Network Online Change Point Localization

点(几何) 计算机科学 人工智能 数学 几何学
作者
Yi Yu,Oscar Hernán Madrid Padilla,Daren Wang,Alessandro Rinaldo
出处
期刊:SIAM journal on mathematics of data science [Society for Industrial and Applied Mathematics]
卷期号:6 (1): 176-198
标识
DOI:10.1137/22m1529816
摘要

.We study the following online network change point detection settings: A time series of independent, possibly sparse Bernoulli networks whose distributions might change at an unknown time are observed in a sequential manner, and at each time point, a determination has to be made on whether a change has taken place in the near past. The goal is to detect the change point event (if any has occurred) as quickly as possible, subject to prespecified constraints on the probability or number of false alarms. We propose a CUSUM-based procedure and derive two high-probability upper bounds on its detection delay, i.e., detection delay \(\{ \gtrsim \log(1/\alpha)\frac{1}{\kappa_0^2 n \rho}\); \(\lesssim \log(\Delta/\alpha) \frac{r}{\kappa_0^2 n \rho}\), under a low-rank assumption; \(\lesssim \log(\Delta/\alpha) \frac{\max\{r^2/n, \log(r), 1\}}{\kappa_0^2 n \rho}\), under a block-constancy assumption, where \(\kappa_0, n, \rho, r\), and \(\alpha\) are the normalized jump size, network size, entrywise sparsity, rank sparsity, and overall type I error upper bound. All the model parameters are allowed to vary as \(\Delta\), the unknown change point, diverges. We further establish a minimax lower bound on the detection delay. Under the low-rank assumption and when the rank is of constant order or under the block-constancy assumption when the number of blocks \(r \lesssim \sqrt{n}\), we obtain minimax rates. The above upper bounds are achieved by novel procedures proposed in this paper, designed for quick detection under two different forms of type I error control. The first is based on controlling the overall probability of a false alarm when there are no change points, and the second is based on specifying a lower bound on the expected time of the first false alarm. Extensive experiments show that under different scenarios and the aforementioned forms of type I error control, our proposed approaches well outperform state-of-the-art methods.Keywordsdynamic networksonline change point detectionminimax optimalityMSC codes62C2062L99
最长约 10秒,即可获得该文献文件

科研通智能强力驱动
Strongly Powered by AbleSci AI
更新
PDF的下载单位、IP信息已删除 (2025-6-4)

科研通是完全免费的文献互助平台,具备全网最快的应助速度,最高的求助完成率。 对每一个文献求助,科研通都将尽心尽力,给求助人一个满意的交代。
实时播报
平常星星完成签到 ,获得积分10
刚刚
量子星尘发布了新的文献求助10
刚刚
一个美女完成签到,获得积分10
刚刚
熊博士完成签到,获得积分10
2秒前
7890733完成签到,获得积分10
3秒前
欣喜的尔曼完成签到,获得积分20
3秒前
王叮叮发布了新的文献求助20
4秒前
kunkun完成签到,获得积分10
5秒前
6秒前
crrr完成签到,获得积分10
6秒前
xiaoguan完成签到,获得积分10
8秒前
Owen应助欣喜的尔曼采纳,获得10
10秒前
独特的翠芙完成签到,获得积分10
10秒前
11秒前
12秒前
13秒前
姜惠完成签到 ,获得积分10
13秒前
小脸红扑扑完成签到 ,获得积分10
14秒前
奋斗的凡完成签到 ,获得积分10
14秒前
15秒前
zhutier完成签到,获得积分10
16秒前
wrr完成签到,获得积分10
16秒前
WxChen完成签到,获得积分10
16秒前
开朗艳一完成签到,获得积分10
18秒前
Wonder完成签到,获得积分10
19秒前
yang完成签到,获得积分10
21秒前
123123完成签到 ,获得积分10
22秒前
温暖宛筠完成签到,获得积分10
22秒前
小欣6116完成签到,获得积分10
23秒前
请叫我风吹麦浪应助冬月采纳,获得10
23秒前
LIUYONG发布了新的文献求助10
24秒前
24秒前
肖雪依完成签到,获得积分10
24秒前
影子完成签到,获得积分10
25秒前
26秒前
晨珂完成签到,获得积分10
26秒前
Florencia发布了新的文献求助10
28秒前
xiezhuochun发布了新的文献求助10
29秒前
29秒前
同瓜不同命完成签到,获得积分10
31秒前
高分求助中
【提示信息,请勿应助】关于scihub 10000
Les Mantodea de Guyane: Insecta, Polyneoptera [The Mantids of French Guiana] 3000
徐淮辽南地区新元古代叠层石及生物地层 3000
The Mother of All Tableaux: Order, Equivalence, and Geometry in the Large-scale Structure of Optimality Theory 3000
Handbook of Industrial Diamonds.Vol2 1100
Global Eyelash Assessment scale (GEA) 1000
Picture Books with Same-sex Parented Families: Unintentional Censorship 550
热门求助领域 (近24小时)
化学 材料科学 医学 生物 工程类 有机化学 生物化学 物理 内科学 纳米技术 计算机科学 化学工程 复合材料 遗传学 基因 物理化学 催化作用 冶金 细胞生物学 免疫学
热门帖子
关注 科研通微信公众号,转发送积分 4038303
求助须知:如何正确求助?哪些是违规求助? 3576013
关于积分的说明 11374210
捐赠科研通 3305780
什么是DOI,文献DOI怎么找? 1819322
邀请新用户注册赠送积分活动 892672
科研通“疑难数据库(出版商)”最低求助积分说明 815029