CIF: Medium: Collaborative Research: Theory of Optimization Geometry and Algorithms for Neural Networks
CIF:媒介:协作研究:神经网络优化几何理论和算法
基本信息
- 批准号:1901199
- 负责人:
- 金额:$ 40万
- 依托单位:
- 依托单位国家:美国
- 项目类别:Standard Grant
- 财政年份:2019
- 资助国家:美国
- 起止时间:2019-10-01 至 2024-09-30
- 项目状态:已结题
- 来源:
- 关键词:
项目摘要
Deep learning has attracted a significant amount of interest in recent years due to its widespread applicability in computer vision, artificial intelligence and natural language processing, alongside recent strides in autonomous driving. The theoretical underpinnings behind such success, however, remain elusive to a large extent, hindering its further adoption in other applications. This project aims to advance the theoretical foundations of training neural networks in terms of optimization landscape and algorithmic efficacy, which in turn should have a measurable impact on the practice of deep learning by providing guiding principles for network design, algorithm selection, hyperparameter tuning, and adversarial training. This project adopts an interdisciplinary approach fusing ideas from machine learning, optimization, statistical signal processing, high-dimensional statistics, nonparametric statistics, and information theory. This project will likewise develop courses and tutorials on theoretical foundations of large-scale machine learning and provide extensive training opportunities for students at all levels.This project aims to develop a comprehensive theory to characterize the optimization landscape and geometry of loss functions and algorithmic regularizations of major neural network training problems, and explore how the network architecture---including depth, width, and activation functions---affect these properties, thus providing guidelines for the design of algorithms to train these networks more efficiently with theoretical performance guarantees. The project will explore the geometric properties and their impact on the optimization performance in training multi-layer neural networks, auto-encoders, generative adversarial networks, and adversarial training involving non-convex and saddle-point problems.This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.
近年来,由于深度学习在计算机视觉、人工智能和自然语言处理方面的广泛应用,以及最近在自动驾驶方面的进展,深度学习引起了人们的极大兴趣。然而,这种成功背后的理论基础在很大程度上仍然难以捉摸,阻碍了它在其他应用中的进一步采用。该项目旨在从优化景观和算法功效方面推进训练神经网络的理论基础,进而通过为网络设计、算法选择、超参数调优和对抗训练提供指导原则,对深度学习的实践产生可衡量的影响。该项目采用跨学科的方法,融合了机器学习,优化,统计信号处理,高维统计,非参数统计和信息论的思想。该项目还将开发有关大规模机器学习理论基础的课程和教程,并为各级学生提供广泛的培训机会。本项目旨在发展一个全面的理论,以表征主要神经网络训练问题的损失函数和算法正则化的优化景观和几何形状,并探索网络架构(包括深度、宽度和激活函数)如何影响这些属性,从而为设计算法提供指导,从而在理论性能保证的情况下更有效地训练这些网络。该项目将探索几何性质及其对训练多层神经网络、自编码器、生成对抗网络和涉及非凸和鞍点问题的对抗训练的优化性能的影响。该奖项反映了美国国家科学基金会的法定使命,并通过使用基金会的知识价值和更广泛的影响审查标准进行评估,被认为值得支持。
项目成果
期刊论文数量(23)
专著数量(0)
科研奖励数量(0)
会议论文数量(0)
专利数量(0)
SoteriaFL: A Unified Framework for Private Federated Learning with Communication Compression
- DOI:10.48550/arxiv.2206.09888
- 发表时间:2022-06
- 期刊:
- 影响因子:0
- 作者:Zhize Li;Haoyu Zhao;Boyue Li;Yuejie Chi
- 通讯作者:Zhize Li;Haoyu Zhao;Boyue Li;Yuejie Chi
Asynchronous Gradient Play in Zero-Sum Multi-agent Games
- DOI:10.48550/arxiv.2211.08980
- 发表时间:2022-11
- 期刊:
- 影响因子:0
- 作者:Ruicheng Ao;Shicong Cen;Yuejie Chi
- 通讯作者:Ruicheng Ao;Shicong Cen;Yuejie Chi
Communication-Efficient Distributed Optimization in Networks with Gradient Tracking and Variance Reduction
- DOI:
- 发表时间:2019-09
- 期刊:
- 影响因子:0
- 作者:Boyue Li;Shicong Cen;Yuxin Chen;Yuejie Chi
- 通讯作者:Boyue Li;Shicong Cen;Yuxin Chen;Yuejie Chi
Nonconvex Optimization Meets Low-Rank Matrix Factorization: An Overview
- DOI:10.1109/tsp.2019.2937282
- 发表时间:2019-10-15
- 期刊:
- 影响因子:5.4
- 作者:Chi, Yuejie;Lu, Yue M.;Chen, Yuxin
- 通讯作者:Chen, Yuxin
Manifold Gradient Descent Solves Multi-Channel Sparse Blind Deconvolution Provably and Efficiently
流形梯度下降可证明且高效地解决多通道稀疏盲反卷积问题
- DOI:10.1109/tit.2021.3075148
- 发表时间:2021
- 期刊:
- 影响因子:2.5
- 作者:Shi, Laixi;Chi, Yuejie
- 通讯作者:Chi, Yuejie
{{
item.title }}
{{ item.translation_title }}
- DOI:
{{ item.doi }} - 发表时间:
{{ item.publish_year }} - 期刊:
- 影响因子:{{ item.factor }}
- 作者:
{{ item.authors }} - 通讯作者:
{{ item.author }}
数据更新时间:{{ journalArticles.updateTime }}
{{ item.title }}
- 作者:
{{ item.author }}
数据更新时间:{{ monograph.updateTime }}
{{ item.title }}
- 作者:
{{ item.author }}
数据更新时间:{{ sciAawards.updateTime }}
{{ item.title }}
- 作者:
{{ item.author }}
数据更新时间:{{ conferencePapers.updateTime }}
{{ item.title }}
- 作者:
{{ item.author }}
数据更新时间:{{ patent.updateTime }}
Yuejie Chi其他文献
Settling the Sample Complexity of Model-Based Offline Reinforcement Learning
解决基于模型的离线强化学习的样本复杂度
- DOI:
10.48550/arxiv.2204.05275 - 发表时间:
2022 - 期刊:
- 影响因子:0
- 作者:
Gen Li;Laixi Shi;Yuxin Chen;Yuejie Chi;Yuting Wei - 通讯作者:
Yuting Wei
Memory-Limited stochastic approximation for poisson subspace tracking
泊松子空间跟踪的内存有限随机近似
- DOI:
- 发表时间:
2017 - 期刊:
- 影响因子:0
- 作者:
Liming Wang;Yuejie Chi - 通讯作者:
Yuejie Chi
Regularized blind detection for MIMO communications
MIMO 通信的正则盲检测
- DOI:
10.1109/isit.2010.5513407 - 发表时间:
2010 - 期刊:
- 影响因子:0
- 作者:
Yuejie Chi;Yiyue Wu;A. Calderbank - 通讯作者:
A. Calderbank
Principal subspace estimation for low-rank Toeplitz covariance matrices with binary sensing
具有二元感知的低秩 Toeplitz 协方差矩阵的主子空间估计
- DOI:
10.1109/acssc.2016.7869594 - 发表时间:
2016 - 期刊:
- 影响因子:0
- 作者:
H. Fu;Yuejie Chi - 通讯作者:
Yuejie Chi
Golay complementary waveforms for sparse delay-Doppler radar imaging
用于稀疏延迟多普勒雷达成像的 Golay 互补波形
- DOI:
- 发表时间:
2009 - 期刊:
- 影响因子:0
- 作者:
Yuejie Chi;Robert Calderbank;A. Pezeshki - 通讯作者:
A. Pezeshki
Yuejie Chi的其他文献
{{
item.title }}
{{ item.translation_title }}
- DOI:
{{ item.doi }} - 发表时间:
{{ item.publish_year }} - 期刊:
- 影响因子:{{ item.factor }}
- 作者:
{{ item.authors }} - 通讯作者:
{{ item.author }}
{{ truncateString('Yuejie Chi', 18)}}的其他基金
Federated Optimization over Bandwidth-Limited Heterogeneous Networks
带宽受限异构网络的联合优化
- 批准号:
2318441 - 财政年份:2023
- 资助金额:
$ 40万 - 项目类别:
Standard Grant
Collaborative Research: Towards a Theoretic Foundation for Optimal Deep Graph Learning
协作研究:为最优深度图学习奠定理论基础
- 批准号:
2134080 - 财政年份:2022
- 资助金额:
$ 40万 - 项目类别:
Continuing Grant
NSF Student Travel Grant for the Fifth Conference on Machine Learning and Systems (MLSys 2022)
第五届机器学习和系统会议 (MLSys 2022) 的 NSF 学生旅费补助金
- 批准号:
2219655 - 财政年份:2022
