Data-driven selection of a convex loss function via shape-constrained estimation
通过形状约束估计来数据驱动选择凸损失函数
基本信息
- 批准号:2311299
- 负责人:
- 金额:$ 20万
- 依托单位:
- 依托单位国家:美国
- 项目类别:Standard Grant
- 财政年份:2023
- 资助国家:美国
- 起止时间:2023-07-01 至 2026-06-30
- 项目状态:未结题
- 来源:
- 关键词:
项目摘要
This research project focuses on the notion of loss functions, which is central to machine learning and statistics. Loss functions measure the difference between the output predicted by the model and the actual output, and they typically satisfy a property called convexity so that they can be easily optimized. Loss functions quantify how accurate a model is at describing the data and therefore, almost all predictive models are computed by learning model parameters which minimize a given loss function. Choosing a good loss function is vitally important; a good loss function not only improves our predictions, but also allows us to build tighter confidence intervals, and gives us greater robustness to outliers. Although there are general guidelines for choosing a suitable loss function, these guidelines are qualitative and imprecise; most people still default to a few standard choices such as the square error loss. The goal of this project is to develop methods to estimate an optimal convex loss function from the data at hand. We will design, implement, and test algorithms that practitioners can use to automatically obtain loss functions specifically optimized to their dataset, which will allow the practitioners to make better predictive models. Successful execution of this project will have far-reaching effects on standard practices in data science. This project will be deeply integrated with the planned educational components at both the undergraduate and graduate levels.The first component of the project will look at linear regression and show that we can learn a data-driven convex loss function by approximating the unknown noise distribution with a log-concave density in a distributional distance known as the Fisher divergence. The proposed approach is computationally simple and, in settings where the noise is non-Gaussian, significantly improves upon the traditional squared error loss in estimation accuracy, inference quality, and robustness. The second component of the project will extend the idea to the setting of multi-task regression where the response is multivariate. The third component of the project will analyze the theoretical properties of score matching–the statistical method that underpins the first two components on convex loss estimation as well as being of fundamental importance in various other applications in statistical learning.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.
损失函数衡量模型预测的输出之间的差异,它们通常满足称为凸度的属性,以便可以轻松地优化它们。损失函数量化了模型在描述数据方面的准确程度,因此,几乎所有预测模型都是通过学习模型参数计算的,该参数可最大程度地减少给定的损失函数。选择良好的损失功能至关重要;良好的损失功能不仅可以改善我们的预测,而且还使我们能够建立更紧密的置信区间,并使我们对异常值的鲁棒性更大。尽管有选择合适的损失函数的一般准则,但这些准则是定性和坚不可摧的。大多数人仍然默认不需要一些标准选择,例如正方形误差丢失。该项目的目的是开发方法,以估算手上数据的最佳凸损失函数。我们将设计,实施和测试算法,实践者可以使用这些算法来自动获取针对其数据集进行特殊优化的损失功能,这将使实践者能够做出更好的预测模型。该项目的成功执行将对数据科学的标准实践产生深远的影响。该项目将与本科和研究生级别的计划的教育组成部分深入融合。该项目的第一部分将介绍线性回归,并表明我们可以通过在被称为Fisher evergence的分布距离中以对数 - 连接密度的未知噪声分布来近似数据驱动的凸损失函数。所提出的方法在计算上很简单,并且在噪声是非高斯的设置中,在估计准确性,推理质量和鲁棒性方面的传统平方误差损失会大大改善。项目的第二个组成部分将把想法扩展到响应为多变量的多任务回归的设置。该项目的第三个组成部分将分析得分匹配的理论属性 - 统计方法是统计损失估计的前两个组成部分,并且在统计学习中其他各种应用中都具有根本重要性。该奖项反映了NSF的立法任务,并被认为是通过基金会的智力优点和广泛的Cravitia和Broaditia的评估值得评估,并且值得通过评估。
项目成果
期刊论文数量(0)
专著数量(0)
科研奖励数量(0)
会议论文数量(0)
专利数量(0)
数据更新时间:{{ journalArticles.updateTime }}
{{
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 }}
Min Xu其他文献
Inpainting of Continuous Frames of Old Movies Based on Deep Neural Network
基于深度神经网络的老电影连续帧修复
- DOI:
10.1109/icalip.2018.8455233 - 发表时间:
2018 - 期刊:
- 影响因子:0
- 作者:
Chang Li;Youdong Ding;Ting Yu;Min Xu;Qianqian Zhang - 通讯作者:
Qianqian Zhang
Numerical Investigations for the Effect of Slender Body on Dynamic Rolling Characteristics of a 80°/60° Double Delta Wing
