A novel feature extraction approach for microarray data based on multi-algorithm fusion

Size: px
Start display at page:

Download "A novel feature extraction approach for microarray data based on multi-algorithm fusion"

Transcription

1 Hypothess Voume 11(1) A nove feature extracton approach for mcroarray data based on mut-agorthm fuson Zhu Jang 1,2 & Rong Xu 1 1State Key Laboratory of Astronautc Dynamcs, X an Satete Contro Center, X an, Chna; 2 Key Laboratory of Fud and Power Machnery, Mnstry of Educaton, Xhua Unversty, Chengdu, Chna; Zhu Jang - Ema: h5525@163.com; * Correspondng author Receved December 09, 2014; Revsed December 31, 2014; Accepted January 08, 2015; Pubshed January 30, 2015 Abstract: Feature extracton s one of the most mportant and effectve method to reduce dmenson n data mnng, wth emergng of hgh dmensona data such as mcroarray gene expresson data. Feature extracton for gene seecton, many serves two purposes. One s to dentfy certan dsease-reated genes. The other s to fnd a compact set of dscrmnatve genes to bud a pattern cassfer wth reduced compexty and mproved generazaton capabtes. Dependng on the purpose of gene seecton, two types of feature extracton agorthms ncudng rankng-based feature extracton and set-based feature extracton are empoyed n mcroarray gene expresson data anayss. In rankng-based feature extracton, features are evauated on an ndvdua bass, wthout consderng nter-reatonshp between features n genera, whe set-based feature extracton evauates features based on ther roe n a feature set by takng nto account dependency between features. Just as earnng methods, feature extracton has a probem n ts generazaton abty, whch s robustness. However, the ssue of robustness s often overooked n feature extracton. In order to mprove the accuracy and robustness of feature extracton for mcroarray data, a nove approach based on mut-agorthm fuson s proposed. By fusng dfferent types of feature extracton agorthms to seect the feature from the sampes set, the proposed approach s abe to mprove feature extracton performance. The new approach s tested aganst gene expresson dataset ncudng Coon cancer data, CNS data, DLBCL data, and Leukema data. The testng resuts show that the performance of ths agorthm s better than exstng soutons. Keywords: feature extracton; robustness; mcroarray data; mut-agorthm fuson Background: Feature extracton s one of the key ssues n the fed of data mnng. Researchers have reazed that n order to use data mnng toos effectvey, data preprocessng s essenta to successfu data mnng. Feature extracton s one of the mportant and frequenty used technques n data preprocessng. It not ony can emnate nformaton redundant, mprove the cassfcaton effcency and acceerate the computatona speed, but aso can reduce the compexty of the cassfer and the error rate of cassfcaton [1]. Feature extracton agorthm can be cassfed nto three fundamentay approaches: wrapper, fter and embedded. Wrapper mode evauates the subset of seected features by usng crtera based on the resuts of earnng agorthms, whe fter methods depends on ntrnsc characterstcs of the data to seect feature subsets wthout nvovng and mnng methods [2]. These methods are often mted n use because they requre a ong computatona tme. To take advantage of these two agorthms, the embedded method s proposed [3]. For a feature extracton agorthms, many rea appcatons, such as computer vson, mcroarray technoogy and vsua recognton, nvove mcroarray data. Robustness of feature extracton for these data aso gets ts attentons n recent years. Gugezen et a. [4] studed the stabty and cassfcaton accuracy of Mnmum Redundancy Maxmum Reevance-based feature extracton. An entropy-based measure for stabty assessment was deveoped by Krzek et a [5]. However, research shows that the extracton resuts of most feature extractons are very senstve to the changes of tranng sets. That s to say, these agorthms have poor robustness [6]. Ths probem s partcuary obvous for mcroarray data set. Even f the ISSN (onne) (prnt) Bonformaton 11 (1): (2015) Bomedca Informatcs

2 tranng data set s sghty changed, the obtaned optmum feature subset w have arge dfference and the performance of the cassfcaton mode w change greaty. Therefore, to mprove the credbty of cassfcaton performance, we need to choose the feature extracton agorthm wth a hgh robustness. Fgure 1: Score-based mut-agorthm fuson. In prevous studes, t s we known that combnng or ntegratng mutpe cassfers, especay uncorreated weak ones, coud greaty mprove the cassfcaton performance [7], but studes for the fuson of varous feature extracton agorthms are a few. Marna Skurchna beeves that there may be usefu nformaton n the unseected features after the feature extracton. The omsson of these features may ead to the poor performance of feature extracton and pattern recognton. So t s suggested to use the method of fuson to utze the usefu nformaton n the negected features [8]. As we a know that feature subsets produced by dfferent feature extracton agorthm may show compementary, and fuson of mutpe agorthms utzes the search abtes of each agorthm to get coser to a goba optma souton. Thus a fuson of the feature subsets may produce a better representaton n feature space. A ot of feature extracton agorthms have been proposed n the terature. But not a feature extracton agorthms can be fused. If two extracton agorthms are smar, the fuson of them can not mprove the stabty of the extracton agorthm greaty [9]. Therefore, the dversty must be consdered whe choosng the feature extracton agorthms. Dfferent types of feature extracton can compement each other and avod overappng. Obvousy, t s not necessary to fuse a feature extracton agorthms, whch s aso mpossbe. To smpfy the cacuaton process and reduce the amount of computaton whe mantanng the dversty of extracton agorthms, n ths study, Fsher Rato, Absoute Weght of SVM (AW-SVM) and Poynoma Support Vector Machne (PSVM) were used to fuse n ths paper. Fsher Rato [9] s one of the basc methods n fter mode of feature extracton. It s a unvarate fter method evauatng each feature ndvduay. Its estmaton standards are drecty obtaned from the data set. It has the characterstcs of sma cacuaton cost, hgh effcency, etc.. AW-SVM [10] s an embedded method that ranks features based on ther correspondng coeffcents n the SVM cassfer. PSVM s a wrapper method based on statstca earnng theory [11]. It has powerfu faut toerance and generazaton abtes. Studes have shown that the generazaton abty of PSVM w not reduce wth the ncreasng of the order. It overcomes the probems of over earnng, ack of earnng, oca mnmum vaue and dmenson dsaster of tradtona machne earnng. Fgure 2: Performance comparsons on coon data: A) Cassfcaton Error; B) AUC; C) Standard Devaton of Error Estmaton; D) Feature Robustness. ISSN (onne) (prnt) Bonformaton 11(1) (2015) Bomedca Informatcs

3 By consderng a these factors and based on the dea of fuson, a nove feature extracton method, poynoma support vector machne based on mut-agorthm fuson (MAF-PSVM), was put forward n ths paper. Fgure 3: Performance comparsons on CNS data: A) Cassfcaton Error; B) AUC; C) Standard Devaton of Error Estmaton; D) Feature Robustness. Methodoogy: Poynoma Support Vector Machne based on Mut-Agorthm Fuson The specfc mpementaton steps of MAF-PSVM are ntroduced n deta. Step one, ntaze the sampe data set. Tranng sampes are custered nto K casses by k-means [12] (The vaue of k n ths paper was 8 by tranng.), dfferent penaty factors are used for each cass. Step two, fuse feature extracton agorthms (n ths paper, the agorthms were Fsher Rato, AW-SVM and PSVM (How to seect approprate kerne functons for dfferent appcatons has been a dffcut probem. Studes have shown that near kerne functon s sutabe for the case of neary separabe data, so poynoma kerne was seected. The other kerne functons can be further studed n the future for us. The generazaton abty of poynoma kerne functons s dfferent. In most cases, the performance of the cassfer can reach to optmum whe the parameter d s taken as 1[13]. Therefore, the vaue of order d of the poynoma kerne was taken as 1)) and conduct feature extracton for custered sampes. The foowng contents ntroduce the fuson method used by ths paper: In our study, we used score-based mut-agorthm fuson methods. Frsty, a score vector contanng scores of a features was produced by each bass crteron. Secondy, a score combnaton agorthm was used to aggregate the mutpe score vectors nto one consensus score vector. Fnay, a feature rankng procedure was empoyed to rank the features based on ther consensus scores. The score-based mut-agorthm fuson procedure s ustrated n Fgure 1. In score aggregatng, the scores produced by dfferent bass crtera w be comparabe. It s essenta to ensure that score normazaton shoud be done before score combnaton s ISSN (onne) (prnt) Bonformaton 11(1) (2015) Bomedca Informatcs

