课题基金 / 基金详情

Assessment of Deep Learning through Meanfield Theory

Assessment of Deep Learning through Meanfield Theory
通过平均场理论评估深度学习
批准号:
462234017
负责人:
Professor Dr. Michael Herty
金额:
$0.0万
依托单位国家:
德国
项目类别:
Priority Programmes
财政年份:
--
资助国家:
德国
项目状态:
未结题
起止时间:

项目摘要

项目成果

Professor Dr. Michael Herty的其他基金

相似基金

相关文献

中文摘要
翻译
动力学和平均场理论已被证明是一种有用的数学工具,用于对各种物理和社会过程进行分层建模。它特别允许研究作为粒子对粒子动力学结果的涌现行为。现代学习方法可以在数学上重新表述,从而出现粒子相互作用结构。例如~在无限多层极限下的剩余深度神经网络类导致神经元激活状态的常微分方程的耦合系统。这个系统可以被重新表述为一个相互作用的“粒子”系统,其中每个粒子的状态对应于神经元在某个时间点的激活状态。 在这个提议中,我们的目标是利用和扩展现有的平均场理论,为现代学习方法提供一个数学框架,从而实现这种描述,并专注于通过深度神经网络进行学习和通过过滤方法进行学习。 动力学和平均场理论的方法将被扩展,以深入了解这些学习方法的属性和机制。所获得的洞察力将用于提出新的,可证明收敛和稳定的方法来解决学习问题。导出的平均场和动力学描述将允许进行理论研究,以深入了解它们的涌现行为。这些包括但不限于快速和稳定的训练程序,关于参数和表达性的不确定性的鲁棒性研究,以及涉及偏微分方程的数据驱动学习问题。
英文摘要
Kinetic and meanfield theory has proven an useful mathematical tool for hierarchical modeling of a variety of physical and sociological processes. It in particular allows to study emergent behavior as consequence of particle--to--particle dynamics. Modern learning methods can mathematically be reformulated such that an particle interaction structure emerges. E.g.~the class of residual deep neural networks in the limit of infinitely many layer leads to a coupled system of ordinary differential equations for the activation state of neurons. This arising system can be reformulated as an interacting 'particle' system where the state of each particle corresponds to the activation state of a neuron at a certain point in time. Within this proposal, we aim to exploit and extend existing meanfield theory to provide a mathematical framework for modern learning methods that allow for this description and hence focusing on learning by deep neural networks and learning through filtering methods. Methods from kinetic and meanfield theory will be extended in order to gain insight on properties and mechanisms of those learning approaches. The gained insight will used to propose novel, provable convergent and stable methods to solve learning problems. The derived meanfield and kinetic description will allow to perform theoretical studies to gain insight on their emergent behaviors. These include but are not limited to fast and stable training procedures, study of robustness with respect to uncertainty in parameters and the expressivity, as well as the data driven learning problems involving partial differential equations.
期刊论文(0)
专著(0)
科研奖励(0)
会议论文
Basic evaluation for simulation-based crash-risk-models - multiscale modelling regarding dynamic traffic flow states
Kinetic Models on Networks with Applications Traffic Flow and Supply Chains
Differentiable programming for flows with discontinuities
Numerical Schemes for Coupled Multi-Scale Problems
国内基金
海外基金
Deep Seek引导下预防肝硬化腹水患者发生腹腔感染的约翰霍普金斯循证实践模型下中医护理策略的构建研究
  • 批准号:
    2026JJ81909
  • 项目类别:
    省市级项目
  • 资助金额:
    --
  • 批准年份:
    2026
  • 负责人:
    胡曦
  • 依托单位:
基于Deep Unrolling的高分辨近红外二区荧光分子断层成像方法研究
  • 批准号:
    12271434
  • 项目类别:
    面上项目
  • 资助金额:
    46万元
  • 批准年份:
    2022
  • 负责人:
    贺小伟
  • 依托单位:
基于深度森林(Deep Forest)模型的表面增强拉曼光谱分析方法研究
  • 批准号:
    2020A151501709
  • 项目类别:
    省市级项目
  • 资助金额:
    10.0万元
  • 批准年份:
    2020
  • 负责人:
    谢怡
  • 依托单位:
面向Deep Web的数据整合关键技术研究
  • 批准号:
    61872168
  • 项目类别:
    面上项目
  • 资助金额:
    62.0万元
  • 批准年份:
    2018
  • 负责人:
    董永权
  • 依托单位: