SVM Application ListSVM Application ListThis list of Support Vector Machineapplications grows thanks to visitors like you who ADDnew entries. Thank you in advance for your contribution. Support vector machines-based generalized predictive controlThis work presents an application of the previously proposed Support Vector Machines Based Generalized Predictive Control (SVM-Based GPC) method to the problem of controlling chaotic dynamics with small parameter perturbations. The Generalized Predictive Control (GPC) method, which is included in the class of Model Predictive Control, necessitates an accurate model of the plant that plays very crucial role in the control loop. On the other hand, chaotic systems exhibit very complex behavior peculiar to them and thus it is considerably difficult task to get their accurate model in the whole phase space. In this work, the Support Vector Machines (SVMs) regression algorithm is used to obtain an acceptable model of a chaotic system to be controlled. SVM-Based GPC exploits some advantages of the SVM approach and utilizes the obtained model in the GPC structure. Simulation results on several chaotic systems indicate that the SVM-Based GPC scheme provides an excellent performance with respect to local stabilization of the target (an originally unstable equilibrium point). Furthermore, it somewhat performs targeting, the task of steering the chaotic system towards the target by applying relatively small parameter perturbations. It considerably reduces the waiting time until the system, starting from random initial conditions, enters the local control region, a small neighborhood of the chosen target. Moreover, SVM-Based GPC maintains its performance in the case that the measured output is corrupted by an additive Gaussian noise. Reference(s): “Support vector machines-based generalized predictive control,” Serdar Iplikci, INTERNATIONAL JOURNAL OF ROBUST AND NONLINEAR CONTROL, Vol. 16, pp. 843-862, 2006 Reference link(s): http://ietfec.oxfordjournals.org/cgi/content/abstract/E89-A/10/2787 Data link(s): Entered by: Serdar Iplikci <iplikci@pau.edu.tr> - Monday, October 23, 2006 at 18:05:17 (GMT)Comments: Dynamic Reconstruction of Chaotic Systems from Inter-spike Intervals Using Least Squares Support Vector MachinesThis work presents a methodology for dynamic reconstruction of chaotic systems from inter-spike interval (ISI) time series obtained via
integrate-and-fire (IF) models. In this methodology, least squares support vector machines (LSSVMs) have been employed for approximating
the dynamic behaviors of the systems under investigation. Reference(s): Physica D, Vol. 216, pp. 282-293, 2006 Reference link(s): Data link(s): Entered by: Serdar Iplikci <iplikci@pau.edu.tr> - Monday, May 29, 2006 at 12:53:56 (GMT)Comments: Application of The Kernel Method to the Inverse Geosounding ProblemDetermining the layered structure of the earth demands the solution of a variety of inverse problems;
in the case of electromagnetic soundings at low induction
numbers, the problem is linear, for the measurements may be represented as a linear functional of the electrical conductivity distribution.
In this work, an application of the Support Vector (SV) Regression technique to the inversion of electromagnetic data is presented.
We take advantage of the regularizing properties of the SV learning algorithm and use it as a modeling technique with synthetic and field data.
The SV method presents better recovery of synthetic models than Tikhonov's regularization.
As the SV formulation is solved in the space of the data, which has a small dimension in this application, a smaller problem than that considered with Tikhonov's regularization is produced.
For field data, the SV formulation develops models similar to those obtained via linear programming techniques, but with the added characteristic of robustness.
