Support vector machines
By Ingo Steinwart, Andreas Christmann. 2008
8,507 citations recorded by the source corpus.
Support vector machines (SVMs) are a family of machine learning methods, originally introduced for the problem of classification and later generalized to various other situations. They are based on principles of statistical learning theory and convex optimization, and are currently used in various domains of application, including bioinformatics, text categorization, and computer vision. Copyright © 2009 John Wiley & Sons, Inc.
Subfields: Artificial Intelligence, Computer Vision and Pattern Recognition, Molecular Biology, Statistical Learning Theory
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
- Yoshua Bengio
- Sergey Levine
- Mohit Bansal
- Percy Liang
- Bernhard Schölkopf
- Quoc V. Le
- Jianfeng Gao
- Ruslan Salakhutdinov
In Computer Vision and Pattern Recognition
- Luc Van Gool
- Dacheng Tao
- Trevor Darrell
- Xiaogang Wang
- Ming-Hsuan Yang
- Jitendra Malik
- Andrew Zisserman
- Li Fei-Fei
In Molecular Biology
- Arthur Gretton
- Kenji Fukumizu
- Bernhard Schölkopf
- Bharath K. Sriperumbudur
- Giuseppe Jurman
- Alex Smola
- Lorenzo Rosasco
- Ingo Steinwart
In Statistical Learning Theory
- Arthur Gretton
- Bharath K. Sriperumbudur
- Bernhard Schölkopf
- Kenji Fukumizu
- Francis R. Bach
- Krikamol Muandet
- Dino Sejdinovic
- Alex Smola
Citations in the corpus
- Visual Genome: Connecting Language and Vision Using Crowdsourced Dense Image Annotations (2016, methodology): “We use the fine-tuned features from the network and train 100 individual SVMs (Hearst et al. 1998) to predict each attribute.”
- Toward Causal Representation Learning (2021, background): “Such algorithms do exist, for instance, nearest neighbor classifiers, support vector machines, and neural networks [67], [221], [239], [257].”
- Kernel Mean Embedding of Distributions: A Review and Beyonds (2016, background): “= x] is a constant function of x, which is not contained in the Gaussian RKHS (Steinwart and Christmann 2008; Corollary 4.44).”
- Dataset Distillation (2018, background): “For example, solutions to many classical linear learning algorithms, e.g., Perceptron (Rosenblatt, 1957) and SVMs (Hearst et al., 1998), are weighted sums of a subset of training examples, which can be viewed as core-sets.”
- Stronger data poisoning attacks break data sanitization defenses (2018, background): “Many ways of choosing which points to flip have been proposed (see Xiao et al., 2015 for a review). Here, we consider the alfa attack from Xiao et al. (2012). alfa seeks to add points that have a high loss under the original (clean) model ∗ def = argmin L( ;Dc) but a low loss under the final (poison”
- Towards Causal Representation Learning (2021, background): “Such algorithms do exist, for instance, nearest neighbor classifiers, support vector machines, and neural networks [257, 217, 239, 66].”
- Causality for Machine Learning (2019, background): “Such algorithms do exist, for instance nearest neighbor classifiers and Support Vector Machines (Vapnik, 1998; Schölkopf and Smola, 2002; Steinwart and Christmann, 2008).”
- Consistent Kernel Mean Estimation for Functions of Random Variables (2016, methodology): “We choose a product kernel kxy ( (x1, y1), (x2, y2) ) = kx(x1, x2)ky(y1, y2), so the corresponding RKHS is a tensor productHkxy = Hkx ⊗Hky (Steinwart and Christmann, 2008, Lemma 4.6) and the mean embedding of the product random variable (X,Y ) is a tensor product of their marginal mean embeddings…”
- Interpretable Distribution Features with Maximum Testing Power (2016, methodology): “(17)]: Hence, applying Lemma 8, and using symmetrization Steinwart and Christmann (2008) (Prop.”
- Bayesian Learning of Kernel Embeddings (2016, background): “Take the convolution operator S k : L 2 ( X ; ν ) → H k with respect to a finite measure ν , defined as It is well known that the adjoint of S k is the inclusion of H k into L 2 (Steinwart and Christmann, 2008, Section 4.3).”
- Minimax Estimation of Kernel Mean Embeddings (2016, background): “lued functions. Then for any 0 < δ ≤ 1 with probability at least 1 −δwe have Z r(x)dPn(x) − Z r(x)dP(x) H ≤ r Ck n + r 2Cklog(1/δ) n . Proof. Note that r : X → H is a H-valued measurable function (Steinwart and Christmann, 2008, Lemma A.5.18) The condition in (44) ensures that R kr(x)kHdQ(x) ≤ √ ”
- A Kernelized Stein Discrepancy for Goodness-of-fit Tests (2016, methodology): “We first prove (12) by applying the reproducing property k ( x, x (cid:48) ) = (cid:104) k ( x, · ) , k ( x (cid:48) , · ) (cid:105) H on (8): where we used the fact that β where we used the fact that ∇ (Zhou, 2008; Steinwart & Christmann, 2008).”
- A Kernel Test of Goodness of Fit (2016, neutral): “Next we show that ξp(x, ·) is Bochner integrable (see Steinwart & Christmann, 2008, Definition A.5.20), Eq‖ξp(Z)‖Fd ≤ √ Eq‖ξp(Z)‖2Fd = √ Eqhp(Z,Z) ∞.”
- Fast Two-Sample Testing with Analytic Representations of Probability Measures (2015, background): “Since R d is separable then by [27, Lemma 4.33] Hilbert Space H k is separable.”
- Optimal Rates for Random Fourier Features (2015, background): “e empirical characteristic function and its applications. Annals of Statistics, 5(1):88–98, 1977. [8] G. B. Folland. Real Analysis: Modern Techniques and Their Applications. Wiley-Interscience, 1999. [9] B. Kulis and K. Grauman. Kernelized locality-sensitivehashing. IEEE Transactions on Pattern Anal”
- Learning Theory for Distribution Regression (2014, methodology): “We use the L2(K) = L2(K,K) shorthand if K := K1 = K2; L2(K) is separable if and only if K is separable (Steinwart and Christmann, 2008, page 506).”
- Kernel Mean Estimation via Spectral Filtering (2014, neutral): “…in [1], i.e., ˆ µ α = αf ∗ + (1 − α )ˆ µ P = α P i h f ∗ , e i i e i + (1 − α ) P i h ˆ µ P , e i i e i , where ( e i ) i ∈ N are the countable orthonormal basis (ONB) of H —countable ONB exist since H is separable which follows from X being separable and k being continuous [20, Lemma 4.33].”