Unsupervised Learning for Hierarchical Clustering Using Statistical Information
|
|
- Herbert Hubbard
- 5 years ago
- Views:
Transcription
1 Unsupervised Learning for Hierarcical Clustering Using Statistical Information Masaru Okamoto, Nan Bu, and Tosio Tsuji Department of Artificial Complex System Engineering Hirosima University Kagamiyama 1-4-1, Higasi-Hirosima, Hirosima, JAPAN {okamoto, bu, ttp:// Abstract. Tis paper proposes a novel ierarcical clustering metod tat can classify given data witout specified knowledge of te number of classes. In tis metod, at eac node of a ierarcical classification tree, log-linearized Gaussian mixture networks [2] are utilized as classifiers to divide data into two subclasses based on statistical information, wic are ten classified into secondary subclasses and so on. Also, unnecessary structure of te tree can be avoided by training in a cross-validation manner. Validity of te proposed metod is demonstrated wit classification experiments on artificial data. 1 Introduction Recently, tere ave been growing interests in using bioelectric signals suc as electromyogram (EMG) to conduct man-macine interface. In order to discriminate an operater s intentions from bioelectric signals efficiently, several attempts ave been made so far [1], [2]. Generally, suc pattern discrimination is performed by estimating te relationsip between te bioelectric signals as feature vectors and te corresponding intentions as class labels. However, difference between classes in te bioelectric signals of elderly or andicapped people is ambiguous, and tis relates to poor reliability of te class labels available. To overcome tis problem, clustering analysis as been widely adopted, in wic a collection of patterns is organized into clusters based on similarity. Te previously proposed clustering analysis tecniques can be dicotomized as eiter k-means algoritm or ierarcical clustering. Te k-means algoritm identifies a partition of te input space. On te oter and, te ierarcical clustering performs a nested series of partitions and finally performs a grouping wit a suitable number of classes. Also, in order to determine te number of class automatically, a clustering algoritm using self organizing maps (SOM) [5] as been proposed [6]. In tis metod, estimation of te number of classes is carried out based on te number of te data belonging to eac node of SOM. However, wen parameters in tis metod were not set up appropriately, suc metod may fail to perform satisfying clustering for complicated data. F. Yin, J. Wang, and C. Guo (Eds.): ISNN 2004, LNCS 3173, pp , c Springer-Verlag Berlin Heidelberg 2004
2 Unsupervised Learning for Hierarcical Clustering 835 X1 (1,1) (1) w1 O1 1,1 O1,1 O1 x1 x2 xd Nonlinear transformation X2 X X H (c,m) w (1) OH 1,M1 c,m C,MC Oc,m O1,M 1 Oc OC OC,M C Fig. 1. Structure of LLGMN In tis paper, a novel ierarcical clustering metod is proposed. In tis metod, a probabilistic NN tat is derived from te Gaussian mixture model (GMM), called a log-linearized Gaussian mixture network (LLGMN) [2], is utilized for partition at eac non-terminal node. Te proposed metod can estimate te number of terminal nodes corresponding to te number of classes according to te statistical information obtained solely from te training data. 2 LLGMN 2.1 Structure of LLGMN Te structure of LLGMN is sown in Fig.1. First, an input vector x R D is converted into a modified vector X as follows: X =[1, x T,x 2 1,x 1 x 2,,x 1 x D,x 2 2,x 2 x 3,,x 2 x D,,x 2 D] T. (1) Te first layer consists of H units corresponding to te dimension of X, and an identity function is used for an activation function of eac unit. Te variable (1) O ( = 1,,H) denotes te output of te t unit in te first layer. Eac unit in te second layer receives te output (1) O weigted by coefficients (c =1,,C; m =1,,M c ). C denotes te number of classes and M c is te number of components belonging to class c. Te relationsip between te input I c,m and te output O c,m of unit {c, m} in te second layer can be defined as w (c,m) I c,m = O c,m = H =1 (1) O w (c,m), exp[ I c,m ] C Mc, c =1 m =1 exp[ I c,m ]
3 836 M. Okamoto, N. Bu, and T. Tsuji were w (C,M C) = 0. Te relationsip between te input I c and te output is described as, O c = I c = M c O c,m. (4) Te output of te tird layer O c P (c x) of class c. m=1 corresponds to te posterior probability 2.2 Supervised Learning Algoritm [2] Consider a training set {x (n), T (n) } (n =1,,N), were T (n) = {T (n) 1,, T (n) C }. If te input vector x(n) belongs to class c, T c (n) = 1, and T (n) c = 0 for all of te oter class c. An energy function according to te minimum log-likeliood training criterion can be derived as: J SV = N C n=1 c=1 T c (n) log O c (n). (5) In te training process, modification of te LLGMN s weigt w (c,m) is defined as: N w (c,m) JSV n = η, (6) n=1 w (c,m) JSV n =( O w (c,m) c,m (n) were η>0 is te learning rate. 3 Hierarcical Clustering O c,m (n) O c (n) T c (n) )X (n), (7) Te divisive clustering starts from a single cluster, and terminates wen a termination criterion as been satisfied, so tat te training data are divided into te appropriate number of clusters. At eac non-terminal node, LLGMN is used to acieve binary splits. Even for data of complicated distributions, interpretable clustering can be made after a nested series of binary splits. In tis section, after te description of te proposed unsupervised learning algoritm of LLGMN, division validation according to te statistical properties of te training data and pruning law are explained. 3.1 Unsupervised Learning Algoritm Given te number of classes, C, te entropy used as cost function is defined as: J SO = N C n=1 c=1 O c (n) log O c (n), (8)
