NON-NEGATIVE MATRIX FACTORIZATION AND CLASSIFIERS: EXPERIMENTAL STUDY

Size: px
Start display at page:

Download "NON-NEGATIVE MATRIX FACTORIZATION AND CLASSIFIERS: EXPERIMENTAL STUDY"

Transcription

1 NON-NEGATIVE MATRIX FACTORIZATION AND CLASSIFIERS: EXPERIMENTAL STUDY Oleg G. Okun Machine Vision Group Infotech Oulu and Department of Electrical and Information Engineering P.O.Box 4500, FIN University of Oulu, FINLAND ABSTRACT Non-negative matrix factorization (NMF) is one of the recently emerged dimensionality reduction methods. Unlike other methods, NMF is based on non-negative constraints, which allow to learn parts from objects. In this paper, we combine NMF with four classifiers (nearest neighbor, kernel nearest neighbor, k-local hyperplane distance nearest neighbor and support vector machine) in order to investigate the influence of dimensionality reduction performed by NMF on the accuracy rate of the classifiers and establish when NMF is useless. Experiments were conducted on three real-world datasets (Japanese Female Facial Expression, UCI Sonar and UCI BUPA liver disorder). The first dataset contains face images as patterns whereas patterns in two others are composed of numerical measurements not constituting any real physical objects when assembled together. The preliminary conclusion is that while NMF turned out to be useful for lowering a dimensionality of face images, it caused a degradation in classification accuracy when applied to other two datasets. It indicates that NMF may not be good for datasets where patterns cannot be decomposed into meaningful parts, though this thought requires further, more detailed, exploration. As for classifiers, k-local hyperplane distance nearest neighbor demonstrated a very good performance, often outperforming other tested classifiers. KEY WORDS Dimensionality reduction, pattern classification, machine learning. 1 Introduction Pattern classification is one of the final operations in the object interpretation and recognition process. Operations preceding to it influence on classification results and therefore selection of appropriate preprocessing techniques is of great importance. An almost unavoidable choice for a preprocessing operation includes dimensionality reduction, since it can reduce the negative effect known as curse of dimensionality. Though it seems that dimensionality reduction followed by classification has been studied for various combinations, we selected a relatively new technique called non-negative matrix factorization [1] and combine it with four advanced classifiers (nearest neighbor, kernel nearest neighbor, k-local hyperplane distance nearest neighbor and support vector machine) in order to find out how NMF influences on the classification error rate. Since its first appearance, NMF and its modifications were studied in a number of works [2, 3, 4, 5, 6, 7, 8], where NMF was applied for image retrieval and classification. However, usefulness of NMF for general object classification (when objects are not images, but rather sets of numerical features, e.g. measurements), was not yet verified. In this paper, we therefore explore this aspect in order to find out if NMF can bring improvements to classification accuracy, even if its definition does not suggest such kind of application. Moreover, all (except nearest neighbor) the selected classifiers were not used with NMF to our best knowledge. Thus, we extend research concerning coupling NMF with different classifiers with the goal to determine appropriate ones. The paper has the following structure. Sections 2 and 3 briefly describe the non-negative matrix factorization and four classifiers, respectively. Experiments are presented in Section 4. Finally Section 5 concludes the paper. 2 Non-negative Matrix Factorization (NMF) Non-negative matrix factorization is a linear dimensionality reduction technique which is distinguished from such methods as PCA by its non-negativity constraints. These constraints lead to a part-based representation because they allow only additive, not subtractive, combinations. Given a non-negative n m matrix V (m vectors, each of n elements), NMF finds a non-negative n r matrix W and a non-negative matrix r m matrix H such that V WH. The value of r is selected according to the rule r < nm n+m in order to obtain dimensionality reduction. Each column of W is a basis vector while each column of H is a reduced representation of the corresponding column of V. In other words, W can be seen as a basis that is optimized for linear approximation of the data in V. NMF provides the following simple learning rule guaranteeing to monotonically converge without the need

2 for setting any adjustable parameters [1]: W ia W ia W ia µ H aµ H aµ V iµ (W H) iµ H aµ, (1) W ia j W ja i, (2) W ia V iµ (W H) iµ. (3) The matrices W and H are initialized with positive random values. Equations (1-3) iterate until convergence to a local maximum of the following objective function: F = n i=1 µ=1 m (V iµ log(w H) iµ (W H) iµ ). (4) In its original form, NMF can be slow to converge to a local maximum for large V. We found that introducing a parameter tol (0 < tol 1) to decide when to stop iterations significantly speeds up the convergence (from several hours to several minutes) without degrading the approximation quality. This is also in accordance with information retrieval results in [2], where the average precision remained almost the same as the number of iterations increased. After learning the NMF basis functions, i.e. the matrix W, new data are projected into r dimensional space by fixing W, randomly initializing H as described above, and iterating (1-3) until convergence [3]. 3 Classifiers 3.1 K Nearest Neighbor (KNN) There is no training step and classification of unseen points is simple. K nearest neighbors are sought for each test point among members of the training set and a point in question is assigned to a class of majority neighbors. To avoid ties, K should be odd. Standard values for K are 1, 3 or 5 in order to keep the computational cost low. 3.2 Support Vector Machine (SVM) Unlike many traditional methods which implement the Empirical Risk Minimization Principle which aims at minimizing the training error, SVM implements the Structural Risk Minimization Principle [9] which seeks to minimize an upper bound of the generalization error, which eventually results in better generalization of SVM than that of traditional techniques. Currently, SVM is considered as the most advanced classifier which can outperform various models of neural networks in many practical tasks. The training of SVM is equivalent to solving a linearly constrained convex quadratic programming problem and therefore the solution of SVM is always globally optimal and free from local minima. In fact, the solution is only determined by the support vectors which are a subset of the training data so that the solution is often very sparse. Another advantage of SVM is that its solution does not depend on a data dimensionality, unlike that of many other methods, and this makes it an attractive choice for dealing with high dimensional datasets. Given that the data cannot be always linearly separated in an input space, SVM performs their mapping into another, a higher dimensional feature space where the data are supposed to be linearly separated. So called kernel trick allows not to calculate this mapping explicitly. Instead, the mapping into the feature space is implicitly defined by a kernel function computing the inner product of two feature vectors corresponding to two inputs. In some cases, if the data are noisy, there can be no linear separation in the feature space. To deal with this obstacle, the following (dual) optimization problem is to be solved 1 : maximize W (α) = m i=1 α i 1 m 2 i,j=1 y iy j α i α j K(x i, x j ), subject to m i=1 y iα i = 0, 0 α i C, i = 1,..., m, (5) where α i are Lagrange multipliers, K(x i, x j ) is a kernel value for inputs x i and x j that correspond to outputs y i and y j, and C is the trade-off between minimizing the number of training errors and maximizing the margin (classifier capacity). Given a test point x, the nonlinear SVM classifier takes the form 2 : f(x) = m α i y i K(x, x i ) + b, i=1 y(x) = sign(f(x)), (6) where the scalar b can be chosen so that y i f(x i ) = 1 for any i with 0 < α i < C. 3.3 Kernel Nearest Neighbor (KernelNN) This algorithm is a modification of the standard KNN when applying kernels [10]. When selecting an appropriate kernel, the kernel nearest neighbor algorithm, via a nonlinear mapping to a high dimensional feature space, may be superior over KNN for some sample distributions. The same kernels as in case of SVM are commonly used, but as remarked in [10], only the polynomial kernel with degree p 1 is actually useful, since the polynomial kernel with p = 1 and radial basis kernel degenerate KernelNN to the conventional KNN. 1 We talk here and further in this section about 2-classes problems for simplicity. 2 In fact, since many α i = 0, the actual upper summation limit is (much) less than m. Points for which 0 < α i < C are called support vectors.