细长体对80°/60°双三角翼动态滚转特性影响的数值研究
- DOI:
10.4028/www.scientific.net/amm.444-445.286 - 发表时间:
2013 - 期刊:
- 影响因子:0
- 作者:
Bing Han;Min Xu;X. Pei;Xiaomin An - 通讯作者:
Xiaomin An
Coherent backscattering of polarized light for tissue diagnostics: an electric field Monte Carlo study
用于组织诊断的偏振光相干反向散射:电场蒙特卡罗研究
- DOI:
- 发表时间:
2008 - 期刊:
- 影响因子:0
- 作者:
Min Xu - 通讯作者:
Min Xu
Local container drayage problem with improved truck platooning operations
通过改进卡车队列操作解决当地集装箱拖运问题
- DOI:
10.1016/j.tre.2022.102992 - 发表时间:
2023-01 - 期刊:
- 影响因子:0
- 作者:
Xiaoyuan Yan;Min Xu;Chi Xie - 通讯作者:
Chi Xie
Spatial variations of the effective elastic thickness and internal load fraction in the Cascadia subduction zone
卡斯卡迪亚俯冲带有效弹性厚度和内部载荷分数的空间变化
- DOI:
10.1093/gji/ggab495 - 发表时间:
2021-12 - 期刊:
- 影响因子:2.8
- 作者:
Chuanhai Yu;Min Xu;Jon F Kirby;Xiaobin Shi;Alberto Jiménez-Díaz - 通讯作者:
Alberto Jiménez-Díaz
Min Xu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('Min Xu', 18)}}的其他基金
CAREER: Cryo-electron tomography derived multiscale integrative modeling of subcellular organization
职业:冷冻电子断层扫描衍生的亚细胞组织多尺度综合模型
- 批准号:
2238093 - 财政年份:2023
- 资助金额:
$ 20万 - 项目类别:
Continuing Grant
Collaborative Research: III: Medium: Systematic De Novo Identification of Macromolecular Complexes in Cryo-Electron Tomography Images
合作研究:III:介质:冷冻电子断层扫描图像中大分子复合物的系统从头识别
- 批准号:
2211597 - 财政年份:2022
- 资助金额:
$ 20万 - 项目类别:
Standard Grant
Inferring the Past on Markovian Models of Networks
根据马尔可夫网络模型推断过去
- 批准号:
2113671 - 财政年份:2021
- 资助金额:
$ 20万 - 项目类别:
Continuing Grant
IIBR Informatics: Reducing the training data annotation cost for learning-based macromolecule identification in cellular electron cryo-tomography
IIBR 信息学:降低细胞电子冷冻断层扫描中基于学习的大分子识别的训练数据注释成本
- 批准号:
1949629 - 财政年份:2020
- 资助金额:
$ 20万 - 项目类别:
Standard Grant
III: Small: Improving automation and speed of macromolecule recognition and localization in cryo-electron tomography using unsupervised deep learning
III:小:使用无监督深度学习提高冷冻电子断层扫描中大分子识别和定位的自动化程度和速度
- 批准号:
2007595 - 财政年份:2020
- 资助金额:
$ 20万 - 项目类别:
Standard Grant
I-Corps: Chemometric fluorescence microscopic imaging and virtual staining for rapid label-free histopathology
I-Corps:化学计量荧光显微成像和虚拟染色,用于快速无标记组织病理学
- 批准号:
2017396 - 财政年份:2020
- 资助金额:
$ 20万 - 项目类别:
Standard Grant
RUI: Cell Growth Laws and Quantitative Microscopy for Cancer Aggressiveness Imaging
RUI:细胞生长规律和癌症侵袭性成像的定量显微镜
- 批准号:
1920617 - 财政年份:2018
- 资助金额:
$ 20万 - 项目类别:
Standard Grant
RUI: Cell Growth Laws and Quantitative Microscopy for Cancer Aggressiveness Imaging
RUI:细胞生长规律和癌症侵袭性成像的定量显微镜
- 批准号:
1607664 - 财政年份:2017
- 资助金额:
$ 20万 - 项目类别:
Standard Grant
相似国自然基金
结合领域知识的数据驱动的消费选择行为研究及零售决策优化
- 批准号:
- 批准年份:2022
- 资助金额:30 万元
- 项目类别:青年科学基金项目
数据驱动下产地型农产品市场平台:机制设计、策略选择与价格决策
- 批准号:72261023
- 批准年份:2022
- 资助金额:28 万元
- 项目类别:地区科学基金项目
结合领域知识的数据驱动的消费选择行为研究及零售决策优化
- 批准号:72201257
- 批准年份:2022
- 资助金额:30.00 万元
- 项目类别:青年科学基金项目
知识驱动下的高维数据进化多任务特征选择方法研究
- 批准号:
- 批准年份:2022
- 资助金额:30 万元
- 项目类别:青年科学基金项目
知识驱动下的高维数据进化多任务特征选择方法研究
- 批准号:62206255
- 批准年份:2022
- 资助金额:30.00 万元
- 项目类别:青年科学基金项目
相似海外基金
"DNMT and TET1 reprogramming as a targetable mechanism of resistance in advanced prostate cancer"
“DNMT 和 TET1 重编程作为晚期前列腺癌的靶向耐药机制”
- 批准号:
10681632 - 财政年份:2023
- 资助金额:
$ 20万 - 项目类别:
Characterizing the immune infiltrate in muscle-invasive urothelial carcinoma
肌层浸润性尿路上皮癌免疫浸润的特征
- 批准号:
10738992 - 财政年份:2023
- 资助金额:
$ 20万 - 项目类别:
Novel synthetic lethality strategy for TP53 mutant colorectal cancer
TP53突变结直肠癌的新型合成致死策略
- 批准号:
10718572 - 财政年份:2023
- 资助金额:
$ 20万 - 项目类别:
A TCF1:Glucocorticoid regulatory circuit controls IL-23-driven Th17 pathogenicity
TCF1:糖皮质激素调节电路控制 IL-23 驱动的 Th17 致病性
- 批准号:
10635984 - 财政年份:2023
- 资助金额:
$ 20万 - 项目类别:
A Modular Framework for Data-Driven Neurogenetics to Predict Complex and Multidimensional Autistic Phenotypes
数据驱动神经遗传学预测复杂和多维自闭症表型的模块化框架
- 批准号:
10826595 - 财政年份:2023
- 资助金额:
$ 20万 - 项目类别: