人工智能培训

搜索

深度学习论文:从辅助判别任务转移学习以进行无监督的异常检测(Transfer Learning from an Auxiliary Discriminative

[复制链接]
1449231467 发表于 2019-12-9 14:45:51 | 显示全部楼层 |阅读模式
1449231467 2019-12-9 14:45:51 636 0 显示全部楼层
深度学习论文:从辅助判别任务转移学习以进行无监督的异常检测(Transfer Learning from an Auxiliary Discriminative Task for Unsupervised  Anomaly Detection)从诸如移动网络之类的高维数据中进行无监督的异常检测是一项艰巨的任务。从这样的高维数据研究不同的特征工程方法已经成为该领域的研究重点。这项研究旨在调查通过网络分类学习的特征到无监督异常检测的可传递性。我们建议使用辅助分类任务,通过监督学习从未标记的数据中提取特征,该特征可用于无监督的异常检测。我们通过设计实验来检测来自纽约和台北的移动网络数据中的异常来验证这种方法,并将结果与​​PCA和自动编码器的传统无监督特征学习方法进行比较。我们发现,我们的特征学习方法对两个数据集都具有最佳的异常检测性能,优于其他研究方法。这确立了该方法在特征工程中的效用,可以将其应用于类似性质的其他问题。
Unsupervised anomaly detection from high dimensional data like mobilitynetworks is a challenging task.Study of different approaches of featureengineering from such high dimensional data have been a focus of research inthis field.This study aims to investigate the transferability of featureslearned by network classification to unsupervised anomaly detection.We proposeuse of an auxiliary classification task to extract features from unlabelleddata by supervised learning, which can be used for unsupervised anomalydetection.We validate this approach by designing experiments to detectanomalies in mobility network data from New York and Taipei, and compare theresults to traditional unsupervised feature learning approaches of PCA andautoencoders.We find that our feature learning approach yields best anomalydetection performance for both datasets, outperforming other studiedapproaches.This establishes the utility of this approach to featureengineering, which can be applied to other problems of similar nature.深度学习论文:从辅助判别任务转移学习以进行无监督的异常检测(Transfer Learning from an Auxiliary Discriminative Task for Unsupervised  Anomaly Detection)
URL地址:https://arxiv.org/abs/1912.02864     ----pdf下载地址:https://arxiv.org/pdf/1912.02864    ----深度学习论文:从辅助判别任务转移学习以进行无监督的异常检测(Transfer Learning from an Auxiliary Discriminative Task for Unsupervised  Anomaly Detection)
回复

使用道具 举报

您需要登录后才可以回帖 登录 | 立即注册

本版积分规则 返回列表 发新帖

1449231467当前离线
新手上路

查看:636 | 回复:0

快速回复 返回顶部 返回列表