Reference(s): "Application of the kernel method to the inverse geosounding problem", Hugo Hidalgo, Sonia Sosa and E. Gómez-Treviño, Neural Networks, vol. 16, pp. 349-353, 2003 Reference link(s): http://cienciascomp.cicese.mx/recopat/articulos/NeuralNetworks03.pdf Data link(s): Entered by: Hugo Hidalgo <hugo@cicese.mx> - Wednesday, March 22, 2006 at 14:04:25 (MST)Comments: Support Vector Machines Based Modeling of Seismic Liquefaction PotentialThis paper investigate the potential of support vector machines based classification approach to assess the liquefaction potential from actual standard penetration test (SPT) and cone penetration test (CPT) field data. Support vector machines are based on statistical learning theory and found to work well in comparison to neural networks in several other applications. Both CPT and SPT field data sets is used with support vector machines for predicting the occurrence and nonoccurrence of liquefaction based on different input parameter combination. With SPT and CPT test data sets, highest accuracy of 96% and 97% respectively was achieved with support vector machines. This suggests that support vector machines can effectively be used to model the complex relationship between different soil parameter and the liquefaction potential. Several other combinations of input variable were used to assess the influence of different input parameters on liquefaction potential. Proposed approach suggest that neither normalized cone resistance value with CPT data nor the calculation of standardized SPT value is required with SPT data. Further, support vector machines required few user-defined parameters and provide better performance in comparison to neural network approach. Reference(s): Goh ATC. Seismic Liquefaction Potential Assessed by Neural Networks. Journal of Geotechnical Engineering 1994; 120(9): 1467-1480. Goh ATC. Neural-Network Modeling of CPT Seismic Liquefaction Data. Journal of Geotechnical Engineering 1996; 122(1): 70-73 Reference link(s): Accepted for publication in International Journal for Numerical and Analytical Methods in Geomechanics. Data link(s): Entered by: Mahesh Pal <mpce_pal@yahoo.co.uk> - Wednesday, February 22, 2006 at 06:50:07 (GMT)Comments: SVM for Geo- and Environmental SciencesStatistical learning theory for geo(spatial) and spatio-temporal environmental data analysis and modelling.
Comparisons with geostatistical predictions and simulations Reference(s): 1. N. Gilardi, M. Kanevski, M. Maignan and E. Mayoraz. Environmental and Pollution Spatial Data Classification with Support Vector Machines and Geostatistics. Workshop W07 “Intelligent techniques for Spatio-Temporal Data Analysis in Environmental Applications”. ACAI99, Greece, July, 1999. pp. 43-51. www.idiap.ch 2. M Kanevski, N Gilardi, E Mayoraz, M Maignan. Spatial Data Classification with Support Vector Machines. Geostat 2000 congress. South Africa, April 2000.3. Kanevski M., Wong P., Canu S. Spatial Data Mapping with Support Vector Regression and Geostatistics. 7th International Conference on Neural Information Processing, Taepon, Korea. Nov. 14-18, 2000. Pp. 1307-1311.4. N GILARDI, Alex GAMMERMAN, Mikhail KANEVSKI, Michel MAIGNAN, Tom MELLUISH, Craig SAUNDERS, Volodia VOVK. Application des méthodes d’apprentissage pour l’étude des risques de pollution dans le Lac Léman. 5e Colloque transfrontalier CLUSE. Risques majeurs: perception, globalisation et management. Université de Genève, 2000. 