4 performed. The scores are normazed to the range of [0, 1] n ths study. Assume u s the score vector produced by bass crteron, the score normazaton s performed as foows: ' u u mn u u max u mn (1) where u mn and u max are the mnmum and maxmum vaues n vector u. For a the bass crtera, t s assumed that the arger the score, the better the feature. A score combnaton method w be used as foowng: m 1 ' u u m 1 (2) where m s the number of bass crtera used n fuson. Step three, feature extracton resuts of the step two w be used to tran the PSVM cassfer. If the numbers of negatve cass ponts and postve cass ponts have a arge dfference n the tranng data set, and f appyng the same penaty parameter C to the set of postve cass ponts and the set of negatve cass ponts, t means that the one wth more cass ponts w get more attenton. However, we hope that penates to the postve pont and negatve pont are not the same. Accordngy, for propery seected parameter C, C C and C C. Where, and are respectvey the number of postve cass tranng ponts and negatve cass tranng ponts. C s the penaty parameter of postve cass pont. Comparatvey, C s the penaty parameter of negatve cass pont. The PSVM cassfer was but as foowng: 1 d max W ( ) ( x x 1) y y 2 1, 1 s.t. y 0, 1 0 Ccass 1, Cass Index cass1; 0 Ccass2, Cass Index cass2; 0 CcassN, Cass Index cassn; 1,, 1, 1, n C 1,, n (3) Where, represents the Lagrange mutper. cass1,, cassn represent the categores after custerng. Cass Index represents the mark of cass. 1,, n represent the number of sampe ponts n each cass, and C represents the penaty factor of each cass. Step four, regress the sampe data set by traned cassfer, and remove features wth the mnma correaton. Sampe set w be updated. In ths study, n order to gve a more genera and precse measure of the smarty between two feature subsets, the correatons between the dfferent features of two feature subsets w be used. JC ( k) S S SC foowng smarty ndex JC ( [0,1]) that takes n account the ISSN (onne) (prnt) Bonformaton 11(1) (2015) Bomedca Informatcs 0 k (4) Where S and S 0 are feature subsets seected usng the th batch of re-samped data and the fu data respectvey. SC s the sum of absoute correaton vaues between the dssmar features form S and S 0. It w be computed usng the greedy search agorthm. Step fve, see f the codng s over, that s, orgna feature set S= [1, 2,, n] s nu. If so, end the teraton, otherwse repeat step two to four unt achevng the feature extracton. Measurement of feature extracton robustness The defnton of robustness of feature extracton s the senstvty degree of the resuts of feature extracton agorthm to the changes of tranng set. Accordng to ths defnton, the measurement of feature extracton robustness s measurng the smarty among the optma feature subsets seected by the agorthm. The overa robustness of the agorthm can be cacuated as [14]: 2 k k S(s, s ) 1 1 Stot (5) k(k 1) Where, s represents the resuts of feature extracton from tranng set No. ( 1 k). S(s, s ) represents a smarty measure between two feature extracton resuts s and s. At present, there are many types of smarty measurng methods accordng to dfferent representaton ways of feature extracton resuts. The commony used set method [14] was seected by ths study. In ths method, the robustness s measured by Tanmoto dstance [12]. Tanmoto dstance s used to measure the concdence degree of eements between two feature subsets: s s 2 s s Ss(s, s ) 1 s σ s s s s s s I (s I (s s s 1) 0 ) If the parameter s true, the functon I () returns as 1; otherwse t returns as 0. The vaue nterva of S s s [0, 1]. 0 means that the ntersecton of these two sets s an empty set, and a eements are dfferent. 1 means that these two sets are exacty the same, and a eements are the same. The szes of sets measured by Tanmoto dstance can be the same or dfferent. Resuts: Coon cancer data, CNS data, DLBCL data, and Leukema data [15] were separatey adopted for the smuaton test. Amng at these mcroarray data sets, performance evauaton of feature extracton was conducted for the method proposed n ths paper, Fsher Rato, AW-SVM and PSVM n four aspects of dentfcaton error, AUC vaues, standard devaton and robustness. The resuts were shown n Fgure 2 - Fgure 5. It can be seen from the smuaton resuts (n the Fgure 2a, Fgure 3a, Fgure 4a, & Fgure 5a) that the accuracy of feature dentfcaton for the method proposed n ths paper s better than the other three methods. Takng Fgure 2a as an exampe, MAF-PSVM reazes the mnmum dentfcaton error by ony

5 extractng 150 features, and the dentfcaton error at ths tme s 12.96%. However, the dentfcaton errors of Fsher Rato, AW-SVM and PSVM when extractng 150 features are 14.70%, 15.63% and 15.17%, separatey. The area (AUC) n the ROC curve s usuay used to measure the cassfcaton performance. The arger the AUC vaues are, the better the cassfcaton performance w be. So AUC vaues are used to evauate the cassfcaton performance of severa feature extracton methods durng the smuaton test. The AUC vaues of these four methods are shown n Fgure 2b, Fgure 3b, Fgure 4b, & Fgure 5b. It can be seen from these resuts and the dentfcaton errors of a methods shown n Fgure 2a, Fgure 3a, Fgure 4a & Fgure 5a that when extractng 150 features, the AUC vaues of the method proposed n ths paper are better than the other three methods when coon data and CNS data were used. When DLBCL data and Leuk data are used to test, the AUC vaues of MAF-PSVM are cose to the resut of FR. It aso ndcates that the cassfcaton performance of MAF-PSVM s better than the other three methods whe achevng the most accurate extracton of features. Fgure 4: Performance comparsons on DLBCL data: A) Cassfcaton Error; B)AUC; C) Standard Devaton of Error Estmaton; D) Feature Robustness. The standard devatons of a methods are shown n Fgure 2c, Fgure 3c, Fgure 4c, and Fgure 5c. By anayzng the smuaton resuts, t s known that the performance of the method proposed n ths paper s better than the other three feature extracton methods when Coon and DLBCL data are used. For exampe, on coon data (Fgure 2c), when extractng 150 features, the standard devaton of MAF-PSVM s ony Its dentfcaton accuracy s second ony to the Fsher Rato, whch standard devaton s at ths tme; the standard devaton of AW-SVM s 0.547, and that of PSVM s the argest, whch reaches to ISSN (onne) (prnt) Bonformaton 11(1) (2015) Bomedca Informatcs

