Parallel Support Vector Machines: The Cascade SVM

By Hans Peter Graf, E. Cosatto, L. Eon Bottou, Igor Durdanovic, Vladimir Naumovich Vapnik. Neural Information Processing Systems, 2004

507 citations recorded by the source corpus.

View Paper: https://www.semanticscholar.org/paper/61abf3eb0f653b67c0eb42c527b6620db51d4f3f

How researchers see this paper

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

Citations in the corpus

Group-Based Alternating Direction Method of Multipliers for Distributed Linear Classification (2017, methodology): “In the first class, gradient methods were adopted to handle the gradients of parallel support vector machines (SVMs) in the Cascade SVM [13].”
Breaking the curse of kernelization: budgeted stochastic gradient descent for large-scale SVM training (2012, background): “Either optimal (e.g., Graf et al., 2005) or approximate (e.g., Zhu et al., 2009) solutions can be obtained by this type of methods.”
Distributed learning in wireless sensor networks (2005, background): “Inspired by the availability of increasingly large data sets, an active area of machine learning research focuses on “scaling up” existing learning algorithms to handle massive training databases; see, for example, [11], [17], [31], and [79] and references thereto and therein.”
Spectral–Spatial Classification of Hyperspectral Imagery Based on Partitional Clustering Techniques (2009, unknown): “However, in recent works, methods and parallel implementations to speed up the SVM training and classification have been proposed [64], [69], [70], [71].”
Interpretable models from distributed data via merging of decision trees (2013, methodology): “This operation is used as the key element of the cascade SVM [19] and plays central role in the approach A decision tree (or simply a tree) F predicts a label via a sequence of tests on attribute-values (each associated with an inner node) where depending on the last test result a next test is selec”
Consensus-Based Distributed Support Vector Machines (2010, background): “Moreover, convergence to the centralized SVM is generally not guaranteed for anypartitioning of the aggregate data set (Graf et al., 2005; Bordes et al., 2005).”
Semantic Pooling for Complex Event Analysis in Untrimmed Videos (2016, methodology): “Then, on top of the extracted low-level features, the cascade SVM [72] was trained for each semantic concept.”
Learning on the border: active learning in imbalanced data classification (2007, methodology): “One possible avenue consists of harnessing the computing power of multiple computers (Graf et al., 2005).”
Distributed GraphLab: A Framework for Machine Learning in the Cloud (2012, background): “tate, and communication protocols while simultaneously developing mathematically complex models and algorithms. Nonetheless, the demand for large-scale computational and storageresources,hasdrivenmany[2,14,15,27,30,35]todevelopnew parallel and distributed MLDM systems targeted at individual modelsan”
Large-scale Learning with SVM and Convolutional for Generic Object Categorization (2006, methodology): “The experiments reported here use the latest version of a fast implementation of SMO-type algorithm [10], running on a 32-node Linux cluster, with each node having dual AMD Athlon 1.”
Fast support vector machine training and classification on graphics processors (2008, methodology): “The cascade SVM (Graf et al., 2005) is another proposed method for parallelizing SVM training on clusters.”
SystemML: Declarative machine learning on MapReduce (2011, background): “Previous attempts at building scalable machine learning algorithms have largely been hand-tuned implementations on specialized hardware/parallel architectures [1], or as noted in [2], clever methods to parallelize individual learning algorithms on a cluster of machines [3], [4], [5].”
Distributed GraphLab: A Framework for Machine Learning in the Cloud (2012, background): “Nonetheless, the demand for large-scale computational and storage resources, has driven many [2, 14, 15, 27, 30, 35] to develop new parallel and distributed MLDM systems targeted at individual models and applications.”
Map-Reduce for Machine Learning on Multicore (2006, background): “6% speed up on average over 16 cores whereas the specialized SVM cascade [11] averages only 4%.”
Modelling input texts: from Tree Kernels to Deep Learning (2015, background): “Previous work in [54] suggests that support vectors collected from locally learned models can encode many of the relevant features retained by models learned globally.”