5. M. Kanevski. Evaluation of SVM Binary Classification with Nonparametric Stochastic Simulations. IDIAP Research Report, IDIAP-RR-01-07, 17 p. 2001. www.idiap.ch 6. M. Kanevski, A. Pozdnukhov, S. Canu, M. Maignan. Advanced Spatial Data Analysis and Modelling with Support Vector Machines. International Journal on Fuzzy Systems 2002. p. 606-615.7. M. Kanevski , A. Pozdnukhov , S. Canu ,M. Maignan , P.M. Wong , S.A.R. Shibli “Support Vector Machines for Classification and Mapping of Reservoir Data”. In: “Soft Computing for Reservoir Characterization and Modelling”. P. Wong, F. Aminzadeh, M. Nikravesh (Eds.). Physica-Verlag, Heidelberg, N.Y. pp. 531-558, 2002.8. Kanevski M., Pozdnukhov A., McKenna S., Murray Ch., Maignan M. Statistical Learning Theory for Spatial Data. In proceedings of GeoENV2002 conference. Barcelona, 2002.9. M. Kanevski et al. Environmental data mining and modelling based on machine learning algorithms and geostatistics. Journal of Environmental Modelling and Software, 2004. vol. 19, pp. 845-855.10. M. Kanevski, M. Maignan et al. Advanced geostatistical and machine learning models for spatial data analysis of radioactively contaminated territories. Journal of Environmental Sciences and Pollution Research, pp.137-149, 2003. 11. Kanevski M., Maignan M. and Piller G. Advanced analysis and modelling tools for spatial environmental data. Case study: indoor radon data n Switzerland. International conference EnviroInfo, 2004. http://www.enviroinfo2004.org/cdrom/Datas/Kanevski.htm 12. Kanevski M., Maignan M. and Pozdnukhov A. Active Learning of Environmental Data Using Support Vector Machines. Conference of the International Association for Mathematical Geology, Toronto 2005. http://www.iamgconference.com/13. M. Kanevski, A. Pozdnukhov, M. Tonini, M. Motelica, E. Savelieva, M. Maignan. Statistical Learning Theory for Geospatial Data. Case study: Aral Sea. 14th European colloquium on Theoretical and Quantitative Geography. Portugal, September 2005. 14. Pozdnukhov A., Kanevski M. Monitoring network optimisation using support vector machines. In: Geostatistics for Environmental applications. (Renard Ph., Demougeot-Renard H and Froidevaux, Eds.). Springer, 2005. pp. 39-50. 15. Pozdnukhov A. and Kanevski M. Monitoring Network Optimisation for Spatial Data Classification Using Support Vector Machines. (2006). International Journal of Environment and Pollution. Vol.28. 20 pp. Reference link(s): www.unil.ch/igarwww.idiap.ch Data link(s): Entered by: Mikhail Kanevski <Mikhail.Kanevski@unil.ch> - Sunday, February 12, 2006 at 16:30:07 (GMT)Comments: SVM for Protein Fold and Remote Homology DetectionMotivation: Protein remote homology detection is a central problem in computational biology. Supervised learning algorithms based on support vector machines are currently one of the most effective methods for remote homology detection. The performance of these methods depends on how the protein sequences are modeled and on the method used to compute the kernel function between them.
Results: We introduce two classes of kernel functions that are constructed by combining sequence profiles with new and existing approaches for determining the similarity between pairs of protein sequences. These kernels are constructed directly from these explicit protein similarity measures and employ effective profile-to-profile scoring schemes for measuring the similarity between pairs of proteins. Experiments with remote homology detection and fold recognition problems show that these kernels are capable of producing results that are substantially better than those produced by all of the existing state-of-the-art SVM-based methods. In addition, the experiments show that these kernels, even when used in the absence of profiles, produce results that are better than those produced by existing non-profile-based schemes. Reference(s): Profile based direct kernels for remote homology detection and fold recognition by Huzefa Rangwala and George Karypis (Bioinformatics 2005) Reference link(s): http://bioinformatics.oxfordjournals.org/cgi/content/abstract/bti687v1 Data link(s): http://bioinfo.cs.umn.edu/supplements/remote-homology/ Entered by: Huzefa Rangwala <rangwala@cs.umn.edu> - Sunday, November 06, 2005 at 06:02:08 (GMT)Comments: content based image retrievalRelevance feedback schemes based on support vector machines (SVM) have been widely used in content-based image retrieval (CBIR). However, the performance of SVM based relevance feedback is often poor when the number of labeled positive feedback samples is small. This is mainly due to three reasons: (1) an SVM classifier is unstable on a small-sized training set; (2) SVM’s optimal hyper-plane may be biased when the positive feedback samples are much less than the negative feedback samples; and (3) overfitting happens because the number of feature dimensions is much higher than the size of the training set. In this paper, we develop a mechanism to overcome these problems. To address the first two problems, we propose an asymmetric bagging based SVM (AB-SVM). For the third problem, we combine the random subspace method and SVM for relevance feedback, which is named random subspacing SVM (RS-SVM). Finally, by integrating AB-SVM and RS-SVM, an asymmetric bagging and random subspacing SVM (ABRS-SVM) is built to solve these three problems and further improve the relevance feedback performance. Reference(s): Dacheng Tao, Xiaoou Tang, Xuelong Li, and Xindong Wu, Asymmetric Bagging and Random Subspacing for Support Vector Machines-based Relevance Feedback in Image Retrieval, IEEE Transactions on Pattern Analysis and Machine Intelligence, accepted, to appear. Reference link(s): Data link(s): Entered by: Dacheng Tao <- Monday, September 20, 1999 at 14:23:51 (PDT)Comments: Drucker-97 finds that SVMs outperform the baselinesystem (bagging) on the Boston housing problem. It is noted that SVMs canmake a real difference when the dimensionality of input space and the orderof the approximation create a dimensionality of feature space which isuntractable with other methods. The results of Drucker et al are furtherimproved in Stitson-99 (overall 35% better than the baseline method).3-D Object Recognition Problems3-D Object Recognition encompasses a wide variety of problems in PatternRecognition that have to do with classifying representations of 3-dimensionalobjects. This ranges from face recognition to Automatic Target Recognition(ATR) from radar images. Some of the challenges include that objects areusually seen from only one angle at a time, and may be partially occulted.After training the system must be able both to classify correctly the objectsof interest and reject other "confusers" or "distractors".Reference(s): Visual Learning and Recognition of 3-D Object from AppearanceH. Murase and S.K. NayarInt. J. Comput. Vision, Vol. 14, 1995, pp. 5--24.Comparison of view-based object recognition algorithms using realistic3D models.V. Blanz, B. Schölkopf,H. Bülthoff, C. Burges, V. Vapnik, T. & Vetter,In: C. von der Malsburg, W. von Seelen, J. C. Vorbrüggen, andB. Sendhoff (eds.): Artificial Neural Networks - ICANN'96.Springer Lecture Notes in Computer Science Vol. 1112, Berlin, 251-256.1996.(also published in Proc of ICANN'96,LNCS Vol. 1112, 1996, pp. 251--256.)Training Support Vector Machines: an Application to Face Detection,Edgar Osuna,Robert Freund and Federico Girosi.Proceedings of CVPR'97, Puerto Rico,1997.Kernel Principal Component Analysis.B. Schölkopf, A. Smola, K.-R.Müller,Proceedings ICANN'97, p.583. Springer Lecture Notes in Computer Science.1997.Schölkopf, B.; 1997. Support Vector Learning. PhD Thesis. Publishedby: R. Oldenbourg Verlag, Munich, 1997.Directaspect-based 3-D object recognition.MassimilianoPontil, and A. Verri.Proc. Int. Conf on Image Analysis and Processing, Firenze,1997.Automatic Target Recognition with Support Vector Machines,Qun Zhao and Jose Principe,NIPS-98 Workshop on Large Margin Classifiers,1998.The kernel adatron algorithm: a fast and simple learning procedure forsupport vector machines.T.