Soft Margins for AdaBoost

By Gunnar Rätsch, Takashi Onoda, K. Müller. 2001

1,384 citations recorded by the source corpus.

Subfields: Artificial Intelligence, Computer Vision and Pattern Recognition

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 Computer Vision and Pattern Recognition

Citations in the corpus

Learning kernel logistic regression in the presence of class label noise (2014, methodology): “We use 13 UCI benchmark data sets [35] in our controlled experiments.”
Ensemble learning: A survey (2018, background): “There are several notable variations of AdaBoost.”
AdaBoost with SVM-based component classifiers (2008, methodology): “In this paper, we shall show that AdaBoost incorporating properly designed RBFSVM (SVM with the RBF kernel) component classifiers, which we call AdaBoostSVM, can perform as well as SVM.”
Coulomb Classifiers: Generalizing Support Vector Machines via an Analogy to Electrostatic Systems (2002, methodology): “The search for hyperparameter was not as intensive as in [7].”
On Equivalence Relationships Between Classification and Ranking Algorithms (2011, methodology): “AdaBoost was originally shown to be a coordinate descent algorithm by Breiman (1997), Friedman et al. (2000), Rätsch et al. (2001), Duffy and Helmbold (1999) and Mason et al. (2000).”
Feature selection via sensitivity analysis of SVM probabilistic outputs (2008, methodology): “datasets obtained from Ratsch (2005) , used also by Mika et al. (1999), Ratsch et”
A nested heuristic for parameter tuning in Support Vector Machines (2014, methodology): “[44] give 100 partitions of each data set into a training sample and a testing sample.”
Preventing Over-Fitting during Model Selection via Bayesian Regularisation of the Hyper-Parameters (2007, methodology): “…and without Bayesian regularisation of the model selection criterion, in this case the PRESS statistic (Allen, 1974), and Gaussian process classiﬁers over thirteen benchmark data sets (R¨atsch et al., 2001), using both standard radial basis function and automatic relevance determination kernels.”
Edited AdaBoost by weighted kNN (2010, methodology): “For example, to use BrownBoost one needs to pre-specify an upper bound ð1=2Þ g on the error of a weak learner and a ‘target’ error e40 should be given a priori [7].”
Boosting bit rates in noninvasive EEG single-trial classifications by feature combination and multiclass paradigms (2004, methodology): “This procedure was tested in several settings without any significant bias on the estimation of the GE, cf., [ 25 ].”
Efficient leave-one-out cross-validation of kernel fisher discriminant classifiers (2003, methodology): “…cross-validation procedure can then be written in closed form as The i th element of r ( i ) is given by kernel Fisher discriminant (KFD) [1], radial basis function (RBF) [35], Ad-aBoost (AB) [34] and regularised AdaBoost (AB R ) [32] classiﬂers on 13 different benchmark datasets [33].”
Different Paradigms for Choosing Sequential Reweighting Algorithms (2004, methodology): “2 This analogy also has led other authors to adapt the “soft margin” SVM principle to the Boosting case to avoid overfitting (Rätsch et al., 2001).”
MCMC for Variationally Sparse Gaussian Processes (2015, methodology): “We first use the image dataset [29] to investigate the benefits of the approach over a Gaussian approximation, and to investigate the effect of changing the number of inducing points, as well as optimizing the inducing points under the Gaussian approximation.”
Evolutionary tuning of multiple SVM parameters (2005, methodology): “For the evaluation of our hyperparameter optimization method we used the common medical benchmark datasets Breast-Cancer , Diabetes , Heart , and Thyroid with input dimension n equal to 9, 8, 13, and 5, respectively, preprocessed and partitioned as in [8].”