6 The robustness resuts of a methods durng the feature extracton process are shown n Fgure 2d, Fgure 3d, Fgure 4d, Fgure 5d. By anayzng the smuaton resuts, t s known that the stabty of the method proposed n ths paper does not perform the best. Ths s because the method proposed n ths paper s an embedded feature extracton method. Compared wth other agorthms, t fuy consders the dependence among features durng the feature extracton process. The resuts of ths treatment are that the feature can be extracted more accuratey and the dentfcaton of patterns can be reazed. The smuaton resuts of the estmated cassfcaton error, AUC, and the standard devaton of error estmaton can fuy confrm ths. It s worth mentonng that whe evauatng the performance of a feature extracton method, we need to comprehensvey consder the accuracy, effcency and stabty of feature dentfcaton of the method. Cassfcaton performance shoud be the frst consderaton because a cassfcaton-neffectve extracton resut does not make any sense. Based on ths and the above smuaton anayss resuts, we can concude that the comprehensve performance of MAF-PSVM proposed n ths paper s better than the other three methods durng feature extracton of mcroarray data. Fgure 5: Performance comparsons on Leukema data: A) Cassfcaton Error; B) AUC; C) Standard Devaton of Error Estmaton; D) Feature Robustness. Concuson: The feature extracton for mcroarray data was dscussed and anayzed n ths paper. Accordng to the dea of custerng and nformaton fuson, a nove feature extracton method, poynoma support vector machne based on mut-agorthm fuson (MAF-PSVM), was put forward. The smuaton resuts of measured data show that the dentfcaton error, AUC vaues, standard devaton of error estmaton of the method proposed n ths paper are better than Fsher Rato, AW-SVM and PSVM. It was found out whe anayzng and comparng the robustness of feature extracton of a methods that the method proposed n ths paper does not perform the best. Ths s because t fuy consdered the dependence among features. The new method buds a compromse between the accuracy of feature dentfcaton and the stabty of feature extracton. For the performance evauaton of a knd of feature extracton method, both the stabty of feature extracton and the dentfcaton performance (such as dentfcaton accuracy and effcency) sha be consdered. ISSN (onne) (prnt) Bonformaton 11(1) (2015) Bomedca Informatcs

7 Therefore, the comprehensve performance of the MAF-PSVM s better than other methods. Ths method s more sutabe for the feature extracton of mcroarray data. Acknowedgement: Fundng: the research s supported by key aboratory of fud and power machnery, mnstry of educaton.xhua unversty (sz ) and aboratory of pattern recognton and artfca ntegence. References: [1] Langey P, Artfca Integence : 245 [2] Guyon I et a. Journa of Machne Learnng Research : 1157 [3] Madonado SR et a. Informaton Scences : 115 [4] Gugezen GZ et a. Book seres n Machne Learnng and Knowedge Dscovery n Databases : 455 [5] Krzek PJ et a. Lecture Notes n Computer Scence : 929 [6] Davs CA et a. Bonformatcs : 2356 [PMID: ] [7] Yang F & Mao KZ, IEEE/ACM Trans Comput Bo Bonform : 1080 [PMID: ] [8] Skurchna M et a. Mutpe Cassfer Systems : 165 [9] Fung ES & Nq MK, Bonformaton : 230 [PMID: ] [10] Yng-Xn L et a. Chnese Journa of Computers : 324 [11] Jang H & Chng WK, Bonformaton : 257 [PMID: ] [12] Ge Y & Seafon SC, Bonformatcs : 2052 [PMID: ] [13] Dwork C et a. In Word Wde Web. 2001: [14] Kaouss A et a. Knowedge and Informaton System : 95 [15] Edted by P Kangueane Ctaton: Jang & Xu, Bonformaton 11(1): (2015) Lcense statement: Ths s an open-access artce, whch permts unrestrcted use, dstrbuton, and reproducton n any medum, for non-commerca purposes, provded the orgna author and source are credted ISSN (onne) (prnt) Bonformaton 11(1) (2015) Bomedca Informatcs

Machinery Fault Diagnosis Based on Improved Algorithm of Support Vector Domain Description and SVMs

Machinery Fault Diagnosis Based on Improved Algorithm of Support Vector Domain Description and SVMs Machnery Faut Dagnoss Based on Improved Agorthm of Support Vector Doman Descrpton and SVMs Qang Wu Marne Engneerng Coege, Daan Martme Unversty, Daan,Laonng, 606,Chna wuqangd@sohu.com Chuanyng Ja Navgaton

More information

A method of SVM with Normalization in Intrusion Detection

A method of SVM with Normalization in Intrusion Detection Avaabe onne at www.scencedrect.com Proceda Envronmenta Scences (20) 256 262 A method of SVM wth ormazaton n Intruson Detecton Wejun, Zhenyu Lu 2 Coege of Automaton Guangdong Unversty of Technoogy Guangzhou,

More information

Automatic Video Based Face Verification and Recognition by Support Vector Machines

Automatic Video Based Face Verification and Recognition by Support Vector Machines Automatc Vdeo Based Face Verfcaton and Recognton by Support Vector Machnes Gang Song *, Hazhou A, Guangyou Xu, L Zhuang Dept of Computer Scence & Technoogy, Tsnghua Unversty, Bejng, Chna ABSTRACT Ths paper

More information

Video Based Face Verification

Video Based Face Verification Vdeo Based Face Verfcaton HONG We AI Hazhou ZHUANG L XU Guangyou Department of Computer Scence and Technoogy, Tsnghua Unversty State Key Laboratory of Integent Technoogy and Systems ABSTRACT Ths paper

More information

Support Vector Machines for Remote-Sensing Image Classification

Support Vector Machines for Remote-Sensing Image Classification Support Vector Machnes for Remote-Sensng Image Cassfcaton Fabo Ro *, Gorgo Fumera Dept. of Eectrca and Eectronc Eng., Unversty of Cagar, Pazza d Arm, 0923, Cagar, Itay ABSTRACT In the ast decade, the appcaton

More information

A Modified Editing k-nearest Neighbor Rule

A Modified Editing k-nearest Neighbor Rule JOURNAL OF COMPUTERS, VOL. 6, NO. 7, JULY 0 493 A Modfed Edtng k-nearest Neghbor Rue Ruqn Chang, Zheng Pe and Chao Zhang Schoo of Mathematcs and Computer Engneerng hua Unversty Chengdu, 60039, Chna Ema:

More information

GA-SVR Prediction of Failure Depth of Coal Seam Floor Based on Small Sample Data

GA-SVR Prediction of Failure Depth of Coal Seam Floor Based on Small Sample Data 23 2nd Internatona Conference on Geoogca and Envronmenta Scences IPCBEE vo.52 (23) (23) IACSIT Press, Sngapore DOI:.7763/IPCBEE. 23. V52. 7 GA-SVR Predcton of Faure Depth of Coa Seam Foor Based on Sma

More information

Study of Deformation-Compensated Modeling for Flexible Material Path Processing Based on Fuzzy Neural Network and Fuzzy Clustering

Study of Deformation-Compensated Modeling for Flexible Material Path Processing Based on Fuzzy Neural Network and Fuzzy Clustering Study of Deformaton-Compensated Modeng for Fexbe Matera Path Processng Based on Fuzzy eura etwor and Fuzzy Custerng Yaohua Deng, Scheng Chen, Jayuan Chen, Hu Chen and Lmng Wu Yaohua Deng, Scheng Chen,

More information

Customer Behavior Clustering Using SVM*

Customer Behavior Clustering Using SVM* Avaabe onne at www.scencedrect.com Physcs Proceda 33 ( ) 489 496 Internatona Conference on Medca Physcs and Bomedca Engneerng Customer Behavor Custerng Usng SVM* Zhongyng Yang,Xaoong Su Dept of Compute

More information

Greedy algorithms of feature selection for multiclass image classification

Greedy algorithms of feature selection for multiclass image classification Greedy agorthms of feature seecton for mutcass mage cassfcaton E F Goncharova and A V Gade,2 Samara Natona Research Unversty, Moskovskoe Shosse 34, Samara, Russa, 443086 2 Image Processng Systems Insttute

More information

An improved hierarchical partitioning fuzzy approach to pattern classification

An improved hierarchical partitioning fuzzy approach to pattern classification An mproved herarchca parttonng fuzzy approach to pattern cassfcaton A dssertaton submtted to The Unversty of Manchester for the degree of MSc Informaton Systems Engneerng n the Facuty of Engneerng and

More information

Optimal Allocation of Multi-Platform Sensor Resources for Multiple Target Tracking

Optimal Allocation of Multi-Platform Sensor Resources for Multiple Target Tracking 4th Internatona Conference on Informaton Fuson Chcago, Inos, USA, Juy 5-8, 0 Optma Aocaton of Mut-Patform Sensor Resources for Mutpe Target Tracng Gary Asns Space and Arborne Systems Raytheon Company E

