Ultraconservative Online Algorithms for Multiclass Problems

Crammer, Koby; Singer, Yoram

doi:10.1007/3-540-44581-1_7

Cited by 220 publications

(327 citation statements)

References 7 publications

Supporting

Mentioning

326

Contrasting

Order By: Relevance

“…5, we show the expected relative improvement as a function of the strength of the base classifiers. More specifically, each position t ∈ [1,15] on the horizontal axis shows the result when the initially generated α ij -values are multiplied with t. The plots are representative for all the initial values that we have generated; in total we generated 200 curves and at least 95% of these curves lie in-between the plotted 20 curves. We may formulate the following three observations.…”

Section: Experimental Setup and Resultsmentioning

confidence: 99%

“…The setting of label ranking can be seen as an extension of the conventional setting of classification [13][14][15]11]. Roughly speaking, the former is obtained from the latter through replacing single class labels by complete label rankings.…”

Section: Generalizing the Classification Settingmentioning

confidence: 99%

“…Essentially, this corresponds to a model known as Babington Smith in the statistical literature [25]. The resulting probability distribution over S m defined by (15) can serve as a point of departure for various types of inferences. In the following, we shall focus on classification, i.e., estimating the top-label.…”

Section: Formal Frameworkmentioning

confidence: 99%

“…First, we propose a formal framework in which the aforementioned aggregation problem for pairwise learning can be studied in a convenient way. This framework is based on the setting of label ranking which has recently received attention in the machine learning literature [13][14][15]11]. Second, within this framework, we develop a new aggregation strategy called adaptive voting.…”

Section: Introductionmentioning

confidence: 99%

See 3 more Smart Citations

Combining predictions in pairwise classification: An optimal adaptive voting strategy and its relation to weighted voting

Hüllermeier

Vanderlooy

2010

Pattern Recognition

111

View full text Add to dashboard Cite

Weighted voting is the commonly used strategy for combining predictions in pairwise classification. Even though it shows good classification performance in practice, it is often criticized for lacking a sound theoretical justification. In this paper, we study the problem of combining predictions within a formal framework of label ranking and, under some model assumptions, derive a generalized voting strategy in which predictions are properly adapted according to the strengths of the corresponding base classifiers. We call this strategy adaptive voting and show that it is optimal in the sense of yielding a MAP prediction of the class label of a test instance. Moreover, we offer a theoretical justification for weighted voting by showing that it yields a good approximation of the optimal adaptive voting prediction. This result is further corroborated by empirical evidence from experiments with real and synthetic data sets showing that, even though adaptive voting is sometimes able to achieve consistent improvements, weighted voting is in general quite competitive, all the more in cases where the aforementioned model assumptions underlying adaptive voting are not met. In this sense, weighted voting appears to be a more robust aggregation strategy.

show abstract

Section: Experimental Setup and Resultsmentioning

confidence: 99%

Section: Generalizing the Classification Settingmentioning

confidence: 99%

Section: Formal Frameworkmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

See 2 more Smart Citations

Combining predictions in pairwise classification: An optimal adaptive voting strategy and its relation to weighted voting

Hüllermeier

Vanderlooy

2010

Pattern Recognition

111

View full text Add to dashboard Cite

show abstract

“…This algorithm, called Winnow, has the advantage of reducing the weights of irrelevant input features quickly toward zero, making the algorithm more effective when there are a large number of features but only few of them are important to the classification task. A number of variations of the Winnow algorithm have since been studied, both in terms of provable error bounds (Littlestone & Warmuth, 1994;Kivinen & Warmuth, 1997;Crammer & Singer, 2001;Mesterharm, 2002) and empirical performance on natural language processing tasks such as document categorization (Dagan, Karov, & Roth, 1997) and spelling correction (Golding & Roth, 1999). However, the error rates for both the additive and multiplicative algorithms are significantly higher when feedback is limited, especially for the important case of k = 1 (simple confirmation), as illustrated by the empirical results presented in this paper.…”

Section: Introductionmentioning

confidence: 99%

Online Multiclass Learning with k-Way Limited Feedback and an Application to Utterance Classification

Alshawi

2005

Mach Learn

View full text Add to dashboard Cite

Abstract. This paper introduces a setting for multiclass online learning with limited feedback and its application to utterance classification. In this learning setting, a parameter k limits the number of choices presented for selection by the environment (e.g. by the user in the case of an interactive spoken system) during each trial of the online learning sequence. New versions of standard additive and multiplicative weight update algorithms for online learning are presented that are more suited to the limited feedback setting, while sharing the efficiency advantages of the standard ones. The algorithms are evaluated on an utterance classification task in two domains. In this utterance classification task, no training material for the domain is provided (for training the speech recognizer or classifier) prior to the start of online learning. We present experiments on the effect of varying k and the weight update algorithms on the learning curve for online utterance classification. In these experiments, the new online learning algorithms improve classification accuracy compared with the standard ones. The methods presented are directly relevant to applications such as building call routing systems that adapt from feedback rather than being trained in batch mode.

show abstract

Automatic Speech and Speaker Recognition

Keshet¹,

Bengio²

2009

View full text Add to dashboard Cite

Designations used by companies to distinguish their products are often claimed as trademarks. All brand names and product names used in this book are trade names, service marks, trademarks or registered trademarks of their respective owners. The publisher is not associated with any product or vendor mentioned in this book. This publication is designed to provide accurate and authoritative information in regard to the subject matter covered. It is sold on the understanding that the publisher is not engaged in rendering professional services. If professional advice or other expert assistance is required, the services of a competent professional should be sought.

show abstract

Ultraconservative Online Algorithms for Multiclass Problems

Cited by 220 publications

References 7 publications

Combining predictions in pairwise classification: An optimal adaptive voting strategy and its relation to weighted voting

Combining predictions in pairwise classification: An optimal adaptive voting strategy and its relation to weighted voting

Online Multiclass Learning with k-Way Limited Feedback and an Application to Utterance Classification

Automatic Speech and Speaker Recognition

Contact Info

Product

Resources

About