.png){kind=icon}
返回
Distributed pairwise algorithms with gradient descent methods
.png){kind=icon}
PRE
AI
DOI：10.1016/j.neucom.2019.01.007.png)
.png){kind=icon}
原文求助.png){kind=icon}
分享.png){kind=icon}
收藏
摘要
En 中文
Pairwise algorithms refer to a learning problem with loss functions depending on pairs of examples. There has been remarkable work on analyzing their generalization properties in batch and online settings such as algorithmic stabilities, robustness or regularization. This paper is concerned with distributed pairwise algorithms for dealing with big data, based on a divide-and-conquer strategy. We show that the global estimator of the distributed pairwise algorithm is as good as that of the classical algorithm processing the whole data on a single machine. We present the optimal convergence rate for the distributed pairwise algorithm and provide a theoretical upper bound for the number of local machines under which the optimal rate is retained. Our analysis is achieved by the integral operator decomposition and distributed U-statistics. (C) 2019 Elsevier B.V. All rights reserved.
Keyword:
Distributed method
Pairwise algorithms
Gradient descent
Reproducing kernel Hilbert spaces
AI总结
对已上传原文的论文进行重点信息的提取，主要内容包括：简要概述、研究摘要、背景介绍、关键亮点、图文解析、展望与总结。
期刊
IF:
6.5
论文数:
2.5W
被引数:
6.5W
机构
引用论文
THE MEANING AND USE OF THE AREA UNDER A RECEIVER OPERATING CHARACTERISTIC (ROC) CURVE受试者工作特征 (ROC) 曲线下面积的含义和用途.png)
RADIOLOGY
RADIOLOGYIF15.2
PREAI
分享收藏Simulating groundstate and dynamical quantum phase transitions on a superconducting quantum computer
OAAI
IF0
OAAI分享收藏The use of the area under the roc curve in the evaluation of machine learning algorithmsroc曲线下面积在机器学习算法评价中的应用PATTERN RECOGNITION
OAAI
PATTERN RECOGNITIONIF7.6
OAAI分享收藏
查看更多内容