More information

A FAST MATCHING APPROACH OF POLYGON FEATURES

A FAST MATCHING APPROACH OF POLYGON FEATURES ISPRS Annas of the Photogrammetry, Remote Sensng and Spata Informaton Scences, Voume I-, XXII ISPRS Congress, 5 August September, Mebourne, Austraa A FAST MATCHING APPROACH OF POLYGON FEATURES Qngxang

More information

Matching Pursuit for Sparse Signal Reconstruction Based on Dual Thresholds

Matching Pursuit for Sparse Signal Reconstruction Based on Dual Thresholds Matchng Pursut for Sparse Sgna Reconstructon Based on Dua Threshods Zhengguang Xe*, Huang Hongwe, Ca Xu Schoo of Eectroncs and Informaton, Nantong Unversty, Nantong Cty, Jangsu Provnce, Chna 69. * Correspondng

More information

A FUZZY BASED DISTRIBUTED NODE MOVEMENT ALGORITHM FOR MAINTAINING NEIGHBORHOOD TOPOLOGY IN MOBILE AD-HOC NETWORKS

A FUZZY BASED DISTRIBUTED NODE MOVEMENT ALGORITHM FOR MAINTAINING NEIGHBORHOOD TOPOLOGY IN MOBILE AD-HOC NETWORKS Internatona Journa of Informaton Technoogy and Knowedge Management Juy-December 2009, Voume 2, No. 2, pp. 387-391 A FUZZY BASED DISTRIBUTED NODE MOVEMENT ALGORITHM FOR MAINTAINING NEIGHBORHOOD TOPOLOGY

More information

An Improved Ordered Subsets Expectation Maximization Reconstruction

An Improved Ordered Subsets Expectation Maximization Reconstruction An Improved Ordered Subsets Expectaton axmzaton Reconstructon Le Xu ), Huafu Chen,2),Dezhong Yao ) ) Schoo of Lfe Scence and Technoogy, Unversty of Eectronc Scence and Technoogy of Chna, 60054, Chna 2)

More information

Forecasting Satellite Attitude Volatility Using Support Vector Regression with Particle Swarm Optimization

Forecasting Satellite Attitude Volatility Using Support Vector Regression with Particle Swarm Optimization IAENG Internatona Journa of Computer Scence, 4:3, IJCS_4_3_ Forecastng Satete Atttude Voatty Usng Support Vector Regresson wth Partce Swarm Optmzaton Zuhua Zhong, Dechang P Abstract Forecastng the voatty

More information

Photo aesthetics evaluation system: an application of CNN and SVM

Photo aesthetics evaluation system: an application of CNN and SVM Photo aesthetcs evauaton system: an appcaton of CNN and SVM Chen Qan chenq@stanford.edu Zh L zh@stanford.edu Abstract In ths project, we apped two machne earnng technques: CNN (Convoutona Neura Network)

More information

ENHANCING PERFORMANCE OF FACE RECOGNITION SYSTEM BY USING NEAR SET APPROACH FOR SELECTING FACIAL FEATURES

ENHANCING PERFORMANCE OF FACE RECOGNITION SYSTEM BY USING NEAR SET APPROACH FOR SELECTING FACIAL FEATURES Journa of Theoretca and Apped Informaton Technoogy 005-008 JATIT. A rghts reserved. www.att.org ENHANCING PERFORMANCE OF FACE RECOGNITION SYSTEM BY USING NEAR SET APPROACH FOR SELECTING FACIAL FEATURES

More information

Using Deep Learning and Gaussian Mixture Models for Road Scene Segmentation Yu-Chen Lin a, *, Peng-Cheng Chen b, Cheng-Hsien Wang c,

Using Deep Learning and Gaussian Mixture Models for Road Scene Segmentation Yu-Chen Lin a, *, Peng-Cheng Chen b, Cheng-Hsien Wang c, ISO 9001:2008 Certfed Internatona Journa of Engneerng Scence and Innovatve Technoogy (IJESIT) Usng Deep Learnng and Gaussan Mxture Modes for Road Scene Segmentaton Yu-Chen Ln a, *, Peng-Cheng Chen b, Cheng-Hsen

More information

UTILITY-BASED DYNAMIC CAMERA ASSIGNMENT AND HAND-OFF IN A VIDEO NETWORK

UTILITY-BASED DYNAMIC CAMERA ASSIGNMENT AND HAND-OFF IN A VIDEO NETWORK UTILITY-BASED DYNAMIC CAMERA ASSIGNMENT AND HAND-OFF IN A VIDEO NETWORK Ymng L and Br Bhanu Center for Research n Integent Systems Unversty of Caforna at Rversde Rversde, CA 9252, USA E-ma: ym@vsab.ucr.edu;

More information

Improving the Generalization of Fisherface by Training Class Selection Using SOM 2

Improving the Generalization of Fisherface by Training Class Selection Using SOM 2 mprovng the Generazaton of Fsherface by Tranng Cass Seecton Usng SOM 2 Jayan Jang 1, Lmng Zhang 1, and Tetsuo Furuawa 2 1 E.E. Dept. Fudan Unversty, 220 Handan Road, Shangha, Chna jangjayan@ctz.net, mzhang@fudan.edu.cn

More information

Using a Combined Intuitionistic Fuzzy Set-TOPSIS Method for Evaluating Project and Portfolio Management Information Systems

Using a Combined Intuitionistic Fuzzy Set-TOPSIS Method for Evaluating Project and Portfolio Management Information Systems Usng a Combned Intutonstc Fuzzy Set-TOPSIS Method for Evauatng Proect and Portfoo Management Informaton Systems Vasss C. Geroganns 1,, Panos Ftss 1,, and Aches D. Kameas 1 Proect Management Department,

More information

Hadoop-Based Similarity Computation System for Composed Documents

Hadoop-Based Similarity Computation System for Composed Documents Journa of Computer an Communcatons, 2015, 3, 196-202 Pubshe Onne May 2015 n ScRes. http://www.scrp.org/ourna/cc http://x.o.org/10.4236/cc.2015.35025 Haoop-Base Smarty Computaton System for Compose Documents

More information

Training of Fuzzy Neural Networks via Quantum- Behaved Particle Swarm Optimization and Rival Penalized Competitive Learning

Training of Fuzzy Neural Networks via Quantum- Behaved Particle Swarm Optimization and Rival Penalized Competitive Learning 36 The Internatona Arab Journa of Informaton Technoogy, Vo. 9, No. 4, Juy Tranng of Fuzzy Neura Networks va Quantum- Behaved Partce Swarm Optmzaton and Rva Penazed Compettve earnng Saeed Farz Department

More information

Induction of fuzzy rules based on scattered data modeling and multidimensional interpolation: A novel approach

Induction of fuzzy rules based on scattered data modeling and multidimensional interpolation: A novel approach Inducton of fuzzy rues based on scattered data modeng and mutdmensona nterpoaton: A nove approach D. S. Sfrs Department of Cv Engneerng, Democrtus Unversty of Thrace, 67100 Xanth, Greece Abstract ths paper

More information

KANEVSKI M. (1), ARUTYNYAN R. (1), BOLSHOV L. (1), CHERNOV S. (1), DEMYANOV V. (1), SAVELIEVA E. (1), TIMONIN V. (1), MAIGNAN M. (2), BRUNO R.

