Detecting Change in Data Streams

By Daniel Kifer, Shai Ben-David, Johannes Gehrke. 2004

964 citations recorded by the source corpus.

Subfields: Artificial Intelligence, Anomaly & Outlier Detection, Signal Processing

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Artificial Intelligence

In Anomaly & Outlier Detection

In Signal Processing

Citations in the corpus

Multiple-Source Adaptation with Domain Classifiers (2020, background): “The general theoretical problem of adaptation from a single domain to a target domain has been studied in a series of publications in the last two decades or so [Kifer et al., 2004, Ben-David et al., 2007, Blitzer et al., 2008, Mansour et al., 2009b, Cortes and Mohri, 2014, Cortes et al., 2019b].”
A Theory of Multiple-Source Adaptation with Limited Target Labeled Data (2020, methodology): “It was shown by Mansour et al. (2009d) (see also Cortes and Mohri (2011)) that the discrepancy is the appropriate divergence between distributions in adaptation.”
The Variational Fair Autoencoder (2015, methodology): “However, there is a subtle difference between the approach of Kingma et al. (2014) and our model.”
Estimating Counterfactual Treatment Outcomes over Time Through Adversarially Balanced Representations (2020, neutral): “Kifer et al. (2004)”
CoSCL: Cooperation of Small Continual Learners is Stronger than a Big One (2022, neutral): “4 of [19] and Lemma 1 of [4]) Let Θ be a cover of a parameter space with VC dimension d.”
A Kernel Method for the Two-Sample-Problem (2006, background): “can be used for the purpose of distribution comparison. For instance, Ben-David et al. (2007, Section 4) use the error of a hyperplane classiﬁer to approximate the A-distance between distributions of Kifer et al. (2004). 2.5 Examples of Non-RKHS Function Classes Otherfunction spaces F inspired by th”
Simultaneous Deep Transfer Across Domains and Tasks (2015, background): “In fact, the size of a domain shift is often measured by the distance between the source and target subspace representations [5, 11, 21, 25, 27].”
Censoring Representations with an Adversary (2015, methodology): “In this appendix we describe the notion of a H -divergence, as developed in Kifer et al. (2004) and Ben-David et al. (2007).”
Deep Domain Confusion: Maximizing for Domain Invariance (2014, background): “In fact, the size of a domain shift is often measured by the distance between the source and target subspace representations [6, 13, 23, 26, 28].”
An embarrassingly simple approach to zero-shot learning (2015, neutral): “• dH D,D′ is known as the A-distance between distributions D and D′ over the subsets defined inH [20]: dH D,D′ = 2sup h∈H |PD(h) − PD′(h)| ,”
Mind the Gap: Assessing Temporal Generalization in Neural Language Models (2021, background): “Detecting changes in data streams, also known as concept drift, has a long history in machine learning (Widmer and Kubat, 1996; Kifer et al., 2004; Baena-Garcıa et al., 2006; Dries and Rückert, 2009; Lu et al., 2019).”
Domain-Adversarial Neural Networks (2014, methodology): “Deﬁnition 1 (Ben-David et al. (2006; 2010); Kifer et al. (2004)) .”
On the Ability of Neural Nets to Express Distributions (2017, background): “One way to model a discriminator is as a function f in a certain class F that maximizes the diﬀerence between E f on the real distribution µ and the generated distribution ν , This is called the maximal mean discrepancy ([KBG04], [DRG15]).”
Learning with Feature and Distribution Evolvable Streams (2020, background): “A series of pioneering works propose different discrepancy measures to estimate the discrepancy between two distributions, including the A-discrepancy (Kifer et al., 2004),H∆H-discrepancy (Ben-David et al.”
Change-Point Detection in Time-Series Data by Relative Density-Ratio Estimation (2012, background): “The problem of discovering time points at which properties of time-series data change is attracting a lot of attention in the data mining community [5, 2, 12, 18].”
UDALM: Unsupervised Domain Adaptation through Language Modeling (2021, background): “Then for any h ∈ H : where d H ∆ H ( D S , D T ) is the H ∆ H -divergence (Kifer et al., 2004) between two domains, that is a measure of distance between domains that can be estimated from ﬁnite samples.”
Learning under Concept Drift: A Review (2019, methodology): “According to the literature, the first formal treatment of change detection in data streams was proposed by [37].”