4 Unsupervised Learning for Hierarcical Clustering 837 were N is te number of total data. Te proposed unsupervised learning algoritm modifies weigts by minimizing Eq. (8), However, for some initial weigts, te LLGMN may be trained to cluster all training data into one class, and cost function, J SO, may converge to suc a local minimum. Terefore, in te proposed metod, te initialization of te weigts is carried out to prevent te LLGMN to converge to one of te local minima, and te number of classes is restricted to two. Let us consider tat LLGMN clusters data into two classes: C 1 and C 2. First, x 1 and x 2 are cosen for te initialization of weigts from te total training data set A according to te following equation, (x 1, x 2 ) = argmax ( x (i) x (j) ). (9) x (i),x (j) A Ten te set B, wic means te set of te utilized data, is set as {x 1, x 2 }. Assuming tat x 1 and x 2 are labeled wit C 1 and C 2, respectively. Training of LLGMN is performed using te supervised learning rule [2] in order to classify x 1 and x 2 into C 1 and C 2 respectively. Ten, wit te initialized weigts, unsupervised learning of te LLGMN is performed using te set B. Te mean values of x C1 and x C2 are calculated using te training data clustered into C 1 and C 2, respectively. One datum x A B, from wic te distance to eiter x C1 or x C2 w (c,m) is te smallest, is added into te set B. Ten, modification of te weigt is defined as: w (c,m) J (n) SO w (c,m) = η J(n) SO w (c,m), (10) = (J SO log O c (n) ) O c,mx (n) (n). (11) After training wit a pre-defined number of times, anoter training datum is selected from te set A B and added into te set B. Tis step of training repeats, until all te training data is added into B, tat is to say, B = A. 3.2 Division Validation Wit te proposed metod, unnecessary splits may occur wen te ierarcy of te tree becomes too deep. In tis metod, cross-validation is adopted and te posterior probabilities of te validation data is utilized to determine weter to split a node or not. First, te validation data is prepared and te entropy H(x) is defined as: C H(x) = O c (n) log O c (n). (12) c=1 Ten, te average value H E of H(x) is utilized as te termination criterion. H E = 1 H(x (n) ), (13) N c x (n) N c
5 838 M. Okamoto, N. Bu, and T. Tsuji C2 C4 C5 y C3 0.2 C x Fig. 2. Examples of artificial data were N c stands for te set of validation data belonging to te node under consideration, and N c is te number of validation data in N c.ifh E is iger tan a tresold H T, splitting of te corresponding node is terminated. On te oter and, if all validation data of te node in consideration are clustered into one class, outliers may exist in te training data and te division of tis node must be terminated. Also, for occasions wen tere is only one training data in a node, furter split of tis node must be terminated, since division is impossible. Wit tis validation, a classification tree can be constructed based on te statistical properties of te training data, and can cluster complicated data into a proper number of classes. 3.3 Pruning Law In te proposed metod, outliers are always classified into some terminal nodes (clusters) separated from oter major clusters. Especially, wen te ierarcy of te tree grows too large, te influence of outliers becomes prominent because of a decrease of te number of training data in eac node. After te classification tree is constructed, pruning is conducted to improve te clustering efficiency. Te number of training data left in eac terminal node is utilized as a decision index of pruning. If te ratio of te number of training data in a terminal node to te total training data number is lower tan tresold α T, tis node and its counter are merged into teir fater node. Wit tis pruning law, excessive splits may be prevented, and te number of clustering may not increase corresponding to te number of outlier data. 3.4 Experiments Numerical simulations were carried out in order to verify te proposed metod. Te feature data is illustrated in Fig. 2: Tere are 2-dimensional data x R 2, and generated from five classes, C i (i =1, 2,, 5). Eac class consists of one normal distribution. Te number of training data for eac class is 100, and te number of validation data for eac class is 200. Te LLGMN includes seven units