- 资助金额:
$ 40万 - 项目类别:
Standard Grant
Collaborative Research: CIF: Medium: Statistical and Algorithmic Foundations of Efficient Reinforcement Learning
合作研究:CIF:媒介:高效强化学习的统计和算法基础
- 批准号:
2106778 - 财政年份:2021
- 资助金额:
$ 40万 - 项目类别:
Continuing Grant
Taming Nonlinear Inverse Problems: Theory and Algorithms
驯服非线性反问题:理论与算法
- 批准号:
2126634 - 财政年份:2021
- 资助金额:
$ 40万 - 项目类别:
Standard Grant
CIF: Small: Resource-Efficient Statistical Inference in Networked Environments
CIF:小型:网络环境中资源高效的统计推断
- 批准号:
2007911 - 财政年份:2020
- 资助金额:
$ 40万 - 项目类别:
Standard Grant
EAGER-DynamicData: Subspace Learning From Binary Sensing
EAGER-DynamicData:从二进制感知中学习子空间
- 批准号:
1833553 - 财政年份:2018
- 资助金额:
$ 40万 - 项目类别:
Standard Grant
CIF: Small: Inverse Methods for Parametric Mixture Models
CIF:小:参数混合模型的逆方法
- 批准号:
1826519 - 财政年份:2018
- 资助金额:
$ 40万 - 项目类别:
Standard Grant
CIF: Medium: Collaborative Research: Nonconvex Optimization for High-Dimensional Signal Estimation: Theory and Fast Algorithms
CIF:中:协作研究:高维信号估计的非凸优化:理论和快速算法
- 批准号:
1806154 - 财政年份:2018
- 资助金额:
$ 40万 - 项目类别:
Continuing Grant
CAREER: Robust Methods for High-Dimensional Signal Processing under Geometric Constraints
职业:几何约束下高维信号处理的鲁棒方法
- 批准号:
1818571 - 财政年份:2018
- 资助金额:
$ 40万 - 项目类别:
Standard Grant
相似海外基金
Collaborative Research: CIF: Medium: Snapshot Computational Imaging with Metaoptics
合作研究:CIF:Medium:Metaoptics 快照计算成像
- 批准号:
2403122 - 财政年份:2024
- 资助金额:
$ 40万 - 项目类别:
Standard Grant
Collaborative Research: CIF-Medium: Privacy-preserving Machine Learning on Graphs
合作研究:CIF-Medium:图上的隐私保护机器学习
- 批准号:
2402815 - 财政年份:2024
- 资助金额:
$ 40万 - 项目类别:
Standard Grant
Collaborative Research: CIF-Medium: Privacy-preserving Machine Learning on Graphs
合作研究:CIF-Medium:图上的隐私保护机器学习
- 批准号:
2402817 - 财政年份:2024
- 资助金额:
$ 40万 - 项目类别:
Standard Grant
Collaborative Research: CIF-Medium: Privacy-preserving Machine Learning on Graphs
合作研究:CIF-Medium:图上的隐私保护机器学习
- 批准号:
2402816 - 财政年份:2024
- 资助金额:
$ 40万 - 项目类别:
Standard Grant
Collaborative Research: CIF: Medium: Snapshot Computational Imaging with Metaoptics
合作研究:CIF:Medium:Metaoptics 快照计算成像
- 批准号:
2403123 - 财政年份:2024
- 资助金额:
$ 40万 - 项目类别:
Standard Grant
Collaborative Research: CIF: Medium: Fundamental Limits of Cache-aided Multi-user Private Function Retrieval
协作研究:CIF:中:缓存辅助多用户私有函数检索的基本限制
- 批准号:
2312229 - 财政年份:2023
- 资助金额:
$ 40万 - 项目类别:
Continuing Grant
Collaborative Research: CIF: Medium: Statistical and Algorithmic Foundations of Distributionally Robust Policy Learning
合作研究:CIF:媒介:分布式稳健政策学习的统计和算法基础
- 批准号:
2312205 - 财政年份:2023
- 资助金额:
$ 40万 - 项目类别:
Continuing Grant
Collaborative Research: CIF: Medium: Fundamental Limits of Privacy-Enhancing Technologies
合作研究:CIF:中:隐私增强技术的基本限制
- 批准号:
2312666 - 财政年份:2023
- 资助金额:
$ 40万 - 项目类别:
Continuing Grant
Collaborative Research: CIF: Medium: Fundamental Limits of Cache-aided Multi-user Private Function Retrieval
协作研究:CIF:中:缓存辅助多用户私有函数检索的基本限制
- 批准号:
2312228 - 财政年份:2023
- 资助金额:
$ 40万 - 项目类别:
Continuing Grant
Collaborative Research: CIF: Medium: Robust Learning over Graphs
协作研究:CIF:媒介:图上的鲁棒学习
- 批准号:
2312547 - 财政年份:2023
- 资助金额:
$ 40万 - 项目类别:
Continuing Grant