3 The kernel approach to KNN consists of two steps: kernel computation, followed by distance computation in the feature space expressed via a kernel K. After that, a nearest neighbor rule is applied just as in case of KNN. In (7), KernelNN redefines a distance metric for using in the feature space. d(φ(x), φ(y)) = K(x, x) 2K(x, y) + K(y, y), (7) where K(x, y) = φ(x), φ(y) and φ(x) is a mapping of x into the feature space. Similarly to SVM, the problem of choosing the optimal kernel and its parameters influences on the error rate. However, given the optimal kernel, results of KernelNN will be not worse than those of KNN, since KNN is a specific kind of KernelNN. 3.4 K-Local Hyperplane Distance Nearest Neighbor (HKNN) This algorithm was proposed in [11] in order to improve KNN in its competition with SVM. Unlike (nonlinear) SVMs which map the input data into a higher dimensional feature space and build a decision surface (hyperplane maximizing a margin) in this space, an alternative approach exploited in [11] is to construct a nonlinear decision surface directly in the input space. It is often convenient to consider each class as forming a low dimensional manifold embedded in a higher dimensional input space. This low dimensional manifold can be reasonably assumed locally linear. The reason why SVM often outperforms KNN is that in case of a finite number of samples, i.e. when data are relatively sparse, missing samples, appearing as holes, bring distortions (e.g. complicated nonlinearity leading to overfitting) to the decision surface produced by KNN 3, and this negatively affects generalization performance of KNN. The idea of HKNN is to fantasize the missing points to remedy this drawback. Given a test point, its K nearest neighbors, found among training points belonging to a certain class, approximately form the local hyperplane passing through them. Thus this hyperplane would include all the missing points of the manifold representing this class. Having such a hyperplane constructed for each class, a given test point x is assigned to the class whose hyperplane is closest to x. To compute a hyperplane, one needs to solve a small linear system (see [11] for details) so that the overall computational load is still low and comparable to that of KNN. Except K whose value should be typically larger than that in KNN (in order to have enough points for constructing hyperplanes), HKNN needs to set another (regularization) parameter, λ, responsible for tolerance of hyperplane fitting to noise and outliers. SVM. 3 However, they do not influence on the decision surface produced by 4 Experiments Four classifiers were applied to the data of the original and reduced dimensionality. The main goal of experiments was to verify the influence of NMF on the error rate of each classifier as well as to compare the classifiers between themselves. 4.1 Data Japanese Female Facial Expression (JAFFE) dataset This dataset 4 contains grayscale ( pixels) images of 10 Japanese females. The total number of images is 213 with the number of images per female ranging from 20 to 23. Three or four examples of one of the seven facial expressions are demonstrated by each female. Our interest was not to recognize facial expressions, but to distinguish different females, independently of expressions. To lower a data dimensionality, each image was downsampled to pixels by using bilinear interpolation. We called a new dataset JAFFE16. No further processing was applied UCI Sonar dataset (Sonar) It is one of the UCI datasets 5. There are dimensional feature vectors divided into two classes (97 and 111 samples, respectively) UCI BUPA liver disorder dataset (Liver) It is another UCI dataset 6. It contains dimensional feature vectors belonging to two classes (145 and 200 samples, respectively). Unlike two other datasets, the Liver data were normalized by dividing each feature by its mean so as to keep features non-negative Parameter Values of Algorithms In this section we list parameter values of each algorithm when applied to each dataset. Cross-validation was used in order to determine the optimal values. Since the polynomial kernel can be only used with KernelNN, the same kernel (and the same polynomial degree) was also used with SVM (C-SVM from OSU-SVM toolbox was used 8 ). Except for C, other parameters had default values as given in the toolbox, since they did not significantly influence on the results. Parameter r of NMF varied depending on the 4 mlyons/jaffe.html 5 ftp://ftp.ics.uci.edu/pub/machine-learningdatabases/undocumented/connectionist-bench/sonar 6 ftp://ftp.ics.uci.edu/pub/machine-learning-databases/liver-disorders 7 Recall that NMF requires non-negative inputs. 8 maj/osu svm/

4 original dimensionality. For Liver data, K in KNN and KernelNN and p in KernelNN were all set to 3 in order to keep them identical to those used in [10]. Table 1. Parameter values used in tests with the JAFFE16 dataset. KNN K = 1 KernelNN K = 1, p = 3 HKNN K = 10, λ = 10 SVM C = 1, p = 3 NMF r = 25, 50, 75, 100, 125, tol = 0.05 can be used to perform classification with this dataset. It is difficult to compare our results with those of other methods, since this dataset was tested for facial expression, but not for face, recognition. Error rate (%) fold cross validation error vs. dimensionality KNN KernelNN HKNN SVM NMF+KNN NMF+KernelNN NMF+HKNN NMF+SVM Table 2. dataset. Parameter values used in tests with the Sonar 6 4 KNN K = 1 KernelNN K = 1, p = 3 HKNN K = 5, λ = 25 SVM C = 100, p = 3 NMF r = 6, 12, 18, 24, 30, tol = Figure 1. Cross-validation error versus dimensionality of a reduced space for JAFFE16 data. Table 3. dataset. Parameter values used in tests with the Liver KNN K = 3 KernelNN K = 3, p = 3 HKNN K = 10, λ = 10 SVM C = 1, p = 3 NMF r = 2, 3, 4, 5, tol = Classification In all experiments, 10-fold cross-validation was used in order to compute error rates. Let us assume the dimensionality of the original space is n, whereas that of the space obtained after dimensionality reduction is r with r < n. Cross-validation errors for each value of r are collected in groups of eight bars in figures below. The first four bars are associated with KNN, KernelNN, HKNN, and SVM, respectively, while other bars correspond to these classifiers coupled with NMF. For JAFFE16 KNN and HKNN demonstrated the best performance in R n, though SVM closely followed it. Cross-validation errors are provided in Fig. 1. It is seen that NMF is useful for KNN and HKNN (the accuracy was sometimes lower in R r than in R n ), but not for KernelNN and SVM. We also experimented by applying all the four classifiers without NMF to images of various resolution (32 32, 64 64, , and pixels). It turned out that the resolution did not influence on error rate. That is, low resolution images (as low as pixels) For Sonar cross-validation errors are shown in Fig. 2. Error rates for all methods in R n were approximately identical and equal to 17.5% in aspect-angle independent experiments 9. They were, however, worse (but not always) in R r. Unlike JAFFE16, it is hard to say here when NMF could be used and for what classifier. Error rate (%) fold cross validation error vs. dimensionality KNN KernelNN HKNN SVM NMF+KNN NMF+KernelNN NMF+HKNN NMF+SVM Figure 2. Cross-validation error versus dimensionality of a reduced space for Sonar data. As remarked in [10], Liver data are highly nonlinear 9 This result closely resembles results obtained by other researchers on this dataset.

