target recognition etc...), and large datasets in general. It has also applications to the development of new algorithms for learning and artificial intelligence, for the automation of complex tasks. The investigator aims at strenghtening his existing collaborations, and establishing new ones, with other institutions, both in the United States and abroad, across several disciplines, in particular computer science, astronomy, biology, and medicine. He will continue his existing collaborations with companies developing next-generation instrumentation, for applications to hyperspectral imaging. He will continue to actively participate in multi- and inter-disciplinary conferences, workshops and research activities, and effectively communicating and disseminating ideas and techniques to multi-disciplinary audiences, making his work, including papers and computer code for the corresponding algorithms, easily accessible electronically.
期刊论文(0)
专著(0)
科研奖励(0)
会议论文
BIGDATA: F: Compositional Learning, Maps and Transfer: Statistical and Machine Learning on Collections of Data Sets
  • 批准号:
    1837991
  • 项目类别:
    Standard Grant
  • 资助金额:
    $70.0万
  • 财政年份:
    2019
  • 负责人:
    Mauro Maggioni
  • 依托单位:
ATD: Estimation and Anomaly Detection for high-dimensional Data, Maps and Dynamic Processes
  • 批准号:
    1737984
  • 项目类别:
    Standard Grant
  • 资助金额:
    $25.0万
  • 财政年份:
    2017
  • 负责人:
    Mauro Maggioni
  • 依托单位:
ATD: Online Multiscale Algorithms for Geometric Density Estimation in High-Dimensions and Persistent Homology of Data for Improved Threat Detection
  • 批准号:
    1756892
  • 项目类别:
    Standard Grant
  • 资助金额:
    $37.99万
  • 财政年份:
    2016
  • 负责人:
    Mauro Maggioni
  • 依托单位:
Collaborative Proposal: SI2-CHE: ExTASY Extensible Tools for Advanced Sampling and analYsis
  • 批准号:
    1708353
  • 项目类别:
    Standard Grant
  • 资助金额:
    $14.56万
  • 财政年份:
    2016
  • 负责人:
    Mauro Maggioni
  • 依托单位:
海外基金