6 Unsupervised Learning for Hierarcical Clustering 839 in te first layer, two units in te second layer corresponding to te total number of components, and two units in te tird layer. To construct te classification tree, tresold of entropy H T is set as 0.2, tresold of pruning α t as 0.01, learning rate η as 0.01, and training times in eac addition of training data as 100. Te classification tree starts from te root node, were training data are divided into two nodes at eac non-terminal node, and finally, a ierarcical tree is constructed from five terminal nodes. To validate te generalization ability, 300 samples for eac class tat are not used in training process were clusterd, and te discrimination rate for 20 independent trials was 98.5 ± 0.64%. It can be found tat te proposed metod can estimate te number of classes and acieve ig classfication rate. 4 Conclusion In tis paper, to deal wit te discrimination problem of ambiguous teacer signals, a ierarcical clustering metod was proposed. In tis metod, entoropy of te LLGMN s outputs at eac node are used as te termination criterion, and unnecessary splits in te structure of te classification tree can be avoided, so tat te proposed metod can make an interpretable and reasonable partition of te training data according solely to its statistical caracteristics. In future works, we would like to carry out discrimination experiments on various data and to examine te influence of te parameters to te clustering result. Furtermore, we would like to establis an improved metod tat determines te value of teresolds suc as α T automatically. References 1. Hiraiwa, A., Simoara, K., Tokunaga, Y.: EMG Pattern Analysis and Classification by Neural Network. IEEE International Conference on Syst., Man and Cybern., (1989) Tsuji, T., Fukuda, O., Icinobe, H., Kaneko, M.: A Log-linearized Gaussian Mixture Network and its Application to EEG Pattern Classification. IEEE Trans. on System, Man and Cybernetics-Part C: Applications and Reviews, Vol. 29., No. 1. (1999) Anderberg, M.R.: Cluster Analysis for Applications. Academic Press, New York (1974) 4. Ward, J.H.: Hierarcical Grouping to Optimize an Objective Function. Journal of te American Statistical Association, Vol. 58., No (1963) Koonen, T.: Self-organization and Associative Memory. Tird Edition, Springer- Verlag, Berlin (1994) 6. Terasima, M., Siratani, F., Yamamoto, K.: Unsupervised Cluster Segmentation Metod Using Data Density Histogram on Self-organizing Feature Map. IEICE Transactions on Information and Systems, PT. 2, Vol. J79., No. 7., (1996) (in Japanese)
Bounding Tree Cover Number and Positive Semidefinite Zero Forcing Number
Bounding Tree Cover Number and Positive Semidefinite Zero Forcing Number Sofia Burille Mentor: Micael Natanson September 15, 2014 Abstract Given a grap, G, wit a set of vertices, v, and edges, various
More informationFast Calculation of Thermodynamic Properties of Water and Steam in Process Modelling using Spline Interpolation
P R E P R N T CPWS XV Berlin, September 8, 008 Fast Calculation of Termodynamic Properties of Water and Steam in Process Modelling using Spline nterpolation Mattias Kunick a, Hans-Joacim Kretzscmar a,
More informationProceedings of the 8th WSEAS International Conference on Neural Networks, Vancouver, British Columbia, Canada, June 19-21,
Proceedings of te 8t WSEAS International Conference on Neural Networks, Vancouver, Britis Columbia, Canada, June 9-2, 2007 3 Neural Network Structures wit Constant Weigts to Implement Dis-Jointly Removed
More informationCS 234. Module 6. October 25, CS 234 Module 6 ADT Dictionary 1 / 22
CS 234 Module 6 October 25, 2016 CS 234 Module 6 ADT Dictionary 1 / 22 Case study Problem: Find a way to store student records for a course, wit unique IDs for eac student, were records can be accessed,
More informationOur Calibrated Model has No Predictive Value: An Example from the Petroleum Industry
Our Calibrated Model as No Predictive Value: An Example from te Petroleum Industry J.N. Carter a, P.J. Ballester a, Z. Tavassoli a and P.R. King a a Department of Eart Sciences and Engineering, Imperial
More informationVector Processing Contours
Vector Processing Contours Andrey Kirsanov Department of Automation and Control Processes MAMI Moscow State Tecnical University Moscow, Russia AndKirsanov@yandex.ru A.Vavilin and K-H. Jo Department of
More informationTwo Modifications of Weight Calculation of the Non-Local Means Denoising Method
Engineering, 2013, 5, 522-526 ttp://dx.doi.org/10.4236/eng.2013.510b107 Publised Online October 2013 (ttp://www.scirp.org/journal/eng) Two Modifications of Weigt Calculation of te Non-Local Means Denoising
More informationLinear Interpolating Splines
Jim Lambers MAT 772 Fall Semester 2010-11 Lecture 17 Notes Tese notes correspond to Sections 112, 11, and 114 in te text Linear Interpolating Splines We ave seen tat ig-degree polynomial interpolation
More informationDesign of PSO-based Fuzzy Classification Systems
Tamkang Journal of Science and Engineering, Vol. 9, No 1, pp. 6370 (006) 63 Design of PSO-based Fuzzy Classification Systems Cia-Cong Cen Department of Electronics Engineering, Wufeng Institute of Tecnology,
More information, 1 1, A complex fraction is a quotient of rational expressions (including their sums) that result
RT. Complex Fractions Wen working wit algebraic expressions, sometimes we come across needing to simplify expressions like tese: xx 9 xx +, xx + xx + xx, yy xx + xx + +, aa Simplifying Complex Fractions
More informationMulti-Stack Boundary Labeling Problems