5 and hard to classify: accuracies achieved in [10] with KernelNN (K = 3, p = 3) and SVM (p = 3) are 71% and 68%, respectively. Thus, this set provides a challenge for classifiers and this is why we selected this dataset despite its low dimensionality. As with two previous examples, cross-validation errors are shown in Fig. 3. Again you can see that the accuracy degraded after dimensionality reduction, but not always dramatically. In fact, the accuracy of NMF+classifier mostly lagged behind the results reported in [10]. Sammon stress Sammon stress vs. dimensionality 10 fold cross validation error vs. dimensionality KNN KernelNN HKNN SVM NMF+KNN NMF+KernelNN NMF+HKNN NMF+SVM Error rate (%) Figure 4. Sammon-like stress versus dimensionality of a reduced space for JAFFE16 data Sammon stress vs. dimensionality Figure 3. Cross-validation error versus dimensionality of a reduced space for Liver data. Sammon stress Topology Preservation In addition to classification error rates we computed a Sammon-like stress [12], E, in (8) measuring topology preservation of the original set after dimensionality reduction. Measurements were done for each dataset for different r s and results are shown in Figs E = m 1 i=1 1 m j=i+1 δ ij m 1 m i=1 j=i+1 (δ ij d ij ) 2 δ ij, (8) where E [0, 1] with 0 indicating a lossless mapping R n R r i.e. the closer E to zero, the better topology of the original space is preserved in a reduced space; d ij and δ ij are the distances between the ith and jth points in R n and R r, respectively. All plots naturally indicate that topology is best preserved in higher dimensional than in lower dimensional spaces. However, there are distinctions between datasets. For example, when the dimensionality is reduced by half, the smallest E ( 0.23 for r = 3) was attained for Liver data whereas the largest E (slightly over 0.76 for r = 125) for JAFFE16 data. For Sonar data E 0.44 for r = 30. It Figure 5. Sammon-like stress versus dimensionality of a reduced space for Sonar data. means that if one reduces the dimensionality by half by using NMF, the topology is best preserved for Liver data, followed by Sonar and JAFFE16 data. In other words, Liver data allow higher dimensionality reduction than the other two sets. Additional information about acceptable levels for dimensionality reduction can be learned from error rates associated with certain r s for each classifier individually. Perhaps, a relative decrease of E is more important than its absolute value. 5 Discussion and Conclusion NMF coupled with four classifiers was tested on three realworld datasets with the aim to figure out when it will be

6 Sammon stress Sammon stress vs. dimensionality [2] S. Tsuge, M. Shishibori, S. Kuroiwa, & K. Kita, Dimensionality reduction using non-negative matrix factorization for information retrieval, Proc IEEE Int. Conf. on Systems, Man, and Cybernetics, Tucson, AZ, 2001, [3] D. Guillamet, & J. Vitrià, Discriminant basis for object classification, Proc. 11th Int. Conf. on Image Analysis and Processing, Palermo, Italy, 2001, [4] D. Guillamet, & J. Vitrià, Evaluation of distance metrics for recognition based on non-negative matrix factorization, Pattern Recognition Letters, 24(9-10), 2003, Figure 6. Sammon-like stress versus dimensionality of a reduced space for Liver data. useful. It turned out to be almost useless, except for classifying JAFFE16 data in R r with KNN and especially with HKNN 10. The reason of this failure can be that NMF is designed to learn parts from objects. Only JAFFE16 is composed of objects - faces, while other two sets just contain numerical features which cannot be considered as real objects when combined together. That is, NMF can provide any valuable results only if a dataset includes images of physical objects/subjects, such as those of people, cars, etc. However, the standard NMF, though being claimed to learn parts, in fact, can be holistic, i.e. parts of objects learned by NMF are not necessarily always localized. Local NMF (LNMF) [6, 8] and Weighted NMF (WNMF) [5] both try to remedy this problem and therefore are worthy for consideration instead of NMF. Another deficiency of NMF lowering its practical value is that the standard NMF often produces correlated basis functions, i.e. despite dimensionality reduction the curse of dimensionality may be not totally avoided. In this case, correlated features can still deteriorate the performance of a classifier. It could be an explanation why NMF did not show clear advantages over PCA (it sometimes even lose to PCA) when classifying MNIST characters [4], Corel images [5], and ORL face images [6]. However, further research involving more datasets is definitely necessary before drawing stronger conclusions. References [1] D.D. Lee, & H.S. Seung, Learning the parts of objects by non-negative matrix facorization, Nature, 401(6755), 1999, [5] D. Guillamet, J. Vitrià & B. Schiele, Introducing a weighted non-negative matrix factorization for image classification, Pattern Recognition Letters, 24(14), 2003, [6] S.Z. Li, X.W. Hou, H.J. Zhang, & Q.S. Cheng, Learning spatially localized, parts-based representation, Proc IEEE Conf. on Computer Vision and Pattern Recognition, Kauai, HW, 2001, [7] B. Xu, J. Lu, & G. Huang, A constrained non-negative matrix factorization in information retrieval, Proc IEEE Int. Conf. on Information Reuse and Integration, Las Vegas, NV, 2003, [8] Y. Wang, Y. Jia, C. Hu, & M. Turk, Fisher nonnegative matrix factorization for learning local features, Proc. 6th Asian Conf. on Computer Vision, Jeju Island, Korea, [9] V. Vapnik, The nature of statistical learning theory (Berlin: Springer-Verlag, 1995). [10] K. Yu, L. Ji, & X. Zhang, Kernel nearest-neighbor algorithm, Neural Processing Letters, 15(2), 2002, [11] P. Vincent, & Y. Bengio, K-local hyperplane and convex distance nearest neighbor algorithms, in T.G. Dietterich, S. Becker, & Z. Ghahramani (Ed.) Advances in Neural Information Processing Systems, 14, (Cambridge, MA: MIT Press, 2002) [12] E. Pȩkalska, D. de Ridder, R.P.W. Duin, & M.A. Kraaijveld, A new method of generalizing Sammon mapping with application to algorithm speedup, Proc. 5th Annual Conf. of the Advanced School for Computing and Imaging, Heijen, the Netherlands, 1999, In fact, HKNN demonstrated an excellent performance in both original and reduced spaces and can be considered as a strong alternative to SVM.