KANEVSKI M. (1), ARUTYNYAN R. (1), BOLSHOV L. (1), CHERNOV S. (1), DEMYANOV V. (1), SAVELIEVA E. (1), TIMONIN V. (1), MAIGNAN M. (2), BRUNO R. Scentfc regstraton number : 255 Symposum n : 25 Presentaton : poster Geostatstca Assessment of Post-Chernoby So Pouton n Bransk Hghy Pouted Area, Russa Evauaton géostatstque de a pouton des sos après accdent

More information

Classifier Selection Based on Data Complexity Measures *

Classifier Selection Based on Data Complexity Measures * Classfer Selecton Based on Data Complexty Measures * Edth Hernández-Reyes, J.A. Carrasco-Ochoa, and J.Fco. Martínez-Trndad Natonal Insttute for Astrophyscs, Optcs and Electroncs, Lus Enrque Erro No.1 Sta.

More information

RPNCH: A Method for Constructing Rooted Phylogenetic Networks from Rooted Triplets based on Height Function

RPNCH: A Method for Constructing Rooted Phylogenetic Networks from Rooted Triplets based on Height Function Orgna Artce: RPNCH: A Method for Constructng Rooted Phyogenetc Networs from Rooted Trpets based on Heght Functon Mohammad Hossen Reyhan 1,*, Had Poormohammad 2,3 1 Department of Mathematcs, Isamc Azad

More information

Determining the Optimal Bandwidth Based on Multi-criterion Fusion

Determining the Optimal Bandwidth Based on Multi-criterion Fusion Proceedngs of 01 4th Internatonal Conference on Machne Learnng and Computng IPCSIT vol. 5 (01) (01) IACSIT Press, Sngapore Determnng the Optmal Bandwdth Based on Mult-crteron Fuson Ha-L Lang 1+, Xan-Mn

More information

A NEW STATISTIC PREDICTION MODEL AND APPLICATION IN ARITHMETIC CODING

A NEW STATISTIC PREDICTION MODEL AND APPLICATION IN ARITHMETIC CODING A NEW STATISTIC PREDICTION MODEL AND APPLICATION IN ARITHMETIC CODING Hua Gang Zheng Nanng the Insttute of Artfca Integence and Robotcs, an Jaotong Unv. P.O.Bo 999# 70049 ganghua@263.net Abstract: In ths

More information

Automatic Construction of Web Directory using Hyperlink and Anchor Text

Automatic Construction of Web Directory using Hyperlink and Anchor Text Automatc Constructon of Web Drectory usng Hypern and Anchor Text Yusue Suzu, Shge Matsubara and Masatosh Yoshawa Graduate Schoo of Informaton Scence, Nagoya Unversty, Furo-cho, Chusa-u, Nagoya Informaton

More information

SOMEA: self-organizing map based extraction algorithm for DNA motif identification with heterogeneous model

SOMEA: self-organizing map based extraction algorithm for DNA motif identification with heterogeneous model http://www.bomedcentra.com/47-205/2/s/s6 RESEARCH Open Access SOMEA: sef-organzng map based extracton agorthm for DNA motf dentfcaton wth heterogeneous mode Nung Kon Lee, Danhu Wang * From The Nnth Asa

More information

Video Watermarking Based on 3D Scale Invariant Spatio-Temporal Feature Points

Video Watermarking Based on 3D Scale Invariant Spatio-Temporal Feature Points Sensors & Transducers 04 by IFSA Pubshng, S. L. http://www.sensorsporta.com Vdeo Watermarkng Based on Scae Invarant Spato-Tempora Feature Ponts, Hongbo BI, Yng LIU, Yubo ZHANG, Xufang WANG, Dongme WANG

More information

Cluster Analysis of Electrical Behavior

Cluster Analysis of Electrical Behavior Journal of Computer and Communcatons, 205, 3, 88-93 Publshed Onlne May 205 n ScRes. http://www.scrp.org/ournal/cc http://dx.do.org/0.4236/cc.205.350 Cluster Analyss of Electrcal Behavor Ln Lu Ln Lu, School

More information

Resource based build direction in additive manufacturing processes

Resource based build direction in additive manufacturing processes Abstract Resource based bud drecton n addtve manufacturng processes Nazmu Ahsan, Ahasan Habb and Bashr Khoda * Industra & Manufacturng Engneerng Dept. North Dakota State Unversty, Fargo, ND 58104, USA

More information

Subspace clustering. Clustering. Fundamental to all clustering techniques is the choice of distance measure between data points;

Subspace clustering. Clustering. Fundamental to all clustering techniques is the choice of distance measure between data points; Subspace clusterng Clusterng Fundamental to all clusterng technques s the choce of dstance measure between data ponts; D q ( ) ( ) 2 x x = x x, j k = 1 k jk Squared Eucldean dstance Assumpton: All features

More information

index.pdf March 20,

index.pdf March 20, ndex.pdf March 20, 20 ITI 2. Introducton to omputng II Impementaton -2-: usng an nner cass We use an nner cass to create the terators: Marce Turcotte Schoo of Informaton Technoogy and Engneerng Iterator

More information

Fuzzy Sliding Mode Controller with Neural Network Applications to Mobile Robot

Fuzzy Sliding Mode Controller with Neural Network Applications to Mobile Robot Fuzzy Sdng Mode Controer wth Neura Network Appcatons to Mobe Robot Lon-Chen Hung and Hung-Yuan Chung* Department of Eectrca Engneerng, Natona Centra Unversty, Chung-L, Tao-Yuan, 3, Tawan, R.O.C TEL: 886-3-475

More information

Effects of Color Characterization on Computational Efficiency of Feature Detection with Live-object Handling Applications

Effects of Color Characterization on Computational Efficiency of Feature Detection with Live-object Handling Applications Proceedngs of the 005 IEEE/ASME Internatona Conference on Advanced Integent Mechatroncs Monterey, Caforna, USA, 4-8 Juy, 005 MB4-04 Effects of Coor Characterzaton on Computatona Effcency of Feature Detecton

More information

GSLM Operations Research II Fall 13/14

GSLM Operations Research II Fall 13/14 GSLM 58 Operatons Research II Fall /4 6. Separable Programmng Consder a general NLP mn f(x) s.t. g j (x) b j j =. m. Defnton 6.. The NLP s a separable program f ts objectve functon and all constrants are

More information

Economic Computation and Economic Cybernetics Studies and Research, Issue 2/2018; Vol. 52

Economic Computation and Economic Cybernetics Studies and Research, Issue 2/2018; Vol. 52 Economc Computaton and Economc Cybernetcs Studes and Research Issue 2/208; Vo. 52 Assstant Professor Sohe RANA PhD East West Unversty Dhaka-22 Bangadesh E-ma: srana_stat@yahoo.com Lecturer Waeed DHHAN

More information

Modular PCA Face Recognition Based on Weighted Average

Modular PCA Face Recognition Based on Weighted Average odern Appled Scence odular PCA Face Recognton Based on Weghted Average Chengmao Han (Correspondng author) Department of athematcs, Lny Normal Unversty Lny 76005, Chna E-mal: hanchengmao@163.com Abstract

More information

Content Based Image Retrieval Using 2-D Discrete Wavelet with Texture Feature with Different Classifiers

Content Based Image Retrieval Using 2-D Discrete Wavelet with Texture Feature with Different Classifiers IOSR Journal of Electroncs and Communcaton Engneerng (IOSR-JECE) e-issn: 78-834,p- ISSN: 78-8735.Volume 9, Issue, Ver. IV (Mar - Apr. 04), PP 0-07 Content Based Image Retreval Usng -D Dscrete Wavelet wth

More information

Framework for Generalization and Improvement of Relational Data Warehouses

Framework for Generalization and Improvement of Relational Data Warehouses 32 IJCSNS Internatona Journa of Computer Scence and Network Securty, OL.8 No.3, arch 2008 Framework for Generazaton and Improvement of Reatona Data Warehouses Goran enov, Dano Ggorosk and argta Kon Popovska,

More information

An Example-based Prior Model for Text Image Super-resolution

An Example-based Prior Model for Text Image Super-resolution An ampe-based Pror Mode for Tet Image Super-resouton Jangkun Park Younghee Kwon and Jn Hung Km Dvson of Computer Scence KAIST Guseong-dong Yuseong-gu Daeeon Korea -ma: {kparkkheekm@a.kast.ac.kr Abstract