-T. Friess, N. Cristianini,C. Campbell.15th Intl. Conf. Machine Learning, Morgan Kaufman Publishers.1998.View-based 3D object recognition with Support Vector Machines.Danny Roobaert & MarcM. Van HulleProc. IEEE Neural Networks for Signal Processing Workshop1999.Improvingthe Generalisation of Linear Support Vector Machines: an Application to3D Object Recognition with Cluttered BackgroundDanny RoobaertProc. Workshop on Support Vector Machines at the 16th InternationalJoint Conference on Artificial Intelligence, July31-August 6, Stockholm, Sweden, p. 29-331999. Reference link(s):Blanz-96,Osuna-97,Schölkopf-thesis-97,Schölkopf-NIPS-97,Pontil-97,Zhao-98,Friess-98,Roobaert-99a,Roobaert-99bData link(s):Chair data setSonar dataEntered by: IsabelleGuyon <isabelle@clopinet.com>- Saturday, September 18, 1999 at 17:54:51 (PDT) PontilMassimiliano <pontil@ai.mit.edu>- Monday, October 04, 1999 at 18:48:38 (PDT) Danny Roobaert <roobaert@nada.kth.se>- Friday, October 08, 1999 at 12:01:21 (PDT) Edited by Isabelle Guyon - Thursday, October 14, 1999.Comments: SVM's have been used either in the classificationstage or in the pre-processing (Kernel Principal Component Analysis).In Blanz-96. Support Vector Classifiers show excellent performance,leaving behind other methods. Osuna-97 demonstrates that SVCs can be trainedon very large data sets (50,000 examples). The classification performancereaches that of one of the best known system while being 30 times fasterat run time.In Schölkopf-97, the advantage of KPCA is more measured in termsof simplicity, ensured convergence, and ease of understanding of the non-linearities.Zhao-98 notes that SVCs with Gaussian kernels handle the rejectionof unknown "confusers" particularly well. Friess-98 reports performanceon the sonar data of Gorman and Sejnowski (1988). Their kernel adatronSVMs has a 95.2% success, compared to 90.2% for the best BackpropagationNeural Networks.Papageorgiou-98 applies SVM with a wavelet preprocessing to face andpeople detection, showing improvements with respect to their base system.Roobaert-99 shows that an SVM system working on raw data, not incorporatingany domain knowledge about the task, matches the performance of their baselinesystem that does incorporate such knowledge.Massimiliano Pontil points out that, as shown by the comparison withother techniques, it appears that SVMs can be effectively trained evenif the number of examples is much lower than the dimensionality of theobject space. In the paper Pontil-Verri-97, linear SVMs are used for 3-Dobject recognition. The potential of SVMs is illustrated on a databaseof 7200 images of 100 different objects. The proposed system does not requirefeature extraction and performs recognition on images regarded as pointsof a space of high dimension without estimating pose. The excellent recognitionrates achieved in all the performed experiments indicate that SVMs arewell-suited for aspect-based recognition.In Roobaert-99, 3 methods for the improvement of Linear Support VectorMachines are presented in the case of Pattern Recognition with a numberof irrelevent dimensions. A method for 3D object recognition without segmentationis proposed.Text CategorizationText categorization is the assignment of natural language texts to oneor more predefined categories based on their content. Applications include:assigning subject categories to documents to support text retrieval, routing,and filtering; email or files sorting into folder hierarchies; web pagesorting into search engine categories.Reference(s):Text Categorization with Support Vector Machines: Learning with Many RelevantFeatures.T. Joachims,European Conference on Machine Learning (ECML),1998.Inductive Learning Algorithms and Representations for Text Categorization,S. Dumais, J.Platt, D. Heckerman, M. Sahami,7th International Conference on Information and Knowledge Management,1998.Support Vector Machines for Spam Categorization. H. Drucker, with D. Wuand V. Vapnik. IEEE Trans. on Neural Networks , vol 10, number 5, pp. 1048-1054.1999.Transductive Inference for Text Classification using Support Vector Machines.Thorsten Joachims.International Conference on Machine Learning (ICML),1999.Reference link(s):Joachims-98Postcript,Joachims-98PDFDumaiset al 98Druckeret al 98Joachins-99PostcriptJoachims-99PDFData link(s):Reuters-21578Entered by: IsabelleGuyon <isabelle@clopinet.com>- Friday, September 17, 1999 at 15:19:48 (PDT). Last modified, October13, 1999.Comments: Joachims-98 reports that SVMs are well suited to learnin very high dimensional spaces (> 10000 inputs). They achieve substantialimprovements over the currently best performing methods, eliminating theneed for feature selection. The tests were run on the Oshumed corpus ofWilliamHersh and Reuter-21578. Dumais et al report that they use linear SVMsbecause they are both accurate and fast (to train and to use). They are35 times faster to train that the next most accurate classifier that theytested (Decision Trees). They have applied SVMs to the Reuter-21578 collection,emails and web pages. Drucker at al classify emails as spam and non spam.They find that boosting trees and SVMs have similar performance in termsof accuracy and speed. SVMs train significatly faster. Joachims-99 reportthat transduction is a very natural setting for many text classificationand information retrieval tasks. Transductive SVMs improve performanceespecially in cases with very small amounts of labelled training data.Time Series Prediction and Dynamic Resconstruction of Chaotic SystemsDynamic reconstruction is an inverse problem that deals with reconstructingthe dynamics of an unknown system, given a noisy time-series representingthe evolution of one variable of the system with time. The reconstructionproceeds by utilizing the time-series to build a predictive model of thesystem and, then, using iterated prediction to test what the model haslearned from the training data on the dynamics of the system.Reference(s):Predicting Time Series with Support Vector Machines.K.-R. Müller, A.Smola,G. Rätsch,B.Schölkopf,J. Kohlmorgen,V.Vapnik.Proceedings ICANN'97, p.999.Springer Lecture Notes in Computer Science, 1997Nonlinear Prediction of Chaotic Time Series using a Support Vector MachineS. Mukherjee, E. Osuna, and F. Girosi NNSP'97, 1997.Using Support Vector Machines for Time Series PredictionMüller, K.-R.; Smola, A.; Rätsch, G.; Schölkopf, B.;Kohlmorgen, J.; Vapnik, V.in Advances in Kernel Methods, B. Schölkopf, C.J.C. Burges, andA.J. Smola Eds.Pages 242-253, MIT Press, 1999. ISBN 0-262-19416-3.Support Vector Machines for Dynamic Reconstruction of a Chaotic SystemDavide Matterra andSimonHaykinin Advances in Kernel Methods, B. Schölkopf, C.J.C. Burges, andA.J. Smola Eds.Pages 211-241, MIT Press, 1999. ISBN 0-262-19416-3.Reference link(s):Müller et alMukherjee et alData link(s):Synthetic data used: Mackey-Glass, Ikewda Map and Lorenz, andSantaFe competition Data Set DEntered by: IsabelleGuyon <isabelle@clopinet.com>- Thursday, September 16, 1999 at 14:54:32 (PDT)Comments: Müller et al report excellent performance ofSVM. They set a new record on the Santa Fe competition data set D, 37%better than the winning approach during the competition. Mattera et alreport that SVM are effective for such tasks and that their main advantageis the possibility of trading off the required accuracy with the numberof Support Vectors.Support Vector Machine Classification of Microarray Gene Expression DataWe introduce a new method of functionally classifying genes using geneexpression data from DNA microarray hybridization experiments. The methodis based on the theory of support vector machines (SVMs). We describe SVMsthat use different similarity metrics, including a simple dot product ofgene expression vectors, polynomial versions of the dot product, and aradial basis function. The radial basis function SVM appears to providesuperior performance in classifying functional classes of genes when comparedto the other SVM similarity metrics. In addition, SVM performance is comparedto four standard machine learning algorithms. SVMs have many features thatmake them attractive for gene expression analysis, including their flexibilityin choosing a similarity function, sparseness of solution when dealingwith large data sets, the ability to handle large feature spaces, and theability to identify outliers.Reference(s):Support Vector Machine Classification of Microarray Gene ExpressionDataM. Brown, W. Grundy, D. Lin, N. Cristianini C. Sugnet, M. Ares Jr.,D. HausslerUniversity of California, Santa Cruz,technical report UCSC-CRL-99-09.Reference link(s):http://www.cse.ucsc.edu/research/compbio/genex/genex.tech.htmlData link(s):http://www.cse.ucsc.edu/research/compbio/genex/Entered by: NelloCristianini <nello.cristianini@bristol.ac.uk>- Friday, September 10, 1999 at 03:11:46 (PDT)Comments: SVMs outperformed all other classifers, when providedwith a specifically designed kernel to deal with very imbalanced data.Handwritten digit recognition problemsSupport vector classifiers were applied to the recognition of isolatedhandwritten digits optically scanned. This is a subtask of of zipcode automaticreading and courtesy amount recognition on checks.Reference(s):An training algorithm for optimal margin classifiers.B. Boser, I.Guyon, and V. Vapnik.In Fifth Annual Workshop on Computational Learning Theory, pages 144--152,Pittsburgh, ACM.1992.Writer adaptation for on-line handwritten character recognition.N. Matic , I. Guyon, J. Denker,and V. VapnikIn Second International Conference on Pattern Recognition and DocumentAnalysis , pages 187--191, Tsukuba, Japan, IEEE Computer Society Press,1993.Support Vector Networks.C. Cortes and V.Vapnik,Machine Learning, 20:273-297,1995.Learning algorithms for classification: A comparison on handwritten digitrecognition.Y. Le Cun, L.D. Jackel, L. Bottou, C. Cortes, J. Denker, H. Drucker,I. Guyon, U.A. Muller, E.Sackinger, P. Simard, and V. Vapnik.In J.H. Kwon and S. Cho, editors, Neural Networks: The StatisticalMechanics Perspective, pages261--276. World Scientific1995.Discovering informative patterns and data cleaning.I. Guyon, N. Matic , and V. Vapnik,In U.M. Fayyad, G. Piatetsky-Shapiro, P. Smyth, and R. Uthurusamy,editors, Advances in Knowledge Discovery and Data Mining, pages 181--203.MIT Press.1996.Incorporating Invariances in Support Vector Learning Machines.B. Schölkopf, C.Burges, and V. Vapnik,In: C. von der Malsburg, W. von Seelen, J. C. Vorbr|ggen, and B. Sendhoff(eds.): Artificial Neural Networks - ICANN'96. Springer Lecture Notes inComputer Science Vol. 1112, Berlin,47-521996.Prior Knowledge in Support Vector KernelsB. Schölkopf, P. Simard, A. Smola, and V. Vapnik,NIPS'971997.Pairwise Classification and Support Vector MachinesUlrich H.-G. Kresselin Advances in Kernel Methods, B. Schölkopf , C.J.C. Burges, andA.J. Smola Eds.Pages 255-268, MIT Press, 1999. ISBN 0-262-19416-3.The kernel adatron algorithm: a fast and simple learning procedure forsupport vector machines.T.-T. Friess, N. Cristianini,C. Campbell.15th Intl. Conf. Machine Learning, Morgan Kaufman Publishers.1998.Reference link(s):Boser-92,Matic-93,Guyon-96,Schölkopf-96,Schölkopf-98Data link(s):http://www.clopinet.com/isabelle/Projects/LITTLE1200/http://www.research.att.com/~yann/ocr/mnist/Entered by: IsabelleGuyon <isabelle@clopinet.com>- Thursday, September 09, 1999 at 16:27:38 (PDT)Comments: This is one of the first applications of SVCs. Itwas demonstrated that SCVs could be applied directly to pixel maps andnearly match or outperform other techniques requiring elaborate pre-processing,architecture design (structured neural networks), and/or a metric incorporatingprior knowledge of the task (tangent distance) -- see e.g. Lecun-95. Elaboratemetrics such as tangent distance can be used in combination with SVCs (Schölkopf-96-97)and yield improved performance. SVCs are also attractive for handwritingrecognition tasks because they lend themself to easy writer adaptationand data cleaning, by making use of the support vectors (Matic-93 and Guyon-96).In Friess-98, the kernel Adatron SVM slightly outperforms the originalSVM on the USPS character recognition benchmark.Breast cancer diagnosis and prognosisSupport vector machines have been applied to breast cancer diagnosis andprognosis. The Wisconsin breast cancer dataset contains 699 patterns with10 attributes for a binary classification task (the tumor is malignantor benign).Reference(s):O. L. Mangasarian, W. Nick Street and W. H. Wolberg: ``Breast cancerdiagnosis and prognosis via linear programming", Operations Research, 43(4),July-August 1995, 570-577.P. S. Bradley, O. L. Mangasarian and W. Nick Street: ``Feature selectionvia mathematical programming", INFORMS Journal on Computing 10, 1998, 209-217.P. S. Bradley, O. L. Mangasarian and W. Nick Street: ``Clustering viaconcave minimization", in ``Advances in Neural Information Processing Systems-9-", (NIPS*96), M. C. Mozer and M. I. Jordan and T. Petsche, editors,MIT Press, Cambridge, MA, 1997, 368-374.T.-T. Friess; N. Cristianini;C. Campbell.The kernel adatron algorithm: a fast and simple learning procedurefor support vector machines.15th Intl. Conf. Machine Learning, Morgan Kaufman Publishers.1998.Reference link(s):ftp://ftp.cs.wisc.edu/math-prog/tech-reports/94-10.psftp://ftp.cs.wisc.edu/math-prog/tech-reports/95-21.psftp://ftp.cs.wisc.edu/math-prog/tech-reports/96-03.pshttp://svm.first.gmd.de/papers/FriCriCam98.ps.gzData link(s):WDBC:Wisconsin Diagnostic Breast Cancer DatabaseBC:Wisconsin Prognostic Breast Cancer DatabaseEntered by: Prof. Olvi L.Mangasarian <olvi@cs.wisc.ed>- Thursday, September 09, 1999 at 15:25:50 (PDT)Modified by: Isabelle Guyon <isabelle@clopinet.com>- Monday, September 20, 1999 at 9:30 (PDT)Comments: Mangasarian et al use a linear programming formulationunderlying that can be interpreted as an SVM. Their system (XCYT) is ahighly accurate non-invasive breast cancer diagnostic program currentlyin use at University of Wisconsin Hospital. Friess et al report that theWisconsin breat cancer dataset has been extensively studied. Their system,which uses Adatron SVMs, has 99.48% success rate, compared to 94.2% (CART),95.9% (RBF), 96% (linear discriminant), 96.6% (Backpropagation network),all results reported elsewhere in the literature.Support Vector Decision Tree Methods for Database MarketingWe introduce a support vector decision tree method for customer targetingin the framework of large databases (database marketing). The goal is toprovide a tool to identify the best customers based on historical data(model development). Then this tool is used to forecast the best potentialcustomers among a pool of prospects through a process of scoring. We beginby recursively constructing a decision tree. Each decision consists ofa linear combination of the independent attributes. A linear program motivatedby the support vector machine method from Vapnik's Statistical LearningTheory is used to construct each decision. A gainschart table is used toverify the goodness of fit of the targeting, the likely prospects, andthe expected utility of profit. Successful results are given for threeindustrial problems. The linear program automatically performs dimensionalityreduction. The method consistently produced trees with a very small numberof decision nodes. Each decision consisted of a relatively small numberof attributes. The trees produced a clear division of the population intolikely prospects, unlikely prospects, and ambiguous prospects. The largesttraining dataset tested contained 15,700 points with 866 attributes. Commercialoptimization package used, CPLEX, is capable of solving even larger problems.Reference(s):On Support Vector Decision Trees for Database Marketing K.P. Bennett,D. Wu, and L. Auslender Report No. 98-100, Rensselaer Polytechnic Institute,Troy, NY, 1998.Reference link(s):http://www.rpi.edu/~bennek/mr98100.psData link(s):Data is proprietary.Entered by: Kristin P Bennett <bennek@rpi.edu>- Thursday, September 09, 1999 at 15:16:07 (PDT)Comments: The support vector decision tree performed betterthan C4.5. SVDT produced very simple trees using few attributes.Add a new entry to the SVMApplication List.Read more about SVM applications on the Kernel Machines website.This page was powered by MattWright Other resources: Artificial Intelligence resources -directory of Artificial Intelligence related websites. |
|