Kernel Combination Versus Classifier Combination

Kernel Combination Versus Classifier Combination Kernel Combination Versus Classifier Combination Wan-Jui Lee 1, Sergey Verzakov 2, and Robert P.W. Duin 2 1 EE Department, National Sun Yat-Sen University, Kaohsiung, Taiwan wrlee@water.ee.nsysu.edu.tw

More information

Data Mining: Concepts and Techniques. Chapter 9 Classification: Support Vector Machines. Support Vector Machines (SVMs)

Data Mining: Concepts and Techniques. Chapter 9 Classification: Support Vector Machines. Support Vector Machines (SVMs) Data Mining: Concepts and Techniques Chapter 9 Classification: Support Vector Machines 1 Support Vector Machines (SVMs) SVMs are a set of related supervised learning methods used for classification Based

More information

Naïve Bayes for text classification

Naïve Bayes for text classification Road Map Basic concepts Decision tree induction Evaluation of classifiers Rule induction Classification using association rules Naïve Bayesian classification Naïve Bayes for text classification Support

More information

K-Local Hyperplane and Convex Distance Nearest Neighbor Algorithms

K-Local Hyperplane and Convex Distance Nearest Neighbor Algorithms K-Local Hyperplane and Convex Distance Nearest Neighbor Algorithms Pascal Vincent and Yoshua Bengio Dept. IRO, Université de Montréal C.P. 6128, Montreal, Qc, H3C 3J7, Canada {vincentp,bengioy}@iro.umontreal.ca

More information

Feature scaling in support vector data description

Feature scaling in support vector data description Feature scaling in support vector data description P. Juszczak, D.M.J. Tax, R.P.W. Duin Pattern Recognition Group, Department of Applied Physics, Faculty of Applied Sciences, Delft University of Technology,

More information

Table of Contents. Recognition of Facial Gestures... 1 Attila Fazekas

Table of Contents. Recognition of Facial Gestures... 1 Attila Fazekas Table of Contents Recognition of Facial Gestures...................................... 1 Attila Fazekas II Recognition of Facial Gestures Attila Fazekas University of Debrecen, Institute of Informatics

More information

Using Analytic QP and Sparseness to Speed Training of Support Vector Machines

Using Analytic QP and Sparseness to Speed Training of Support Vector Machines Using Analytic QP and Sparseness to Speed Training of Support Vector Machines John C. Platt Microsoft Research 1 Microsoft Way Redmond, WA 9805 jplatt@microsoft.com Abstract Training a Support Vector Machine

More information

All lecture slides will be available at CSC2515_Winter15.html

All lecture slides will be available at  CSC2515_Winter15.html CSC2515 Fall 2015 Introduc3on to Machine Learning Lecture 9: Support Vector Machines All lecture slides will be available at http://www.cs.toronto.edu/~urtasun/courses/csc2515/ CSC2515_Winter15.html Many

More information

Basis Functions. Volker Tresp Summer 2017

Basis Functions. Volker Tresp Summer 2017 Basis Functions Volker Tresp Summer 2017 1 Nonlinear Mappings and Nonlinear Classifiers Regression: Linearity is often a good assumption when many inputs influence the output Some natural laws are (approximately)

More information

Bagging and Boosting Algorithms for Support Vector Machine Classifiers

Bagging and Boosting Algorithms for Support Vector Machine Classifiers Bagging and Boosting Algorithms for Support Vector Machine Classifiers Noritaka SHIGEI and Hiromi MIYAJIMA Dept. of Electrical and Electronics Engineering, Kagoshima University 1-21-40, Korimoto, Kagoshima

More information

SUPPORT VECTOR MACHINES

SUPPORT VECTOR MACHINES SUPPORT VECTOR MACHINES Today Reading AIMA 18.9 Goals (Naïve Bayes classifiers) Support vector machines 1 Support Vector Machines (SVMs) SVMs are probably the most popular off-the-shelf classifier! Software

More information

DECISION-TREE-BASED MULTICLASS SUPPORT VECTOR MACHINES. Fumitake Takahashi, Shigeo Abe

DECISION-TREE-BASED MULTICLASS SUPPORT VECTOR MACHINES. Fumitake Takahashi, Shigeo Abe DECISION-TREE-BASED MULTICLASS SUPPORT VECTOR MACHINES Fumitake Takahashi, Shigeo Abe Graduate School of Science and Technology, Kobe University, Kobe, Japan (E-mail: abe@eedept.kobe-u.ac.jp) ABSTRACT

More information

Face Recognition Based on LDA and Improved Pairwise-Constrained Multiple Metric Learning Method

Face Recognition Based on LDA and Improved Pairwise-Constrained Multiple Metric Learning Method Journal of Information Hiding and Multimedia Signal Processing c 2016 ISSN 2073-4212 Ubiquitous International Volume 7, Number 5, September 2016 Face Recognition ased on LDA and Improved Pairwise-Constrained

More information

Training Data Selection for Support Vector Machines

Training Data Selection for Support Vector Machines Training Data Selection for Support Vector Machines Jigang Wang, Predrag Neskovic, and Leon N Cooper Institute for Brain and Neural Systems, Physics Department, Brown University, Providence RI 02912, USA

More information

Kernel Methods & Support Vector Machines

Kernel Methods & Support Vector Machines & Support Vector Machines & Support Vector Machines Arvind Visvanathan CSCE 970 Pattern Recognition 1 & Support Vector Machines Question? Draw a single line to separate two classes? 2 & Support Vector

More information

Applying Supervised Learning

Applying Supervised Learning Applying Supervised Learning When to Consider Supervised Learning A supervised learning algorithm takes a known set of input data (the training set) and known responses to the data (output), and trains

More information

Binarization of Color Character Strings in Scene Images Using K-means Clustering and Support Vector Machines

Binarization of Color Character Strings in Scene Images Using K-means Clustering and Support Vector Machines 2011 International Conference on Document Analysis and Recognition Binarization of Color Character Strings in Scene Images Using K-means Clustering and Support Vector Machines Toru Wakahara Kohei Kita

More information

SELECTION OF THE OPTIMAL PARAMETER VALUE FOR THE LOCALLY LINEAR EMBEDDING ALGORITHM. Olga Kouropteva, Oleg Okun and Matti Pietikäinen

SELECTION OF THE OPTIMAL PARAMETER VALUE FOR THE LOCALLY LINEAR EMBEDDING ALGORITHM. Olga Kouropteva, Oleg Okun and Matti Pietikäinen SELECTION OF THE OPTIMAL PARAMETER VALUE FOR THE LOCALLY LINEAR EMBEDDING ALGORITHM Olga Kouropteva, Oleg Okun and Matti Pietikäinen Machine Vision Group, Infotech Oulu and Department of Electrical and

More information

Introduction to Support Vector Machines

Introduction to Support Vector Machines Introduction to Support Vector Machines CS 536: Machine Learning Littman (Wu, TA) Administration Slides borrowed from Martin Law (from the web). 1 Outline History of support vector machines (SVM) Two classes,