More information

Journal of Chemical and Pharmaceutical Research, 2014, 6(6): Research Article. A selective ensemble classification method on microarray data

Journal of Chemical and Pharmaceutical Research, 2014, 6(6): Research Article. A selective ensemble classification method on microarray data Avalable onlne www.ocpr.com Journal of Chemcal and Pharmaceutcal Research, 2014, 6(6):2860-2866 Research Artcle ISSN : 0975-7384 CODEN(USA) : JCPRC5 A selectve ensemble classfcaton method on mcroarray

More information

Learning the Kernel Parameters in Kernel Minimum Distance Classifier

Learning the Kernel Parameters in Kernel Minimum Distance Classifier Learnng the Kernel Parameters n Kernel Mnmum Dstance Classfer Daoqang Zhang 1,, Songcan Chen and Zh-Hua Zhou 1* 1 Natonal Laboratory for Novel Software Technology Nanjng Unversty, Nanjng 193, Chna Department

More information

Term Weighting Classification System Using the Chi-square Statistic for the Classification Subtask at NTCIR-6 Patent Retrieval Task

Term Weighting Classification System Using the Chi-square Statistic for the Classification Subtask at NTCIR-6 Patent Retrieval Task Proceedngs of NTCIR-6 Workshop Meetng, May 15-18, 2007, Tokyo, Japan Term Weghtng Classfcaton System Usng the Ch-square Statstc for the Classfcaton Subtask at NTCIR-6 Patent Retreval Task Kotaro Hashmoto

More information

BOOSTING CLASSIFICATION ACCURACY WITH SAMPLES CHOSEN FROM A VALIDATION SET

BOOSTING CLASSIFICATION ACCURACY WITH SAMPLES CHOSEN FROM A VALIDATION SET 1 BOOSTING CLASSIFICATION ACCURACY WITH SAMPLES CHOSEN FROM A VALIDATION SET TZU-CHENG CHUANG School of Electrcal and Computer Engneerng, Purdue Unversty, West Lafayette, Indana 47907 SAUL B. GELFAND School

More information

Research Article Improving Streaming Capacity in Multi-Channel P2P VoD Systems via Intra-Channel and Cross-Channel Resource Allocation

Research Article Improving Streaming Capacity in Multi-Channel P2P VoD Systems via Intra-Channel and Cross-Channel Resource Allocation Internatona Journa of Dgta Mutmeda Broadcastng Voume 202, Artce ID 807520, 9 pages do:0.55/202/807520 Research Artce Improvng Streamng Capacty n Mut-Channe P2P VoD Systems va Intra-Channe and Cross-Channe

More information

Design of Fuzzy Multiobjective Linear Program Integrated with Fuzzy Vikor for Facility Location

Design of Fuzzy Multiobjective Linear Program Integrated with Fuzzy Vikor for Facility Location ndan Journa of Scence and Technoogy, Vo 7(), 25 34, January 204 Desgn of Fuzzy Mutobjectve near Program ntegrated wth Fuzzy Vkor for Facty ocaton C. Kavtha * and C. Vjayaakshm 2 Department of Mathematcs,

More information

Expectation Maximization (EM). Mixtures of Gaussians. Learning probability distribution

Expectation Maximization (EM). Mixtures of Gaussians. Learning probability distribution S 2750 Macne Learnng Lecture 7 ectaton Mamzaton M. Mtures of Gaussans. Mos auskrect mos@cs.tt.edu 5329 Sennott Square S 2750 Macne Learnng Learnng robabty dstrbuton Basc earnng settngs: A set of random

More information

Outline. Discriminative classifiers for image recognition. Where in the World? A nearest neighbor recognition example 4/14/2011. CS 376 Lecture 22 1

Outline. Discriminative classifiers for image recognition. Where in the World? A nearest neighbor recognition example 4/14/2011. CS 376 Lecture 22 1 4/14/011 Outlne Dscrmnatve classfers for mage recognton Wednesday, Aprl 13 Krsten Grauman UT-Austn Last tme: wndow-based generc obect detecton basc ppelne face detecton wth boostng as case study Today:

More information

An Improved Image Segmentation Algorithm Based on the Otsu Method

An Improved Image Segmentation Algorithm Based on the Otsu Method 3th ACIS Internatonal Conference on Software Engneerng, Artfcal Intellgence, Networkng arallel/dstrbuted Computng An Improved Image Segmentaton Algorthm Based on the Otsu Method Mengxng Huang, enjao Yu,

More information

The Greedy Method. Outline and Reading. Change Money Problem. Greedy Algorithms. Applications of the Greedy Strategy. The Greedy Method Technique

The Greedy Method. Outline and Reading. Change Money Problem. Greedy Algorithms. Applications of the Greedy Strategy. The Greedy Method Technique //00 :0 AM Outlne and Readng The Greedy Method The Greedy Method Technque (secton.) Fractonal Knapsack Problem (secton..) Task Schedulng (secton..) Mnmum Spannng Trees (secton.) Change Money Problem Greedy

More information

Machine Learning: Algorithms and Applications

Machine Learning: Algorithms and Applications 14/05/1 Machne Learnng: Algorthms and Applcatons Florano Zn Free Unversty of Bozen-Bolzano Faculty of Computer Scence Academc Year 011-01 Lecture 10: 14 May 01 Unsupervsed Learnng cont Sldes courtesy of

More information

Optimal Online Scheduling with Arbitrary Hard Deadlines in Multihop Communication Networks

Optimal Online Scheduling with Arbitrary Hard Deadlines in Multihop Communication Networks Optma Onne Schedung wth Arbtrary Hard Deadnes n Muthop Communcaton Networks Zhouja Mao, C. Emre Koksa, Ness B. Shroff E-ma: maoz@ece.osu.edu, koksa@ece.osu.edu, shroff@ece.osu.edu Abstract The probem of

More information

IT is common to perform high-dimensional data clustering

IT is common to perform high-dimensional data clustering 1532 IEEE TRANSACTIONS ON PATTERN ANALYSIS AND MACHINE INTELLIGENCE, VOL. 33, NO. 8, AUGUST 2011 Feature Seecton and Kerne Learnng for Loca Learnng-Based Custerng Hong Zeng, Member, IEEE, and Yu-mng Cheung,

More information

Problem Definitions and Evaluation Criteria for Computational Expensive Optimization

Problem Definitions and Evaluation Criteria for Computational Expensive Optimization Problem efntons and Evaluaton Crtera for Computatonal Expensve Optmzaton B. Lu 1, Q. Chen and Q. Zhang 3, J. J. Lang 4, P. N. Suganthan, B. Y. Qu 6 1 epartment of Computng, Glyndwr Unversty, UK Faclty

More information

Data Mining: Model Evaluation

Data Mining: Model Evaluation Data Mnng: Model Evaluaton Aprl 16, 2013 1 Issues: Evaluatng Classfcaton Methods Accurac classfer accurac: predctng class label predctor accurac: guessng value of predcted attrbutes Speed tme to construct

More information

Error Awareness Data Mining

Error Awareness Data Mining Error Awareness Data nng Xngquan Zhu, ember, IEEE, and Xndong Wu, Senor ember, IEEE Abstract Rea-word data mnng appcatons often dea wth ow-quaty nformaton sources where data coecton naccuracy, devce mtatons,

More information

Two novel procedures for aggregating randomized model ensemble outcomes for robust signal reconstruction in nuclear power plants monitoring systems

Two novel procedures for aggregating randomized model ensemble outcomes for robust signal reconstruction in nuclear power plants monitoring systems Two nove procedures for aggregatng randomzed mode ensembe outcomes for robust sgna reconstructon n nucear power pants montorng systems Pero Barad, Enrco Zo, Guo Goa, Davde Roverso, M. Hoffmann To cte ths

