Fast Robust Non-Negative Matrix Factorization for Large-Scale Human Action Data Clustering
|
|
- Aleesha Haynes
- 6 years ago
- Views:
Transcription
1 roceedngs of the Twenty-Ffth Internatonal Jont Conference on Artfcal Intellgence IJCAI-6) Fast Robust Non-Negatve Matrx Factorzaton for Large-Scale Human Acton Data Clusterng De Wang, Fepng Ne, Heng Huang Department of Computer Scence and Engneerng Unversty of Texas at Arlngton Abstract Human acton recognton s mportant n mprovng human lfe n varous aspects. However, the outlers and nose n data often bother the clusterng tasks. Therefore, there s a great need for the robust data clusterng technques. Nonnegatve matrx factorzaton ) and Nonnegatve Matrx Tr-Factorzaton NMTF) methods have been wdely researched these years and appled to many data clusterng applcatons. Wth the presence of outlers, most prevous /NMTF models fal to acheve the optmal clusterng performance. To address ths challenge, n ths paper, we propose three new and NMTF models whch are robust to outlers. Effcent algorthms are derved, whch converge much faster than prevous methods and as fast as K-means algorthm, and scalable to large-scale data sets. Expermental results on both synthetc and real world data sets show that our methods outperform other and NMTF methods n most cases, and n the meanwhle, take much less computatonal tme. Introducton Human acton recognton s very useful n mprovng human lfe n varous aspects, lke smart home applcatons, rehabltaton, human-computer nteracton, etc. To recognze human acton s helpful to know one s ntent, such that an ntellgent system could make correspondng reactons. However, n the real world lfe, the data are not as clean as we want, and there mght be many outlers and nose. For example, t s common that part of the human body s shelded behnd an obstacle, whch makes the human acton dffcult to recognze. Therefore, robust data mnng technques should be developed n order to acheve good performance. Non-negatve matrx factorzaton ) and nonnegatve matrx tr-factorzaton NMTF) are models whch am to factorze a matrx nto non-negatve matrces wth mnmum reconstructon error. It has been wdely researched To whom all correspondence should be addressed. Ths work was partally supported by US NSF-IIS 7965, NSF-IIS 32675, NSF-IIS 34452, NSF-DBI , NIH R AG4937. and used n varous knds of applcatons, lke document coclusterng [L and Dng, 26], computer vson [Lee and Seung, 999], socal network analyss [Huang et al., 23], bonformatcs [Wang et al., 22; 24b], knowledge transfer [Wang et al., 2; 25] and many others. Dfferent knds of and NMTF models are proposed these years. The standard factorze a non-negatve matrx nto two non-negatve matrces. [Dng et al., 25] dscussed the equvalence relatonshp between K-means, and spectral clusterng. Later, [Dng et al., 2] proposed two models: sem- and convex. In many stuatons, the data matrx may contan negatve elements. In such stuaton, t s not sutable to enforce the non-negatve constrants on model factors. Dfferent from standard, sem- relaxed the non-negatve constrant on models, allowed one factor to be mx-sgned. In convex, the author restrcts that the centrod to be a lnear combnaton of samples. Such model has better nterpretablty, however, the performance may not be as good as other models [L and Dng, 26]. [Dng et al., 26] proposed to add orthogonalty constrants n the cluster ndcator factor G. Ths property wll enforce the soluton of G to have a clear cluster structure, whch s desrable for usng for clusterng. It s ponted out n [Gu et al., 2] that the orthogonal constrant can prevent trval soluton of models. [Gu and Zhou, 29] combnes Laplacan graph based algorthms wth to enforce the smoothness on data manfolds. [Kong et al., 2; Gao et al., 25] proposed robust models whch are robust to outlers. Most prevous works can not acheve good clusterng results when there exst outlers n the data. resence of outlers s very common n real world applcatons. For example, n human acton recognton, t s common that part of the human body s shelded behnd an obstacle, whch makes the human acton dffcult to recognze. Therefore, robust data mnng technques should be developed n order to acheve good performance. In vew of the above consderatons, we propose three and NMTF models whch converge as fast as K-means, and are robust to outlers. Notatons: Gven a matrx 2 R n m, ts -th row and j-th column are denoted by.,.j, respectvely. The `r,p -norm of a matrx s defned as: 24
2 kk r,p = n = m j= x j r ) p r ) p. Accordng to ths defnton, we can get the followng norms: v n u m n kk 2, = t x 2 j = x ) 2 = kk = j= n = j= = m x j 2) v u n m kk F = kk = t x 2 j 3) = j= 2 New Fast Algorthms for Robust and NMTF Models 2. Motvatons models often have the followng form of objectve functons, subjectng to dfferent knds of constrants. mn F,G FGT 2 s.t. F,G 4) F where 2 R d n + s a data matrx wth d features and n samples. Such models have close relaton to K-means clusterng as ponted out n [Dng et al., 25]. F 2 R d c + can be vewed as cluster centrods, and G 2 R n c + can be vewed as clusterng ndcator matrx. Many prevous work [Dng et al., 2; 26; Lee and Seung, 2; Kong et al., 2] use a soft ndcator matrx G,.e., elements n G are contnuous values. Algorthms for those models are based on matrx multplcatve updatng rules. However, these algorthms converge very slowly, often need hundreds or even thousands of teratons before convergence. Consderng the computatonal cost of large matrx multplcaton, such algorthms usually take a long tme to converge. In addton, an addtonal postprocessng step need to be performed to get the fnal clusterng labels usng the soft ndcator matrx G. On the other hand, outlers are very common n real world data problems. For example, n mage recognton problems, there maybe dfferent extent of nose due to dfferent condtons of llumnatons, vewponts, wearngs, etc. Also, human measurement and recordng errors may also ncur many outlers. revous models usng Frobenus norm as loss measurement can not get good clusterng results f there exst outlers/nose n the data. Ths s because the loss term wll be domnated by outlers, thus the loss of other normal data ponts wll be dsregarded. We wll show ths usng a synthetc data set n the expermental secton. Therefore, developng robust models to handle data wth outlers s mportant. To address the above challenges, n ths paper, we propose three new fast robust and NMTF models. 2.2 New Fast Robust and NMTF Models To make the /NMTF models robust to outlers, nstead of usng Frobenus norm, we propose to use the `2, -norm and `-norm as loss measurements. The new robust and fast models am to mnmze the followng objectve functons: mn F,G2Ind FGT 5) mn F,G2Ind FGT 6) 2, mn F 2Ind,G2Ind,S FSGT 7) where G 2 Ind or F 2 Ind ndcates that G and F are ndcator matrces,.e. g j =f x belongs to class j, and g j =otherwse. There s only one element can be nonzero n each row of bnary ndcator matrx. Note that n ths way, the constrant G T G = I n orthogonal [Dng et al., 26] s automatcally satsfed. In addton, clusterng labels are drectly obtaned wthout the need for further postprocessng as n prevous works. Whle F ndcates that F s a non-negatve matrx wth all elements greater or equal to zero. In the tr-factorzaton model 7), snce both F and G are restrcted to be bnary ndcators, S s ntroduced to absorb the magntude n the orgnal data matrx. Snce our methods are robust and fast /NMTF models, we call the three models as L,, and, respectvely. `2, -norm and `-norm are often used for enforcng sparsty when appled to regularze parameter matrx [Ne et al., 2; Wang et al., 24a; 23], and achevng robustness to outlers when appled to loss functon [Kong et al., 2]. Laplacan Nose Interpretaton for L: We show the probablstc motvaton of our model from Laplacan nose dstrbuton. Suppose x s an observed data pont contamnated by an addtonal nose : x = + 8) where s the unobservable true data, n clusterng t s the clusterng centrod,.e. = FG T., G. 2 Ind. s the nose. Suppose the nose s drawn from Laplacan dstrbuton wth zero mean, then we have: px )= 2b exp kx k ) 9) b where b s the scale parameter of Laplacan dstrbuton. Suppose we have an observed data set =[x,x 2,...,x N ], the maxmum lkelhood estmate MLE) of should be: max log N N px ) ) max = b kx = k ) N N mn kx k kx = F,G.2Ind = FG. k N k FGk ) F,G2Ind = Therefore, the MLE under Laplacan nose assumpton s equvalent to the L model. In the algorthm secton, we wll show that the soluton of cluster centrods F s exactly fndng the sample medans, whch concdes wth the MLE of the locaton parameter of Laplacan dstrbuton. Smlar to the Laplacan nose nterpretaton of L, t s also not dffcult to see that: the 25