More information

Lecture 9: Support Vector Machines

Lecture 9: Support Vector Machines Lecture 9: Support Vector Machines William Webber (william@williamwebber.com) COMP90042, 2014, Semester 1, Lecture 8 What we ll learn in this lecture Support Vector Machines (SVMs) a highly robust and

More information

Robust PDF Table Locator

Robust PDF Table Locator Robust PDF Table Locator December 17, 2016 1 Introduction Data scientists rely on an abundance of tabular data stored in easy-to-machine-read formats like.csv files. Unfortunately, most government records

More information

Introduction to object recognition. Slides adapted from Fei-Fei Li, Rob Fergus, Antonio Torralba, and others

Introduction to object recognition. Slides adapted from Fei-Fei Li, Rob Fergus, Antonio Torralba, and others Introduction to object recognition Slides adapted from Fei-Fei Li, Rob Fergus, Antonio Torralba, and others Overview Basic recognition tasks A statistical learning approach Traditional or shallow recognition

More information

The Effects of Outliers on Support Vector Machines

The Effects of Outliers on Support Vector Machines The Effects of Outliers on Support Vector Machines Josh Hoak jrhoak@gmail.com Portland State University Abstract. Many techniques have been developed for mitigating the effects of outliers on the results

More information

Technical Report. Title: Manifold learning and Random Projections for multi-view object recognition

Technical Report. Title: Manifold learning and Random Projections for multi-view object recognition Technical Report Title: Manifold learning and Random Projections for multi-view object recognition Authors: Grigorios Tsagkatakis 1 and Andreas Savakis 2 1 Center for Imaging Science, Rochester Institute

More information

Support Vector Machines

Support Vector Machines Support Vector Machines RBF-networks Support Vector Machines Good Decision Boundary Optimization Problem Soft margin Hyperplane Non-linear Decision Boundary Kernel-Trick Approximation Accurancy Overtraining

More information

Face Recognition using Eigenfaces SMAI Course Project

Face Recognition using Eigenfaces SMAI Course Project Face Recognition using Eigenfaces SMAI Course Project Satarupa Guha IIIT Hyderabad 201307566 satarupa.guha@research.iiit.ac.in Ayushi Dalmia IIIT Hyderabad 201307565 ayushi.dalmia@research.iiit.ac.in Abstract

More information

A Maximal Margin Classification Algorithm Based on Data Field

A Maximal Margin Classification Algorithm Based on Data Field Send Orders for Reprints to reprints@benthamscience.ae 1088 The Open Cybernetics & Systemics Journal, 2015, 9, 1088-1093 Open Access A Maximal Margin Classification Algorithm Based on Data Field Zhao 1,*,

More information

Multiresponse Sparse Regression with Application to Multidimensional Scaling

Multiresponse Sparse Regression with Application to Multidimensional Scaling Multiresponse Sparse Regression with Application to Multidimensional Scaling Timo Similä and Jarkko Tikka Helsinki University of Technology, Laboratory of Computer and Information Science P.O. Box 54,

More information

Robust Face Recognition via Sparse Representation

Robust Face Recognition via Sparse Representation Robust Face Recognition via Sparse Representation Panqu Wang Department of Electrical and Computer Engineering University of California, San Diego La Jolla, CA 92092 pawang@ucsd.edu Can Xu Department of

More information

Support Vector Machines

Support Vector Machines Support Vector Machines RBF-networks Support Vector Machines Good Decision Boundary Optimization Problem Soft margin Hyperplane Non-linear Decision Boundary Kernel-Trick Approximation Accurancy Overtraining

More information

Classification by Support Vector Machines

Classification by Support Vector Machines Classification by Support Vector Machines Florian Markowetz Max-Planck-Institute for Molecular Genetics Computational Molecular Biology Berlin Practical DNA Microarray Analysis 2003 1 Overview I II III

More information

DM6 Support Vector Machines

DM6 Support Vector Machines DM6 Support Vector Machines Outline Large margin linear classifier Linear separable Nonlinear separable Creating nonlinear classifiers: kernel trick Discussion on SVM Conclusion SVM: LARGE MARGIN LINEAR

More information

Linear methods for supervised learning

Linear methods for supervised learning Linear methods for supervised learning LDA Logistic regression Naïve Bayes PLA Maximum margin hyperplanes Soft-margin hyperplanes Least squares resgression Ridge regression Nonlinear feature maps Sometimes

More information

LECTURE 5: DUAL PROBLEMS AND KERNELS. * Most of the slides in this lecture are from

LECTURE 5: DUAL PROBLEMS AND KERNELS. * Most of the slides in this lecture are from LECTURE 5: DUAL PROBLEMS AND KERNELS * Most of the slides in this lecture are from http://www.robots.ox.ac.uk/~az/lectures/ml Optimization Loss function Loss functions SVM review PRIMAL-DUAL PROBLEM Max-min

More information

Data Mining Practical Machine Learning Tools and Techniques. Slides for Chapter 6 of Data Mining by I. H. Witten and E. Frank

Data Mining Practical Machine Learning Tools and Techniques. Slides for Chapter 6 of Data Mining by I. H. Witten and E. Frank Data Mining Practical Machine Learning Tools and Techniques Slides for Chapter 6 of Data Mining by I. H. Witten and E. Frank Implementation: Real machine learning schemes Decision trees Classification

More information

Kernels + K-Means Introduction to Machine Learning. Matt Gormley Lecture 29 April 25, 2018

Kernels + K-Means Introduction to Machine Learning. Matt Gormley Lecture 29 April 25, 2018 10-601 Introduction to Machine Learning Machine Learning Department School of Computer Science Carnegie Mellon University Kernels + K-Means Matt Gormley Lecture 29 April 25, 2018 1 Reminders Homework 8:

More information

Face Recognition via Sparse Representation

Face Recognition via Sparse Representation Face Recognition via Sparse Representation John Wright, Allen Y. Yang, Arvind, S. Shankar Sastry and Yi Ma IEEE Trans. PAMI, March 2008 Research About Face Face Detection Face Alignment Face Recognition

More information

Support Vector Machines

Support Vector Machines Support Vector Machines Chapter 9 Chapter 9 1 / 50 1 91 Maximal margin classifier 2 92 Support vector classifiers 3 93 Support vector machines 4 94 SVMs with more than two classes 5 95 Relationshiop to

More information

Selecting Models from Videos for Appearance-Based Face Recognition

Selecting Models from Videos for Appearance-Based Face Recognition Selecting Models from Videos for Appearance-Based Face Recognition Abdenour Hadid and Matti Pietikäinen Machine Vision Group Infotech Oulu and Department of Electrical and Information Engineering P.O.

More information

Robot Learning. There are generally three types of robot learning: Learning from data. Learning by demonstration. Reinforcement learning

Robot Learning. There are generally three types of robot learning: Learning from data. Learning by demonstration. Reinforcement learning Robot Learning 1 General Pipeline 1. Data acquisition (e.g., from 3D sensors) 2. Feature extraction and representation construction 3. Robot learning: e.g., classification (recognition) or clustering (knowledge

More information

Support Vector Machines.