Multi-Stack Boundary Labeling Problems Micael A. Bekos 1, Micael Kaufmann 2, Katerina Potika 1 Antonios Symvonis 1 1 National Tecnical University of Atens, Scool of Applied Matematical & Pysical Sciences,
More informationClassification of Osteoporosis using Fractal Texture Features
Classification of Osteoporosis using Fractal Texture Features V.Srikant, C.Dines Kumar and A.Tobin Department of Electronics and Communication Engineering Panimalar Engineering College Cennai, Tamil Nadu,
More informationSoftware Fault Prediction using Machine Learning Algorithm Pooja Garg 1 Mr. Bhushan Dua 2
IJSRD - International Journal for Scientific Researc & Development Vol. 3, Issue 04, 2015 ISSN (online): 2321-0613 Software Fault Prediction using Macine Learning Algoritm Pooja Garg 1 Mr. Busan Dua 2
More informationAlternating Direction Implicit Methods for FDTD Using the Dey-Mittra Embedded Boundary Method
Te Open Plasma Pysics Journal, 2010, 3, 29-35 29 Open Access Alternating Direction Implicit Metods for FDTD Using te Dey-Mittra Embedded Boundary Metod T.M. Austin *, J.R. Cary, D.N. Smite C. Nieter Tec-X
More informationInvestigating an automated method for the sensitivity analysis of functions
Investigating an automated metod for te sensitivity analysis of functions Sibel EKER s.eker@student.tudelft.nl Jill SLINGER j..slinger@tudelft.nl Delft University of Tecnology 2628 BX, Delft, te Neterlands
More informationSymmetric Tree Replication Protocol for Efficient Distributed Storage System*
ymmetric Tree Replication Protocol for Efficient Distributed torage ystem* ung Cune Coi 1, Hee Yong Youn 1, and Joong up Coi 2 1 cool of Information and Communications Engineering ungkyunkwan University
More informationAn Algorithm for Loopless Deflection in Photonic Packet-Switched Networks
An Algoritm for Loopless Deflection in Potonic Packet-Switced Networks Jason P. Jue Center for Advanced Telecommunications Systems and Services Te University of Texas at Dallas Ricardson, TX 75083-0688
More informationUNSUPERVISED HIERARCHICAL IMAGE SEGMENTATION BASED ON THE TS-MRF MODEL AND FAST MEAN-SHIFT CLUSTERING
UNSUPERVISED HIERARCHICAL IMAGE SEGMENTATION BASED ON THE TS-MRF MODEL AND FAST MEAN-SHIFT CLUSTERING Raffaele Gaetano, Giuseppe Scarpa, Giovanni Poggi, and Josiane Zerubia Dip. Ing. Elettronica e Telecomunicazioni,
More information2.8 The derivative as a function
CHAPTER 2. LIMITS 56 2.8 Te derivative as a function Definition. Te derivative of f(x) istefunction f (x) defined as follows f f(x + ) f(x) (x). 0 Note: tis differs from te definition in section 2.7 in
More informationCommunicator for Mac Quick Start Guide
Communicator for Mac Quick Start Guide 503-968-8908 sterling.net training@sterling.net Pone Support 503.968.8908, option 2 pone-support@sterling.net For te most effective support, please provide your main
More informationSome Handwritten Signature Parameters in Biometric Recognition Process
Some Handwritten Signature Parameters in Biometric Recognition Process Piotr Porwik Institute of Informatics, Silesian Uniersity, Bdziska 39, 41- Sosnowiec, Poland porwik@us.edu.pl Tomasz Para Institute
More informationDensity Estimation Over Data Stream
Density Estimation Over Data Stream Aoying Zou Dept. of Computer Science, Fudan University 22 Handan Rd. Sangai, 2433, P.R. Cina ayzou@fudan.edu.cn Ziyuan Cai Dept. of Computer Science, Fudan University
More informationEnergy efficient temporal load aware resource allocation in cloud computing datacenters
Vakilinia Journal of Cloud Computing: Advances, Systems and Applications (2018) 7:2 DOI 10.1186/s13677-017-0103-2 Journal of Cloud Computing: Advances, Systems and Applications RESEARCH Energy efficient
More informationMinimizing Memory Access By Improving Register Usage Through High-level Transformations
Minimizing Memory Access By Improving Register Usage Troug Hig-level Transformations San Li Scool of Computer Engineering anyang Tecnological University anyang Avenue, SIGAPORE 639798 Email: p144102711@ntu.edu.sg
More informationGrid Adaptation for Functional Outputs: Application to Two-Dimensional Inviscid Flows
Journal of Computational Pysics 176, 40 69 (2002) doi:10.1006/jcp.2001.6967, available online at ttp://www.idealibrary.com on Grid Adaptation for Functional Outputs: Application to Two-Dimensional Inviscid
More informationUtilizing Call Admission Control to Derive Optimal Pricing of Multiple Service Classes in Wireless Cellular Networks
Utilizing Call Admission Control to Derive Optimal Pricing of Multiple Service Classes in Wireless Cellular Networks Okan Yilmaz and Ing-Ray Cen Computer Science Department Virginia Tec {oyilmaz, ircen}@vt.edu
More informationThe Euler and trapezoidal stencils to solve d d x y x = f x, y x
restart; Te Euler and trapezoidal stencils to solve d d x y x = y x Te purpose of tis workseet is to derive te tree simplest numerical stencils to solve te first order d equation y x d x = y x, and study
More informationAn Anchor Chain Scheme for IP Mobility Management
An Ancor Cain Sceme for IP Mobility Management Yigal Bejerano and Israel Cidon Department of Electrical Engineering Tecnion - Israel Institute of Tecnology Haifa 32000, Israel E-mail: bej@tx.tecnion.ac.il.
More informationTraffic Sign Classification Using Ring Partitioned Method
Traffic Sign Classification Using Ring Partitioned Metod Aryuanto Soetedjo and Koici Yamada Laboratory for Management and Information Systems Science, Nagaoa University of Tecnology 603- Kamitomioamaci,
More informationLaser Radar based Vehicle Localization in GPS Signal Blocked Areas
International Journal of Computational Intelligence Systems, Vol. 4, No. 6 (December, 20), 00-09 Laser Radar based Veicle Localization in GPS Signal Bloced Areas Ming Yang Department of Automation, Sangai
More informationCubic smoothing spline
Cubic smooting spline Menu: QCExpert Regression Cubic spline e module Cubic Spline is used to fit any functional regression curve troug data wit one independent variable x and one dependent random variable
More informationHASH ALGORITHMS: A DESIGN FOR PARALLEL CALCULATIONS
HASH ALGORITHMS: A DESIGN FOR PARALLEL CALCULATIONS N.G.Bardis Researc Associate Hellenic Ministry of te Interior, Public Administration and Decentralization 8, Dragatsaniou str., Klatmonos S. 0559, Greece
More informationComputing geodesic paths on manifolds
Proc. Natl. Acad. Sci. USA Vol. 95, pp. 8431 8435, July 1998 Applied Matematics Computing geodesic pats on manifolds R. Kimmel* and J. A. Setian Department of Matematics and Lawrence Berkeley National
More informationUUV DEPTH MEASUREMENT USING CAMERA IMAGES
ABCM Symposium Series in Mecatronics - Vol. 3 - pp.292-299 Copyrigt c 2008 by ABCM UUV DEPTH MEASUREMENT USING CAMERA IMAGES Rogerio Yugo Takimoto Graduate Scool of Engineering Yokoama National University
More informationCategory Detection Using Hierarchical Mean Shift
Category Detection Using Hierarcical Mean Sift Pavan Vatturi Scool of EECS 118 Kelley Engineering Center Oregon State University Corvallis, OR 97331 vatturi@eecs.oregonstate.edu Weng-Keen Wong Scool of
More informationNotes: Dimensional Analysis / Conversions
Wat is a unit system? A unit system is a metod of taking a measurement. Simple as tat. We ave units for distance, time, temperature, pressure, energy, mass, and many more. Wy is it important to ave a standard?
More informationComparison of the Efficiency of the Various Algorithms in Stratified Sampling when the Initial Solutions are Determined with Geometric Method
International Journal of Statistics and Applications 0, (): -0 DOI: 0.9/j.statistics.000.0 Comparison of te Efficiency of te Various Algoritms in Stratified Sampling wen te Initial Solutions are Determined
More informations e
1 Learning Hierarcical Partially Observable Markov Decision Process Models or Robot Navigation Georgios Teocarous Kasayar Roanimanes Sridar Maadevan teocar@cse.msu.edu kas@cse.msu.edu maadeva@cse.msu.edu
More informationPiecewise Polynomial Interpolation, cont d
Jim Lambers MAT 460/560 Fall Semester 2009-0 Lecture 2 Notes Tese notes correspond to Section 4 in te text Piecewise Polynomial Interpolation, cont d Constructing Cubic Splines, cont d Having determined
More informationYou Try: A. Dilate the following figure using a scale factor of 2 with center of dilation at the origin.
1 G.SRT.1-Some Tings To Know Dilations affect te size of te pre-image. Te pre-image will enlarge or reduce by te ratio given by te scale factor. A dilation wit a scale factor of 1> x >1enlarges it. A dilation
More informationTHE POSSIBILITY OF ESTIMATING THE VOLUME OF A SQUARE FRUSTRUM USING THE KNOWN VOLUME OF A CONICAL FRUSTRUM
THE POSSIBILITY OF ESTIMATING THE VOLUME OF A SQUARE FRUSTRUM USING THE KNOWN VOLUME OF A CONICAL FRUSTRUM SAMUEL OLU OLAGUNJU Adeyemi College of Education NIGERIA Email: lagsam04@aceondo.edu.ng ABSTRACT
More informationMore on Functions and Their Graphs
More on Functions and Teir Graps Difference Quotient ( + ) ( ) f a f a is known as te difference quotient and is used exclusively wit functions. Te objective to keep in mind is to factor te appearing in
More informationSearch-aware Conditions for Probably Approximately Correct Heuristic Search
Searc-aware Conditions for Probably Approximately Correct Heuristic Searc Roni Stern Ariel Felner Information Systems Engineering Ben Gurion University Beer-Seva, Israel 85104 roni.stern@gmail.com, felner@bgu.ac.il
More informationCS 234. Module 6. October 16, CS 234 Module 6 ADT Dictionary 1 / 33
CS 234 Module 6 October 16, 2018 CS 234 Module 6 ADT Dictionary 1 / 33 Idea for an ADT Te ADT Dictionary stores pairs (key, element), were keys are distinct and elements can be any data. Notes: Tis is
More informationIntegrating Constraints and Metric Learning in Semi-Supervised Clustering
Integrating Constraints and Metric Learning in Semi-Supervised Clustering Mikail Bilenko MBILENKO@CS.UTEXAS.EDU Sugato Basu SUGATO@CS.UTEXAS.EDU Raymond J. Mooney MOONEY@CS.UTEXAS.EDU Department of Computer
More informationOn the Use of Radio Resource Tests in Wireless ad hoc Networks
Tecnical Report RT/29/2009 On te Use of Radio Resource Tests in Wireless ad oc Networks Diogo Mónica diogo.monica@gsd.inesc-id.pt João Leitão jleitao@gsd.inesc-id.pt Luis Rodrigues ler@ist.utl.pt Carlos
More informationTest Generation for Acyclic Sequential Circuits with Hold Registers
Test Generation for Acyclic Sequential Circuits wit Hold Registers Tomoo Inoue, Debes Kumar Das, Ciio Sano, Takairo Miara, and Hideo Fujiwara Faculty of Information Sciences Computer Science and Engineering
More informationMAPI Computer Vision
MAPI Computer Vision Multiple View Geometry In tis module we intend to present several tecniques in te domain of te 3D vision Manuel Joao University of Mino Dep Industrial Electronics - Applications -
More informationHash-Based Indexes. Chapter 11. Comp 521 Files and Databases Spring
Has-Based Indexes Capter 11 Comp 521 Files and Databases Spring 2010 1 Introduction As for any index, 3 alternatives for data entries k*: Data record wit key value k
More informationCoarticulation: An Approach for Generating Concurrent Plans in Markov Decision Processes
Coarticulation: An Approac for Generating Concurrent Plans in Markov Decision Processes Kasayar Roanimanes kas@cs.umass.edu Sridar Maadevan maadeva@cs.umass.edu Department of Computer Science, University
More informationAn Interactive X-Ray Image Segmentation Technique for Bone Extraction
An Interactive X-Ray Image Segmentation Tecnique for Bone Extraction Cristina Stolojescu-Crisan and Stefan Holban Politenica University of Timisoara V. Parvan 2, 300223 Timisoara, Romania {cristina.stolojescu@etc.upt.ro
More informationIntroduction to Computer Graphics 5. Clipping
Introduction to Computer Grapics 5. Clipping I-Cen Lin, Assistant Professor National Ciao Tung Univ., Taiwan Textbook: E.Angel, Interactive Computer Grapics, 5 t Ed., Addison Wesley Ref:Hearn and Baker,
More informationAn Effective Sensor Deployment Strategy by Linear Density Control in Wireless Sensor Networks Chiming Huang and Rei-Heng Cheng
An ffective Sensor Deployment Strategy by Linear Density Control in Wireless Sensor Networks Ciming Huang and ei-heng Ceng 5 De c e mbe r0 International Journal of Advanced Information Tecnologies (IJAIT),
More informationPYRAMID FILTERS BASED ON BILINEAR INTERPOLATION
PYRAMID FILTERS BASED ON BILINEAR INTERPOLATION Martin Kraus Computer Grapics and Visualization Group, Tecnisce Universität Müncen, Germany krausma@in.tum.de Magnus Strengert Visualization and Interactive
More informationA UPnP-based Decentralized Service Discovery Improved Algorithm
Indonesian Journal of Electrical Engineering and Informatics (IJEEI) Vol.1, No.1, Marc 2013, pp. 21~26 ISSN: 2089-3272 21 A UPnP-based Decentralized Service Discovery Improved Algoritm Yu Si-cai*, Wu Yan-zi,
More informationCE 221 Data Structures and Algorithms
CE Data Structures and Algoritms Capter 4: Trees (AVL Trees) Text: Read Weiss, 4.4 Izmir University of Economics AVL Trees An AVL (Adelson-Velskii and Landis) tree is a binary searc tree wit a balance
More informationHash-Based Indexes. Chapter 11. Comp 521 Files and Databases Fall
Has-Based Indexes Capter 11 Comp 521 Files and Databases Fall 2012 1 Introduction Hasing maps a searc key directly to te pid of te containing page/page-overflow cain Doesn t require intermediate page fetces
More informationIntra- and Inter-Session Network Coding in Wireless Networks
Intra- and Inter-Session Network Coding in Wireless Networks Hulya Seferoglu, Member, IEEE, Atina Markopoulou, Member, IEEE, K K Ramakrisnan, Fellow, IEEE arxiv:857v [csni] 3 Feb Abstract In tis paper,
More informationFeature-Based Steganalysis for JPEG Images and its Implications for Future Design of Steganographic Schemes
Feature-Based Steganalysis for JPEG Images and its Implications for Future Design of Steganograpic Scemes Jessica Fridric Dept. of Electrical Engineering, SUNY Bingamton, Bingamton, NY 3902-6000, USA fridric@bingamton.edu
More informationEfficient Content-Based Indexing of Large Image Databases
Efficient Content-Based Indexing of Large Image Databases ESSAM A. EL-KWAE University of Nort Carolina at Carlotte and MANSUR R. KABUKA University of Miami Large image databases ave emerged in various
More informationGraph Matching: Fast Candidate Elimination Using Machine Learning Techniques
Graph Matching: Fast Candidate Elimination Using Machine Learning Techniques M. Lazarescu 1,2, H. Bunke 1, and S. Venkatesh 2 1 Computer Science Department, University of Bern, Switzerland 2 School of
More information3.6 Directional Derivatives and the Gradient Vector
288 CHAPTER 3. FUNCTIONS OF SEVERAL VARIABLES 3.6 Directional Derivatives and te Gradient Vector 3.6.1 Functions of two Variables Directional Derivatives Let us first quickly review, one more time, te
More informationCESILA: Communication Circle External Square Intersection-Based WSN Localization Algorithm
Sensors & Transducers 2013 by IFSA ttp://www.sensorsportal.com CESILA: Communication Circle External Square Intersection-Based WSN Localization Algoritm Sun Hongyu, Fang Ziyi, Qu Guannan College of Computer
More informationEfficient Classification of Data Using Decision Tree
Bonfring International Journal of Data Mining, Vol. 2, No. 1, Marc 2012 6 Efficient Classification of Data Using Decision Tree Baskar N. Patel, Satis G. Prajapati and Dr. Kamaljit I. Laktaria Abstract---
More informationFault Localization Using Tarantula
Class 20 Fault localization (cont d) Test-data generation Exam review: Nov 3, after class to :30 Responsible for all material up troug Nov 3 (troug test-data generation) Send questions beforeand so all
More informationOvercomplete Steerable Pyramid Filters and Rotation Invariance
vercomplete Steerable Pyramid Filters and Rotation Invariance H. Greenspan, S. Belongie R. Goodman and P. Perona S. Raksit and C. H. Anderson Department of Electrical Engineering Department of Anatomy
More informationWhen the dimensions of a solid increase by a factor of k, how does the surface area change? How does the volume change?
8.4 Surface Areas and Volumes of Similar Solids Wen te dimensions of a solid increase by a factor of k, ow does te surface area cange? How does te volume cange? 1 ACTIVITY: Comparing Surface Areas and
More information4.1 Tangent Lines. y 2 y 1 = y 2 y 1
41 Tangent Lines Introduction Recall tat te slope of a line tells us ow fast te line rises or falls Given distinct points (x 1, y 1 ) and (x 2, y 2 ), te slope of te line troug tese two points is cange
More informationMulti-Objective Particle Swarm Optimizers: A Survey of the State-of-the-Art
Multi-Objective Particle Swarm Optimizers: A Survey of te State-of-te-Art Margarita Reyes-Sierra and Carlos A. Coello Coello CINVESTAV-IPN (Evolutionary Computation Group) Electrical Engineering Department,
More informationSection 2.3: Calculating Limits using the Limit Laws
Section 2.3: Calculating Limits using te Limit Laws In previous sections, we used graps and numerics to approimate te value of a it if it eists. Te problem wit tis owever is tat it does not always give
More informationA Cost Model for Distributed Shared Memory. Using Competitive Update. Jai-Hoon Kim Nitin H. Vaidya. Department of Computer Science
A Cost Model for Distributed Sared Memory Using Competitive Update Jai-Hoon Kim Nitin H. Vaidya Department of Computer Science Texas A&M University College Station, Texas, 77843-3112, USA E-mail: fjkim,vaidyag@cs.tamu.edu
More informationOptimal In-Network Packet Aggregation Policy for Maximum Information Freshness
1 Optimal In-etwork Packet Aggregation Policy for Maimum Information Fresness Alper Sinan Akyurek, Tajana Simunic Rosing Electrical and Computer Engineering, University of California, San Diego aakyurek@ucsd.edu,
More informationA Novel QC-LDPC Code with Flexible Construction and Low Error Floor
A Novel QC-LDPC Code wit Flexile Construction and Low Error Floor Hanxin WANG,2, Saoping CHEN,2,CuitaoZHU,2 and Kaiyou SU Department of Electronics and Information Engineering, Sout-Central University
More informationTuning MAX MIN Ant System with off-line and on-line methods
Université Libre de Bruxelles Institut de Recerces Interdisciplinaires et de Développements en Intelligence Artificielle Tuning MAX MIN Ant System wit off-line and on-line metods Paola Pellegrini, Tomas
More informationDistributed and Optimal Rate Allocation in Application-Layer Multicast
Distributed and Optimal Rate Allocation in Application-Layer Multicast Jinyao Yan, Martin May, Bernard Plattner, Wolfgang Mülbauer Computer Engineering and Networks Laboratory, ETH Zuric, CH-8092, Switzerland
More informationMulti-View Clustering with Constraint Propagation for Learning with an Incomplete Mapping Between Views
Multi-View Clustering wit Constraint Propagation for Learning wit an Incomplete Mapping Between Views Eric Eaton Bryn Mawr College Computer Science Department Bryn Mawr, PA 19010 eeaton@brynmawr.edu Marie
More informationMTH-112 Quiz 1 - Solutions
MTH- Quiz - Solutions Words in italics are for eplanation purposes onl (not necessar to write in te tests or. Determine weter te given relation is a function. Give te domain and range of te relation. {(,
More informationTHE EVALUATION CRITERION FOR COLOR IMAGE SEGMENTATION ALGORITHMS
Journal of ELECTRICAL ENGINEERING, VOL. 63, NO. 1, 2012, 13 20 THE EVALUATION CRITERION FOR COLOR IMAGE SEGMENTATION ALGORITHMS Peter Lukáč Róbert Hudec Miroslav Benčo Zuzana Dubcová Martina Zacariášová
More informationAnalytical CHEMISTRY
ISSN : 974-749 Grap kernels and applications in protein classification Jiang Qiangrong*, Xiong Zikang, Zai Can Department of Computer Science, Beijing University of Tecnology, Beijing, (CHINA) E-mail:
More informationSoft sensor modelling by time difference, recursive partial least squares and adaptive model updating
Soft sensor modelling by time difference, recursive partial least squares adaptive model updating Y Fu 1, 2, W Yang 2, O Xu 1, L Zou 3, J Wang 4 1 Zijiang College, Zejiang University of ecnology, Hangzou
More informationLehrstuhl für Informatik 10 (Systemsimulation)
FRIEDRICH-ALEXANDER-UNIVERSITÄT ERLANGEN-NÜRNBERG INSTITUT FÜR INFORMATIK (MATHEMATISCHE MASCHINEN UND DATENVERARBEITUNG) Lerstul für Informatik 10 (Systemsimulation) PDE based Video Compression in Real
More informationTREES. General Binary Trees The Search Tree ADT Binary Search Trees AVL Trees Threaded trees Splay Trees B-Trees. UNIT -II
UNIT -II TREES General Binary Trees Te Searc Tree DT Binary Searc Trees VL Trees Treaded trees Splay Trees B-Trees. 2MRKS Q& 1. Define Tree tree is a data structure, wic represents ierarcical relationsip
More informationProceedings. Seventh ACM/IEEE International Conference on Distributed Smart Cameras (ICDSC 2013) Palm Spring, CA
Proceedings Of te Sevent ACM/IEEE International Conference on Distributed Smart Cameras (ICDSC ) Palm Spring, CA October 9 November st Parameter-Unaware Autocalibration for Occupancy Mapping David Van