More information

A New Approach For the Ranking of Fuzzy Sets With Different Heights

A New Approach For the Ranking of Fuzzy Sets With Different Heights New pproach For the ankng of Fuzzy Sets Wth Dfferent Heghts Pushpnder Sngh School of Mathematcs Computer pplcatons Thapar Unversty, Patala-7 00 Inda pushpndersnl@gmalcom STCT ankng of fuzzy sets plays

More information

X- Chart Using ANOM Approach

X- Chart Using ANOM Approach ISSN 1684-8403 Journal of Statstcs Volume 17, 010, pp. 3-3 Abstract X- Chart Usng ANOM Approach Gullapall Chakravarth 1 and Chaluvad Venkateswara Rao Control lmts for ndvdual measurements (X) chart are

More information

Web Database Sampling Based on Dependency of Keywords

Web Database Sampling Based on Dependency of Keywords Send Orders for Rernts to rernts@benthamscence.ae The Oen Cybernetcs & Systemcs Journa, 2015, 9, 375-383 375 Web Database Samng Based on Deendency of Keywords Zhang Ru 1*, Wang Feng 1 and Ln Peguang 2

More information

Collaboratively Regularized Nearest Points for Set Based Recognition

Collaboratively Regularized Nearest Points for Set Based Recognition Academc Center for Computng and Meda Studes, Kyoto Unversty Collaboratvely Regularzed Nearest Ponts for Set Based Recognton Yang Wu, Mchhko Mnoh, Masayuk Mukunok Kyoto Unversty 9/1/013 BMVC 013 @ Brstol,

More information

Graph Tree Decomposition Based Fast Peptide Sequencing and

Graph Tree Decomposition Based Fast Peptide Sequencing and Graph Tree Decomposton Based Fast Peptde Sequencng and Spectra Agnment Internatona Journa of Computatona Scence 1992-6669 (Prnt) 1992-6677 (Onne) www.gp.h/cs 2008 Goba Informaton Pubsher (H.K) Co., Ltd.

More information

Constructing a Maximal Independent Set in Parallel. Mark Goldberg ( ), Thomas Spencer Department of Computer Science Rensselaer Polytechnic Institute

Constructing a Maximal Independent Set in Parallel. Mark Goldberg ( ), Thomas Spencer Department of Computer Science Rensselaer Polytechnic Institute Constructng a Maxma Independent Set n Parae Mark Godberg ( ), Thomas Spencer Department of Computer Scence Rensseaer Poytechnc Insttute ABSTRACT The probem of constructng n parae a maxma ndependent set

More information

User Authentication Based On Behavioral Mouse Dynamics Biometrics

User Authentication Based On Behavioral Mouse Dynamics Biometrics User Authentcaton Based On Behavoral Mouse Dynamcs Bometrcs Chee-Hyung Yoon Danel Donghyun Km Department of Computer Scence Department of Computer Scence Stanford Unversty Stanford Unversty Stanford, CA

More information

THE SIGNIFICANCE OF GEOMETRIC NONLINEARITY ON INTEGRATED THICKNESS AND SHAPE OPTIMAL SHELL SOLUTIONS

THE SIGNIFICANCE OF GEOMETRIC NONLINEARITY ON INTEGRATED THICKNESS AND SHAPE OPTIMAL SHELL SOLUTIONS Proceedngs of the 6th ASMO UK Conference on Engneerng Desgn Optmzaton, Page 275 HE SIGNIFICANCE OF GEOMERIC NONLINEARIY ON INEGRAED HICKNESS AND SHAPE OPIMAL SHELL SOLUIONS Artht Petchsasthon and Peter

More information

Simulation: Solving Dynamic Models ABE 5646 Week 11 Chapter 2, Spring 2010

Simulation: Solving Dynamic Models ABE 5646 Week 11 Chapter 2, Spring 2010 Smulaton: Solvng Dynamc Models ABE 5646 Week Chapter 2, Sprng 200 Week Descrpton Readng Materal Mar 5- Mar 9 Evaluatng [Crop] Models Comparng a model wth data - Graphcal, errors - Measures of agreement

More information

Texture Analysis of Remote Sensing Imagery with Clustering and Bayesian Inference

Texture Analysis of Remote Sensing Imagery with Clustering and Bayesian Inference I.J. Image, Graphcs and Sgna Processng, 015, 9, 1-10 Pubshed Onne August 015 n ECS (http://www.mecs-press.org/) DOI: 10.5815/jgsp.015.09.01 exture Anayss of Remote Sensng Imagery wth Custerng and Bayesan

More information

Support Vector Machines

Support Vector Machines /9/207 MIST.6060 Busness Intellgence and Data Mnng What are Support Vector Machnes? Support Vector Machnes Support Vector Machnes (SVMs) are supervsed learnng technques that analyze data and recognze patterns.

More information

FEATURE EXTRACTION. Dr. K.Vijayarekha. Associate Dean School of Electrical and Electronics Engineering SASTRA University, Thanjavur

FEATURE EXTRACTION. Dr. K.Vijayarekha. Associate Dean School of Electrical and Electronics Engineering SASTRA University, Thanjavur FEATURE EXTRACTION Dr. K.Vjayarekha Assocate Dean School of Electrcal and Electroncs Engneerng SASTRA Unversty, Thanjavur613 41 Jont Intatve of IITs and IISc Funded by MHRD Page 1 of 8 Table of Contents

More information

A Unified Framework for Semantics and Feature Based Relevance Feedback in Image Retrieval Systems

A Unified Framework for Semantics and Feature Based Relevance Feedback in Image Retrieval Systems A Unfed Framework for Semantcs and Feature Based Relevance Feedback n Image Retreval Systems Ye Lu *, Chunhu Hu 2, Xngquan Zhu 3*, HongJang Zhang 2, Qang Yang * School of Computng Scence Smon Fraser Unversty

More information

Relevance Feedback for Image Retrieval

Relevance Feedback for Image Retrieval Vashal D Dhale et al, / (IJCSIT Internatonal Journal of Computer Scence and Informaton Technologes, Vol 4 (2, 203, 39-323 Relevance Feedback for Image Retreval Vashal D Dhale, Dr A R Mahaan, Prof Uma Thakur

More information

Support Vector Machines

Support Vector Machines Support Vector Machnes Decson surface s a hyperplane (lne n 2D) n feature space (smlar to the Perceptron) Arguably, the most mportant recent dscovery n machne learnng In a nutshell: map the data to a predetermned

More information

An Image Fusion Approach Based on Segmentation Region

An Image Fusion Approach Based on Segmentation Region Rong Wang, L-Qun Gao, Shu Yang, Yu-Hua Cha, and Yan-Chun Lu An Image Fuson Approach Based On Segmentaton Regon An Image Fuson Approach Based on Segmentaton Regon Rong Wang, L-Qun Gao, Shu Yang 3, Yu-Hua

More information

MULTISPECTRAL IMAGES CLASSIFICATION BASED ON KLT AND ATR AUTOMATIC TARGET RECOGNITION

MULTISPECTRAL IMAGES CLASSIFICATION BASED ON KLT AND ATR AUTOMATIC TARGET RECOGNITION MULTISPECTRAL IMAGES CLASSIFICATION BASED ON KLT AND ATR AUTOMATIC TARGET RECOGNITION Paulo Quntlano 1 & Antono Santa-Rosa 1 Federal Polce Department, Brasla, Brazl. E-mals: quntlano.pqs@dpf.gov.br and

More information

BioTechnology. An Indian Journal FULL PAPER. Trade Science Inc.