Support Vector Machines. Support Vector Machines srihari@buffalo.edu SVM Discussion Overview 1. Overview of SVMs 2. Margin Geometry 3. SVM Optimization 4. Overlapping Distributions 5. Relationship to Logistic Regression 6. Dealing

More information

Classification by Support Vector Machines

Classification by Support Vector Machines Classification by Support Vector Machines Florian Markowetz Max-Planck-Institute for Molecular Genetics Computational Molecular Biology Berlin Practical DNA Microarray Analysis 2003 1 Overview I II III

More information

A Short SVM (Support Vector Machine) Tutorial

A Short SVM (Support Vector Machine) Tutorial A Short SVM (Support Vector Machine) Tutorial j.p.lewis CGIT Lab / IMSC U. Southern California version 0.zz dec 004 This tutorial assumes you are familiar with linear algebra and equality-constrained optimization/lagrange

More information

Chap.12 Kernel methods [Book, Chap.7]

Chap.12 Kernel methods [Book, Chap.7] Chap.12 Kernel methods [Book, Chap.7] Neural network methods became popular in the mid to late 1980s, but by the mid to late 1990s, kernel methods have also become popular in machine learning. The first

More information

FACE RECOGNITION USING SUPPORT VECTOR MACHINES

FACE RECOGNITION USING SUPPORT VECTOR MACHINES FACE RECOGNITION USING SUPPORT VECTOR MACHINES Ashwin Swaminathan ashwins@umd.edu ENEE633: Statistical and Neural Pattern Recognition Instructor : Prof. Rama Chellappa Project 2, Part (b) 1. INTRODUCTION

More information

A Distance-Based Classifier Using Dissimilarity Based on Class Conditional Probability and Within-Class Variation. Kwanyong Lee 1 and Hyeyoung Park 2

A Distance-Based Classifier Using Dissimilarity Based on Class Conditional Probability and Within-Class Variation. Kwanyong Lee 1 and Hyeyoung Park 2 A Distance-Based Classifier Using Dissimilarity Based on Class Conditional Probability and Within-Class Variation Kwanyong Lee 1 and Hyeyoung Park 2 1. Department of Computer Science, Korea National Open

More information

Content-based image and video analysis. Machine learning

Content-based image and video analysis. Machine learning Content-based image and video analysis Machine learning for multimedia retrieval 04.05.2009 What is machine learning? Some problems are very hard to solve by writing a computer program by hand Almost all

More information

CLASSIFICATION WITH RADIAL BASIS AND PROBABILISTIC NEURAL NETWORKS

CLASSIFICATION WITH RADIAL BASIS AND PROBABILISTIC NEURAL NETWORKS CLASSIFICATION WITH RADIAL BASIS AND PROBABILISTIC NEURAL NETWORKS CHAPTER 4 CLASSIFICATION WITH RADIAL BASIS AND PROBABILISTIC NEURAL NETWORKS 4.1 Introduction Optical character recognition is one of

More information

A Hybrid Face Recognition Approach Using GPUMLib

A Hybrid Face Recognition Approach Using GPUMLib A Hybrid Face Recognition Approach Using GPUMLib Noel Lopes 1,2 and Bernardete Ribeiro 1 1 CISUC - Center for Informatics and Systems of University of Coimbra, Portugal 2 UDI/IPG - Research Unit, Polytechnic

More information

Machine Learning: Think Big and Parallel

Machine Learning: Think Big and Parallel Day 1 Inderjit S. Dhillon Dept of Computer Science UT Austin CS395T: Topics in Multicore Programming Oct 1, 2013 Outline Scikit-learn: Machine Learning in Python Supervised Learning day1 Regression: Least

More information

Leave-One-Out Support Vector Machines

Leave-One-Out Support Vector Machines Leave-One-Out Support Vector Machines Jason Weston Department of Computer Science Royal Holloway, University of London, Egham Hill, Egham, Surrey, TW20 OEX, UK. Abstract We present a new learning algorithm

More information

Transductive Learning: Motivation, Model, Algorithms

Transductive Learning: Motivation, Model, Algorithms Transductive Learning: Motivation, Model, Algorithms Olivier Bousquet Centre de Mathématiques Appliquées Ecole Polytechnique, FRANCE olivier.bousquet@m4x.org University of New Mexico, January 2002 Goal

More information

SUPPORT VECTOR MACHINES

SUPPORT VECTOR MACHINES SUPPORT VECTOR MACHINES Today Reading AIMA 8.9 (SVMs) Goals Finish Backpropagation Support vector machines Backpropagation. Begin with randomly initialized weights 2. Apply the neural network to each training

More information

Support Vector Machines

Support Vector Machines Support Vector Machines . Importance of SVM SVM is a discriminative method that brings together:. computational learning theory. previously known methods in linear discriminant functions 3. optimization

More information

Feature Extractors. CS 188: Artificial Intelligence Fall Some (Vague) Biology. The Binary Perceptron. Binary Decision Rule.

Feature Extractors. CS 188: Artificial Intelligence Fall Some (Vague) Biology. The Binary Perceptron. Binary Decision Rule. CS 188: Artificial Intelligence Fall 2008 Lecture 24: Perceptrons II 11/24/2008 Dan Klein UC Berkeley Feature Extractors A feature extractor maps inputs to feature vectors Dear Sir. First, I must solicit

More information

The Comparative Study of Machine Learning Algorithms in Text Data Classification*

The Comparative Study of Machine Learning Algorithms in Text Data Classification* The Comparative Study of Machine Learning Algorithms in Text Data Classification* Wang Xin School of Science, Beijing Information Science and Technology University Beijing, China Abstract Classification

More information

Non-Bayesian Classifiers Part II: Linear Discriminants and Support Vector Machines

Non-Bayesian Classifiers Part II: Linear Discriminants and Support Vector Machines Non-Bayesian Classifiers Part II: Linear Discriminants and Support Vector Machines Selim Aksoy Department of Computer Engineering Bilkent University saksoy@cs.bilkent.edu.tr CS 551, Spring 2007 c 2007,

More information

GENDER CLASSIFICATION USING SUPPORT VECTOR MACHINES

GENDER CLASSIFICATION USING SUPPORT VECTOR MACHINES GENDER CLASSIFICATION USING SUPPORT VECTOR MACHINES Ashwin Swaminathan ashwins@umd.edu ENEE633: Statistical and Neural Pattern Recognition Instructor : Prof. Rama Chellappa Project 2, Part (a) 1. INTRODUCTION

More information

Unsupervised learning in Vision

Unsupervised learning in Vision Chapter 7 Unsupervised learning in Vision The fields of Computer Vision and Machine Learning complement each other in a very natural way: the aim of the former is to extract useful information from visual