More informationImage Registration via Particle Movement
Image Registration via Particle Movement Zao Yi and Justin Wan Abstract Toug fluid model offers a good approac to nonrigid registration wit large deformations, it suffers from te blurring artifacts introduced
More informationHaar Transform CS 430 Denbigh Starkey
Haar Transform CS Denbig Starkey. Background. Computing te transform. Restoring te original image from te transform 7. Producing te transform matrix 8 5. Using Haar for lossless compression 6. Using Haar
More informationInternational Journal of Advanced Research in Computer Science and Software Engineering
Volume 2, Issue 9, September 2012 ISSN: 2277 128X International Journal of Advanced Researc in Computer Science and Software Engineering Researc Paper Available online at: www.ijarcsse.com Performance
More informationDiscretizing Continuous Attributes Using Information Theory
Discretizing Continuous Attributes Using Information Theory Chang-Hwan Lee Department of Information and Communications, DongGuk University, Seoul, Korea 100-715 chlee@dgu.ac.kr Abstract. Many classification
More information1.4 RATIONAL EXPRESSIONS
6 CHAPTER Fundamentals.4 RATIONAL EXPRESSIONS Te Domain of an Algebraic Epression Simplifying Rational Epressions Multiplying and Dividing Rational Epressions Adding and Subtracting Rational Epressions
More informationMATH 5a Spring 2018 READING ASSIGNMENTS FOR CHAPTER 2
MATH 5a Spring 2018 READING ASSIGNMENTS FOR CHAPTER 2 Note: Tere will be a very sort online reading quiz (WebWork) on eac reading assignment due one our before class on its due date. Due dates can be found
More informationAn Overview of New Features in
6. LS-DYNA Anwenderforum, Frankental 2007 Optimierung An Overview of New Features in LS-OPT Version 3.3 Nielen Stander*, Tusar Goel*, David Björkevik** *Livermore Software Tecnology Corporation, Livermore,
More information4.2 The Derivative. f(x + h) f(x) lim
4.2 Te Derivative Introduction In te previous section, it was sown tat if a function f as a nonvertical tangent line at a point (x, f(x)), ten its slope is given by te it f(x + ) f(x). (*) Tis is potentially
More informationObstacle Avoiding Real-Time Trajectory Generation and Control of Omnidirectional Vehicles
2009 American Control Conference Hyatt Regency Riverfront, St. Louis, MO, USA June 10-12, 2009 FrC10.6 Obstacle Avoiding Real-Time Trajectory Generation and Control of Omnidirectional Veicles Ji-wung Coi,
More informationTraffic Pattern-based Adaptive Routing for Intra-group Communication in Dragonfly Networks
Traffic Pattern-based Adaptive Routing for Intra-group Communication in Dragonfly Networks Peyman Faizian, Md Safayat Raman, Md Atiqul Molla, Xin Yuan Department of Computer Science Florida State University
More informationExcel based finite difference modeling of ground water flow
Journal of Himalaan Eart Sciences 39(006) 49-53 Ecel based finite difference modeling of ground water flow M. Gulraiz Akter 1, Zulfiqar Amad 1 and Kalid Amin Kan 1 Department of Eart Sciences, Quaid-i-Azam
More informationGeorge Xylomenos and George C. Polyzos. with existing protocols and their eæciency in terms of
IP MULTICASTING FOR WIRELESS MOBILE OSTS George Xylomenos and George C. Polyzos fxgeorge,polyzosg@cs.ucsd.edu Computer Systems Laboratory Department of Computer Science and Engineering University of California,
More informationPedestrian Detection Algorithm for On-board Cameras of Multi View Angles
Pedestrian Detection Algoritm for On-board Cameras of Multi View Angles S. Kamijo IEEE, K. Fujimura, and Y. Sibayama Abstract In tis paper, a general algoritm for pedestrian detection by on-board monocular
More informationRedundancy Awareness in SQL Queries
Redundancy Awareness in QL Queries Bin ao and Antonio Badia omputer Engineering and omputer cience Department University of Louisville bin.cao,abadia @louisville.edu Abstract In tis paper, we study QL
More informationREVERSIBLE DATA HIDING USING IMPROVED INTERPOLATION TECHNIQUE
International Researc Journal of Engineering and Tecnology (IRJET) e-issn: 2395-0056 REVERSIBLE DATA HIDING USING IMPROVED INTERPOLATION TECHNIQUE Devendra Kumar 1, Dr. Krisna Raj 2 (Professor in ECE Department
More informationLarge Scale Kernel Machines
Large Scale Kernel Macines Editors: Léon Bottou NEC Labs America Princeton, NJ 08540, USA leon@bottou.org Olivier Capelle capelle@tuebingen.mpg.de Max Planck Institure for Biological Cybernetics 72076
More informationFairness and Load Balancing in Wireless LANs Using Association Control
Fairness and oad Balancing in Wireless ANs Using Association Control Yigal Bejerano, i (Erran) i Bell abs, ucent Tecnologies, NJ USA Seung-Jae Han Yonsei University, Seoul, Korea Abstract: Te traffic load
More information