3 model can also be nterpreted from a Laplacan dstrbuted nose, except s replaced by FSG. nstead of FG. ; the model can be nterpreted from a Gaussan dstrbuted nose. For space reason, we do not show the detal dervaton. The constrants n the objectve functons nvolves combnatoral search over the soluton space of F and G, whch s very hard to optmze. revous works always solve the relaxed verson to make G have contnuous values. In the followng secton, effcent algorthms are proposed for solvng the objectve functons wth bnary ndcator constrants. 3 Optmzaton Algorthms We frst ntroduce a lemma whch wll be used n the followng optmzaton algorthms. Lemma. Consderng the followng objectve functon: mn z a ) z The optmal soluton of z s the medan value of a. Ths lemma can be easly obtaned by settng the dervatve of Eq. to zero: sgnz a )= 2) where sgnx) =f x>, sgnx) = f x<, and sgnx) =f x =. Ths condton can be satsfed f and only f z takes the medan value of a. 3. Effcent Algorthm for roblem 5) Gven an ntal guess of F and G, we teratvely update the model parameters. When G s fxed, problem 5) becomes: mn F FGT 3) Snce `-norm can be decoupled through rows and columns, the above problem can be reduced to: F. F k F k G Ṭ k k j F k G jk = The above problem can be decoupled as: for 8, k, solvng: mn j F k 4) F k G jk = Accordng to Lemma, the optmal soluton of F K can be effcently obtaned by fndng the medan values of samples belong to the k-th cluster. When F s fxed, G can be effcently optmzed by assgnng label to the cluster wth nearest centrod,.e.: j = arg mn k. F.k k g j = k 5) otherwse 3.2 Effcent Algorthm for roblem 6) When G s fxed, usng reweghted method, problem 6) can be reduced to: mn F Tr FGT )D FG T ) T ) 6) where D 2 R n n s a dagonal matrx whose dagonal elements are: D = 2. FG T 7). Settng the dervatve w.r.t. F to zero, we get: F = DGG T DG) 8) When F s fxed, G can be obtaned by assgnng labels to the cluster wth the nearest centrod: j = arg mn k. F.k k 2 g j = k 9) otherwse 3.3 Effcent Algorthm for roblem 7) When F and G are fxed, problem Eq. 7) becomes: S FSG T S S S kl F.k G Ṭ l k,l k,l F k =,G jl = S kl j Ths problem can be decoupled as: 8k, l, solvng: S kl j 2) S kl F k =,G jl = Accordng to Lemma, S can be effcently optmzed by fndng the medan values. When S and F are fxed, G can be obtaned by: j = arg mn k. FS.k k g j = k 2) otherwse When S and G are fxed, F can be obtaned by: j = arg mn. S k. G T f j = k otherwse 22) Therefore, all the three models can be effcently solved by teratvely update F and G. Unlke prevous works whch need to ntalze G usng the clusterng results of K-means, our algorthms works good wth random ntalzaton of G. In practce, we can repeat the algorthm many tmes usng dfferent random ntalzatons, and then take the result wth the mnmum objectve value. Note that our algorthms use bnary ndcator matrx nstead of soft ndcator matrx. Therefore, the clusterng labels can be obtaned drectly by assgnng labels to the cluster wth nearest centrod. Therefore, our algorthms avod the ntensve computaton of matrx multplcaton. More mportantly, prevous /NMTF works converges slowly, often needs hundreds or even thousands teratons before convergence. Our algorthms converge very fast, usually n just tens of teratons, and take much less computatonal tme. 26