More information

Learning based face hallucination techniques: A survey

Learning based face hallucination techniques: A survey Vol. 3 (2014-15) pp. 37-45. : A survey Premitha Premnath K Department of Computer Science & Engineering Vidya Academy of Science & Technology Thrissur - 680501, Kerala, India (email: premithakpnath@gmail.com)

More information

Lab 2: Support vector machines

Lab 2: Support vector machines Artificial neural networks, advanced course, 2D1433 Lab 2: Support vector machines Martin Rehn For the course given in 2006 All files referenced below may be found in the following directory: /info/annfk06/labs/lab2

More information

Content Based Image Retrieval system with a combination of Rough Set and Support Vector Machine

Content Based Image Retrieval system with a combination of Rough Set and Support Vector Machine Shahabi Lotfabadi, M., Shiratuddin, M.F. and Wong, K.W. (2013) Content Based Image Retrieval system with a combination of rough set and support vector machine. In: 9th Annual International Joint Conferences

More information

Multi-label Classification. Jingzhou Liu Dec

Multi-label Classification. Jingzhou Liu Dec Multi-label Classification Jingzhou Liu Dec. 6 2016 Introduction Multi-class problem, Training data (x $, y $ ) ( ), x $ X R., y $ Y = 1,2,, L Learn a mapping f: X Y Each instance x $ is associated with

More information

HW2 due on Thursday. Face Recognition: Dimensionality Reduction. Biometrics CSE 190 Lecture 11. Perceptron Revisited: Linear Separators

HW2 due on Thursday. Face Recognition: Dimensionality Reduction. Biometrics CSE 190 Lecture 11. Perceptron Revisited: Linear Separators HW due on Thursday Face Recognition: Dimensionality Reduction Biometrics CSE 190 Lecture 11 CSE190, Winter 010 CSE190, Winter 010 Perceptron Revisited: Linear Separators Binary classification can be viewed

More information

A Fuzzy C-means Clustering Algorithm Based on Pseudo-nearest-neighbor Intervals for Incomplete Data

A Fuzzy C-means Clustering Algorithm Based on Pseudo-nearest-neighbor Intervals for Incomplete Data Journal of Computational Information Systems 11: 6 (2015) 2139 2146 Available at http://www.jofcis.com A Fuzzy C-means Clustering Algorithm Based on Pseudo-nearest-neighbor Intervals for Incomplete Data

More information

Lecture Linear Support Vector Machines

Lecture Linear Support Vector Machines Lecture 8 In this lecture we return to the task of classification. As seen earlier, examples include spam filters, letter recognition, or text classification. In this lecture we introduce a popular method

More information

Univariate Margin Tree

Univariate Margin Tree Univariate Margin Tree Olcay Taner Yıldız Department of Computer Engineering, Işık University, TR-34980, Şile, Istanbul, Turkey, olcaytaner@isikun.edu.tr Abstract. In many pattern recognition applications,

More information

Bagging for One-Class Learning

Bagging for One-Class Learning Bagging for One-Class Learning David Kamm December 13, 2008 1 Introduction Consider the following outlier detection problem: suppose you are given an unlabeled data set and make the assumptions that one

More information

A generalized quadratic loss for Support Vector Machines

A generalized quadratic loss for Support Vector Machines A generalized quadratic loss for Support Vector Machines Filippo Portera and Alessandro Sperduti Abstract. The standard SVM formulation for binary classification is based on the Hinge loss function, where

More information

Random projection for non-gaussian mixture models

Random projection for non-gaussian mixture models Random projection for non-gaussian mixture models Győző Gidófalvi Department of Computer Science and Engineering University of California, San Diego La Jolla, CA 92037 gyozo@cs.ucsd.edu Abstract Recently,

More information

Using K-NN SVMs for Performance Improvement and Comparison to K-Highest Lagrange Multipliers Selection

Using K-NN SVMs for Performance Improvement and Comparison to K-Highest Lagrange Multipliers Selection Using K-NN SVMs for Performance Improvement and Comparison to K-Highest Lagrange Multipliers Selection Sedat Ozer, Chi Hau Chen 2, and Imam Samil Yetik 3 Electrical & Computer Eng. Dept, Rutgers University,

More information

Preliminary Local Feature Selection by Support Vector Machine for Bag of Features

Preliminary Local Feature Selection by Support Vector Machine for Bag of Features Preliminary Local Feature Selection by Support Vector Machine for Bag of Features Tetsu Matsukawa Koji Suzuki Takio Kurita :University of Tsukuba :National Institute of Advanced Industrial Science and

More information

SVM in Analysis of Cross-Sectional Epidemiological Data Dmitriy Fradkin. April 4, 2005 Dmitriy Fradkin, Rutgers University Page 1

SVM in Analysis of Cross-Sectional Epidemiological Data Dmitriy Fradkin. April 4, 2005 Dmitriy Fradkin, Rutgers University Page 1 SVM in Analysis of Cross-Sectional Epidemiological Data Dmitriy Fradkin April 4, 2005 Dmitriy Fradkin, Rutgers University Page 1 Overview The goals of analyzing cross-sectional data Standard methods used

More information

Generative and discriminative classification techniques

Generative and discriminative classification techniques Generative and discriminative classification techniques Machine Learning and Category Representation 013-014 Jakob Verbeek, December 13+0, 013 Course website: http://lear.inrialpes.fr/~verbeek/mlcr.13.14

More information

The Analysis of Parameters t and k of LPP on Several Famous Face Databases

The Analysis of Parameters t and k of LPP on Several Famous Face Databases The Analysis of Parameters t and k of LPP on Several Famous Face Databases Sujing Wang, Na Zhang, Mingfang Sun, and Chunguang Zhou College of Computer Science and Technology, Jilin University, Changchun

More information

Edge-Directed Image Interpolation Using Color Gradient Information

Edge-Directed Image Interpolation Using Color Gradient Information Edge-Directed Image Interpolation Using Color Gradient Information Andrey Krylov and Andrey Nasonov Laboratory of Mathematical Methods of Image Processing, Faculty of Computational Mathematics and Cybernetics,

More information

A Weighted Non-negative Matrix Factorization for Local Representations Λ

A Weighted Non-negative Matrix Factorization for Local Representations Λ A Weighted Non-negative Matrix Factorization for Local Representations Λ David Guillamet, Marco Bressan and Jordi Vitrià Centre de Visió per Computador, Dept. Informàtica Universitat Autònoma de Barcelona,

More information

Support Vector Regression for Software Reliability Growth Modeling and Prediction

Support Vector Regression for Software Reliability Growth Modeling and Prediction Support Vector Regression for Software Reliability Growth Modeling and Prediction 925 Fei Xing 1 and Ping Guo 2 1 Department of Computer Science Beijing Normal University, Beijing 100875, China xsoar@163.com

More information

Robustness of Selective Desensitization Perceptron Against Irrelevant and Partially Relevant Features in Pattern Classification

