Dynamic Weighted Majority: An Ensemble Method for Drifting Concepts

By J. Kolter, M. Maloof. 2007

790 citations recorded by the source corpus.

Subfields: Artificial Intelligence, Information Systems

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 Information Systems

Citations in the corpus

From amateurs to connoisseurs: modeling the evolution of user expertise through online reviews (2013, background): “The emergence of new products may cause users to change their focus [19]; older movies may be treated more favorably once they are considered ‘classics’ [20]; and users may be inﬂuenced by general trends in the community, or by members of their social networks [26].”
Memory Efficient Experience Replay for Streaming Learning (2018, neutral): “[37], [38], [39], [56], [63] or ensemble methods [7], [9], [27], [44], [46], [51], [54], [61], [70], [72].”
Continual Prototype Evolution: Learning Online from Non-Stationary Data Streams (2020, background): “However, online learning typically makes the rigorous assumption of learning from an iid data stream [41] or does not focus on preventing catastrophic forgetting [34, 26], whereas data incremental learning disregards the iid assumption and additionally generalizes beyond the online learning horizon ”
Learning in Nonstationary Environments: A Survey (2015, background): “Dynamic weighted majority (DWM) [90], is an extension of the weighted majority algorithm (WM) [26] that extends WM to data streams with concept drift, and uses an updated period to add/remove classifiers.”
Reacting to Different Types of Concept Drift: The Accuracy Updated Ensemble Algorithm (2014, methodology): “The Naive Bayes classifier has no drift reaction mechanism; the drift detector in ACE is not triggered, thereby causing poor reaction to gradual changes, while HOT and DWM appear to not be pruning outdated data, with HOT additionally using too much memory.”
Towards Understanding Ensemble, Knowledge Distillation and Self-Distillation in Deep Learning (2020, background): “On the theory side, there are quite a few works focusing on understanding the superior performance of ensemble from a principled perspective [12, 14, 29, 31–34, 37, 44, 47–49, 52, 72, 72, 75, 76].”
Detecting concept drift in data streams using model explanation (2018, background): “Informally, it can be deﬁned as a “situation in which the statistical properties of the target concept change over time” (Wang et al., 2011) or “the event when a set of examples has legitimate class labels at one time and has diﬀerent legitimate labels at time” (Kolter & Maloof, 2007).”
Learning under Concept Drift: an Overview (2010, background): “There is a number of ensembles for concept drift, where the ideas are not speciﬁc to particular type of base learners (although some studies are limited to testing one base learner) [83, 141, 142, 150, 148, 72, 137, 7, 115, 151, 44, 163, 131].”
Exponentially weighted moving average charts for detecting concept drift (2012, background): “The ﬁrst is to design classiﬁers which automatically adapt their behavior to stay up-to-date with the stream dynamics [24, 14, 16].”
Ensemble learning for data stream analysis: A survey (2017, methodology): “One of the most well known approaches under this category is Dynamic Weighted Majority (DWM) [101], proposed by Kolter and Maloof.”
On evaluating stream learning algorithms (2012, background): “Other direct applications are tracking the best expert algorithms (Herbster and Warmuth, 1998) and dynamic weighted majority algorithms (Kolter and Maloof, 2007) that continuously track the evolution of the performance of its components.”
Minimax Classification under Concept Drift with Multidimensional Adaptation and Performance Guarantees (2022, neutral): “…risk minimization algorithm (NORMA) (Kivinen et al., 2004), Nystr¨om online gradient descent (NOGD), Fourier online gradient descent (FOGD) (Lu et al., 2016), λ - perceptron (Pavlidis et al., 2011), dynamic weighted majority (DWM) (Kolter & Maloof, 2007), and Forgetron (Dekel et al., 2006).”
Incremental Learning of Concept Drift in Nonstationary Environments (2011, background): “The cyclical nature of drift is a phenomenon that can be observed in many real-world applications such as climate or electricity demand modeling.”
Incremental on-line learning: A review and comparison of state of the art algorithms (2018, background): “Various algorithms have been published explicitly tackling this challenge [74, 75, 76].”
Adaptive Machine Learning for Credit Card Fraud Detection (2015, background): “If the weight of a classifier drop below a certain threshold, then the classifier is usually replaced, see for examples SEA [160] and DWM [162].”
Credit Card Fraud Detection: A Realistic Modeling and a Novel Learning Strategy (2018, methodology): “Ensemble methods [23], [30], [43], [61], [72] and classifiers trained over a sliding window of the recent supervised samples (like STAGGER [57] and FLORA [67]) are probably the most extensively investigated passive solutions.”