BioTechnology. An Indian Journal FULL PAPER. Trade Science Inc. [Type text] [Type text] [Type text] ISSN : 0974-74 Volume 0 Issue BoTechnology 04 An Indan Journal FULL PAPER BTAIJ 0() 04 [684-689] Revew on Chna s sports ndustry fnancng market based on market -orented

More information

Edge Detection in Noisy Images Using the Support Vector Machines

Edge Detection in Noisy Images Using the Support Vector Machines Edge Detecton n Nosy Images Usng the Support Vector Machnes Hlaro Gómez-Moreno, Saturnno Maldonado-Bascón, Francsco López-Ferreras Sgnal Theory and Communcatons Department. Unversty of Alcalá Crta. Madrd-Barcelona

More information

An Indian Journal FULL PAPER ABSTRACT KEYWORDS. Trade Science Inc.

An Indian Journal FULL PAPER ABSTRACT KEYWORDS. Trade Science Inc. [Type text] [Type text] [Type text] ISSN : 97-735 Volume Issue 9 BoTechnology An Indan Journal FULL PAPER BTAIJ, (9), [333-3] Matlab mult-dmensonal model-based - 3 Chnese football assocaton super league

More information

Parallelism for Nested Loops with Non-uniform and Flow Dependences

Parallelism for Nested Loops with Non-uniform and Flow Dependences Parallelsm for Nested Loops wth Non-unform and Flow Dependences Sam-Jn Jeong Dept. of Informaton & Communcaton Engneerng, Cheonan Unversty, 5, Anseo-dong, Cheonan, Chungnam, 330-80, Korea. seong@cheonan.ac.kr

More information

An Optimal Algorithm for Prufer Codes *

An Optimal Algorithm for Prufer Codes * J. Software Engneerng & Applcatons, 2009, 2: 111-115 do:10.4236/jsea.2009.22016 Publshed Onlne July 2009 (www.scrp.org/journal/jsea) An Optmal Algorthm for Prufer Codes * Xaodong Wang 1, 2, Le Wang 3,

More information

Face Recognition University at Buffalo CSE666 Lecture Slides Resources:

Face Recognition University at Buffalo CSE666 Lecture Slides Resources: Face Recognton Unversty at Buffalo CSE666 Lecture Sldes Resources: http://www.face-rec.org/algorthms/ Overvew of face recognton algorthms Correlaton - Pxel based correspondence between two face mages Structural

More information

Optimizing Document Scoring for Query Retrieval

Optimizing Document Scoring for Query Retrieval Optmzng Document Scorng for Query Retreval Brent Ellwen baellwe@cs.stanford.edu Abstract The goal of ths project was to automate the process of tunng a document query engne. Specfcally, I used machne learnng

More information

semg Based Movement Quantitative Estimation of Joins Using SVM Method

semg Based Movement Quantitative Estimation of Joins Using SVM Method Preprnts of the 9th Word Congress The Internatona Federaton of Automatc Contro Cape Town, South Afrca. August 24-29, 24 semg Based Movement Quanttatve Estmaton of Jons Usng SVM Method Dongsheng Lu Xngang

More information

Fast Feature Value Searching for Face Detection

Fast Feature Value Searching for Face Detection Vol., No. 2 Computer and Informaton Scence Fast Feature Value Searchng for Face Detecton Yunyang Yan Department of Computer Engneerng Huayn Insttute of Technology Hua an 22300, Chna E-mal: areyyyke@63.com

More information

Final Exam. CSE Section M, Winter p. 1 of 16. Family Name: Given Name(s): Student Number:

Final Exam. CSE Section M, Winter p. 1 of 16. Family Name: Given Name(s): Student Number: Fna Exam CSE 1020 3.0 Secton M, Wnter 2010 p. 1 o 16 Famy Name: Gven Name(s): Student Numer: Gudenes and Instructons: 1. Ths s a 90-mnute exam. You can use the textook, ut no eectronc ads such as cacuators,

More information

Helsinki University Of Technology, Systems Analysis Laboratory Mat Independent research projects in applied mathematics (3 cr)

Helsinki University Of Technology, Systems Analysis Laboratory Mat Independent research projects in applied mathematics (3 cr) Helsnk Unversty Of Technology, Systems Analyss Laboratory Mat-2.08 Independent research projects n appled mathematcs (3 cr) "! #$&% Antt Laukkanen 506 R ajlaukka@cc.hut.f 2 Introducton...3 2 Multattrbute

More information

A Fast Visual Tracking Algorithm Based on Circle Pixels Matching

A Fast Visual Tracking Algorithm Based on Circle Pixels Matching A Fast Vsual Trackng Algorthm Based on Crcle Pxels Matchng Zhqang Hou hou_zhq@sohu.com Chongzhao Han czhan@mal.xjtu.edu.cn Ln Zheng Abstract: A fast vsual trackng algorthm based on crcle pxels matchng

More information

Research Article Fault Isolation for Nonlinear Systems Using Flexible Support Vector Regression

Research Article Fault Isolation for Nonlinear Systems Using Flexible Support Vector Regression Mathematca Probems n Engneerng, Artce ID 738, pages http://dxdoorg/55/24/738 Research Artce Faut Isoaton for Nonnear Systems Usng Fexbe Support Vector Regresson Yufang Lu,,2 Bn Jang, Hu Y, 3 and Cume Bo

More information

Signature and Lexicon Pruning Techniques

Signature and Lexicon Pruning Techniques Sgnature and Lexcon Prunng Technques Srnvas Palla, Hansheng Le, Venu Govndaraju Centre for Unfed Bometrcs and Sensors Unversty at Buffalo {spalla2, hle, govnd}@cedar.buffalo.edu Abstract Handwrtten word

More information

Positive Semi-definite Programming Localization in Wireless Sensor Networks

Positive Semi-definite Programming Localization in Wireless Sensor Networks Postve Sem-defnte Programmng Localzaton n Wreless Sensor etworks Shengdong Xe 1,, Jn Wang, Aqun Hu 1, Yunl Gu, Jang Xu, 1 School of Informaton Scence and Engneerng, Southeast Unversty, 10096, anjng Computer

More information

Magnetic Resonance Images (MRI) Classificatıon through Support Vector Machine

Magnetic Resonance Images (MRI) Classificatıon through Support Vector Machine J. App. Envron. Bo. Sc., 7(4)30-36, 07 07, TextRoad Pubcaton ISSN: 090-474 Journa of Apped Envronmenta and Boogca Scences.textroad.com Magnetc Resonance Images (MRI) Cassfcatıon through Support Vector

More information

Abnormal Crowd Behavior Detection Using Size-Adapted Spatio-Temporal Features

Abnormal Crowd Behavior Detection Using Size-Adapted Spatio-Temporal Features Internatona Journa of Contro, Automaton, and Systems (2) 9(5):95-92 DOI.7/s2555--5-x ttp://www.sprnger.com/2555 Abnorma Crowd Beavor Detecton Usng Sze-Adapted Spato-Tempora Features Bo Wang, Mao Ye, Xue

More information

S1 Note. Basis functions.

S1 Note. Basis functions. S1 Note. Bass functons. Contents Types of bass functons...1 The Fourer bass...2 B-splne bass...3 Power and type I error rates wth dfferent numbers of bass functons...4 Table S1. Smulaton results of type

More information

Nearest Neighbor Learning

Nearest Neighbor Learning Nearest Neighbor Learning Cassify based on oca simiarity Ranges from simpe nearest neighbor to case-based and anaogica reasoning Use oca information near the current query instance to decide the cassification

More information

Empirical Distributions of Parameter Estimates. in Binary Logistic Regression Using Bootstrap

Empirical Distributions of Parameter Estimates. in Binary Logistic Regression Using Bootstrap Int. Journal of Math. Analyss, Vol. 8, 4, no. 5, 7-7 HIKARI Ltd, www.m-hkar.com http://dx.do.org/.988/jma.4.494 Emprcal Dstrbutons of Parameter Estmates n Bnary Logstc Regresson Usng Bootstrap Anwar Ftranto*

More information