Robustness of Selective Desensitization Perceptron Against Irrelevant and Partially Relevant Features in Pattern Classification Robustness of Selective Desensitization Perceptron Against Irrelevant and Partially Relevant Features in Pattern Classification Tomohiro Tanno, Kazumasa Horie, Jun Izawa, and Masahiko Morita University

More information

An Introduction to Machine Learning

An Introduction to Machine Learning TRIPODS Summer Boorcamp: Topology and Machine Learning August 6, 2018 General Set-up Introduction Set-up and Goal Suppose we have X 1,X 2,...,X n data samples. Can we predict properites about any given

More information

A Two-phase Distributed Training Algorithm for Linear SVM in WSN

A Two-phase Distributed Training Algorithm for Linear SVM in WSN Proceedings of the World Congress on Electrical Engineering and Computer Systems and Science (EECSS 015) Barcelona, Spain July 13-14, 015 Paper o. 30 A wo-phase Distributed raining Algorithm for Linear

More information

Training-Free, Generic Object Detection Using Locally Adaptive Regression Kernels

Training-Free, Generic Object Detection Using Locally Adaptive Regression Kernels Training-Free, Generic Object Detection Using Locally Adaptive Regression Kernels IEEE TRANSACTIONS ON PATTERN ANALYSIS AND MACHINE INTELLIENCE, VOL.32, NO.9, SEPTEMBER 2010 Hae Jong Seo, Student Member,

More information

CSE 573: Artificial Intelligence Autumn 2010

CSE 573: Artificial Intelligence Autumn 2010 CSE 573: Artificial Intelligence Autumn 2010 Lecture 16: Machine Learning Topics 12/7/2010 Luke Zettlemoyer Most slides over the course adapted from Dan Klein. 1 Announcements Syllabus revised Machine

More information

Data mining with sparse grids

Data mining with sparse grids Data mining with sparse grids Jochen Garcke and Michael Griebel Institut für Angewandte Mathematik Universität Bonn Data mining with sparse grids p.1/40 Overview What is Data mining? Regularization networks

More information

Fast and Effective Interpolation Using Median Filter

Fast and Effective Interpolation Using Median Filter Fast and Effective Interpolation Using Median Filter Jian Zhang 1, *, Siwei Ma 2, Yongbing Zhang 1, and Debin Zhao 1 1 Department of Computer Science, Harbin Institute of Technology, Harbin 150001, P.R.

More information

Fine Classification of Unconstrained Handwritten Persian/Arabic Numerals by Removing Confusion amongst Similar Classes

Fine Classification of Unconstrained Handwritten Persian/Arabic Numerals by Removing Confusion amongst Similar Classes 2009 10th International Conference on Document Analysis and Recognition Fine Classification of Unconstrained Handwritten Persian/Arabic Numerals by Removing Confusion amongst Similar Classes Alireza Alaei

More information

Information Management course

Information Management course Università degli Studi di Milano Master Degree in Computer Science Information Management course Teacher: Alberto Ceselli Lecture 20: 10/12/2015 Data Mining: Concepts and Techniques (3 rd ed.) Chapter

More information

Support Vector Machines

Support Vector Machines Support Vector Machines About the Name... A Support Vector A training sample used to define classification boundaries in SVMs located near class boundaries Support Vector Machines Binary classifiers whose

More information

5 Learning hypothesis classes (16 points)

5 Learning hypothesis classes (16 points) 5 Learning hypothesis classes (16 points) Consider a classification problem with two real valued inputs. For each of the following algorithms, specify all of the separators below that it could have generated

More information

Edge-directed Image Interpolation Using Color Gradient Information

Edge-directed Image Interpolation Using Color Gradient Information Edge-directed Image Interpolation Using Color Gradient Information Andrey Krylov and Andrey Nasonov Laboratory of Mathematical Methods of Image Processing, Faculty of Computational Mathematics and Cybernetics,

More information

Some questions of consensus building using co-association

Some questions of consensus building using co-association Some questions of consensus building using co-association VITALIY TAYANOV Polish-Japanese High School of Computer Technics Aleja Legionow, 4190, Bytom POLAND vtayanov@yahoo.com Abstract: In this paper

More information

ECG782: Multidimensional Digital Signal Processing

ECG782: Multidimensional Digital Signal Processing ECG782: Multidimensional Digital Signal Processing Object Recognition http://www.ee.unlv.edu/~b1morris/ecg782/ 2 Outline Knowledge Representation Statistical Pattern Recognition Neural Networks Boosting

More information

Unsupervised Feature Selection for Sparse Data

Unsupervised Feature Selection for Sparse Data Unsupervised Feature Selection for Sparse Data Artur Ferreira 1,3 Mário Figueiredo 2,3 1- Instituto Superior de Engenharia de Lisboa, Lisboa, PORTUGAL 2- Instituto Superior Técnico, Lisboa, PORTUGAL 3-

More information

Lab 2: Support Vector Machines

Lab 2: Support Vector Machines Articial neural networks, advanced course, 2D1433 Lab 2: Support Vector Machines March 13, 2007 1 Background Support vector machines, when used for classication, nd a hyperplane w, x + b = 0 that separates

More information

Classification: Feature Vectors

Classification: Feature Vectors Classification: Feature Vectors Hello, Do you want free printr cartriges? Why pay more when you can get them ABSOLUTELY FREE! Just # free YOUR_NAME MISSPELLED FROM_FRIEND... : : : : 2 0 2 0 PIXEL 7,12

More information

Generating the Reduced Set by Systematic Sampling

Generating the Reduced Set by Systematic Sampling Generating the Reduced Set by Systematic Sampling Chien-Chung Chang and Yuh-Jye Lee Email: {D9115009, yuh-jye}@mail.ntust.edu.tw Department of Computer Science and Information Engineering National Taiwan

More information

Local Linear Approximation for Kernel Methods: The Railway Kernel

Local Linear Approximation for Kernel Methods: The Railway Kernel Local Linear Approximation for Kernel Methods: The Railway Kernel Alberto Muñoz 1,JavierGonzález 1, and Isaac Martín de Diego 1 University Carlos III de Madrid, c/ Madrid 16, 890 Getafe, Spain {alberto.munoz,

More information

Support Vector Machines + Classification for IR

Support Vector Machines + Classification for IR Support Vector Machines + Classification for IR Pierre Lison University of Oslo, Dep. of Informatics INF3800: Søketeknologi April 30, 2014 Outline of the lecture Recap of last week Support Vector Machines

More information

Machine Learning for NLP

Machine Learning for NLP Machine Learning for NLP Support Vector Machines Aurélie Herbelot 2018 Centre for Mind/Brain Sciences University of Trento 1 Support Vector Machines: introduction 2 Support Vector Machines (SVMs) SVMs

More information

3D Object Recognition using Multiclass SVM-KNN

3D Object Recognition using Multiclass SVM-KNN 3D Object Recognition using Multiclass SVM-KNN R. Muralidharan, C. Chandradekar April 29, 2014 Presented by: Tasadduk Chowdhury Problem We address the problem of recognizing 3D objects based on various

More information