4 3.4 Convergence Analyss In ths secton, we prove the convergence of our algorthms. To begn wth, followng are some smple defntons that wll be used. Jf,f )= x f s f) 23) where x s the -th data, f s the cluster centrod of the prevous teraton, f s the new cluster centrod, s f) denotes the clusterng assgnment of the -th data usng the cluster centrod of prevous step. Obvously, when f = f = f t, Ef t ) = Jf t,f t ) s exactly the objectve value of L n Eq. 5) n the t-th teraton. Accordng to Lemma : the optmum soluton of f s the medan value of samples, and we have Jf,f ) apple Jf,f). In addton, snce by defnton s f ) s the best cluster assgnment usng cluster centrod f, we have: Ef ) Jf,f ) = Jf,f ) Jf,f ) = x fs f ) x fs f) apple 24) So we have: Ef ) Ef) = Ef ) Jf,f )+Jf,f ) Jf,f) apple Jf,f ) Jf,f) apple 25) Therefore, our algorthm for solvng problem 5) monotoncally decreases the objectve value at each teraton untl t reaches the optmum soluton where f = f. In addton, the KKT condton s satsfed. So the algorthm converges to a local mnmum. The convergence proof of the other two algorthms follows a smlar procedure. For space reasons, we omt the detaled proof. 4 Expermental Results 4. Experment Settngs In order to valdate the effectveness of the proposed and NMTF methods, we compare our methods wth some related methods as followng: ) Standard ) [Lee and Seung, 2]: solves the objectve functon n Eq. 4). 2) Orthogonal NMTF Orth) [Dng et al., 26]: factorzes a matrx nto three non-negatve components, and each column of the soft ndcator matrces F and G) are requred to be orthogonal. 3) Sem [Dng et al., 2]: allows the bass matrx F n standard to be mx-sgned 4) Convex Conv) [Dng et al., 2]: restrcts the bass matrx F nto a lnear combnaton of orgnal data ponts. 5) Robust R) [Kong et al., 2]: replaces the loss measurement n standard from Frobenus norm to `2, -norm, whch makes the model robust to outlers. The dfference between our methods and R s that: our methods Table : Average dstance from the centrods for normal data blue and red ponts n fgure a)), outlers, and all data. normal data outlers all data Our methods use a hard ndcator, whch s more dffcult to solve, but has clearer clusterng nterpretaton snce they reduce to drectly assgn cluster labels to samples at each updates of ndcator matrx. Ths approach s better for clusterng tasks. In addton, the algorthm s also much faster as we wll show emprcally n the experment secton. All the above models are solved usng a multplcatve updatng rules, whch converges slowly and nvolves ntensve matrx computatons. K-means serves as a baselne n the clusterng performance comparson. The above comparson methods need to be ntalzed wth the clusterng results of K-means, snce the objectve functons are non-convex, and the fnal clusterng results are senstve to ntalzatons. A lttle perturbaton s needed for preventng these models from stckng at the K-means solutons. As suggested n prevous researches [Kong et al., 2; Dng et al., 26], the ntalzaton of ndcator matrx s set as G =G k +.2, where G k s the result of the K-means algorthm. The soluton of these comparson methods s then get by teratvely update F and G usng the multplcatve rule n Table n [L and Dng, 26]. The teraton number s set as 5. As for our methods, the ntalzaton s set as the followng: for and L, rather than usng the result of K-means, we just randomly ntalze the cluster ndcator matrx G n case of the model stuck at the soluton of K-means. We can see n the experments that our objectve converges good on a local mnmum even wth random ntalzaton. For, the ntalzaton of G and F s usng the clusterng results of the data dmenson and feature dmenson, smlar to the strategy used n tradtonal methods. 4.2 Experments on Synthetc Data We compare our methods wth standard on a synthetc data set wth outlers. As shown n Fgure a), the data set contans two normal clusters blue ponts and red ponts) drawn from two gaussan dstrbutons wth mean -6, ) and 6,), respectvely. Each cluster contans data ponts. Three outlers black ponts) were added far away from these two clusters. Fgure b) shows the clusterng results usng standard algorthm. Because standard s not robust to outlers, the orgnal blue/red ponts n one cluster are splt nto two clusters. Other comparson methods follow the same pattern as standard,.e. splt the orgnal cluster nto two clusters. Fgure c) shows the clusterng results usng the proposed three methods. Our methods are robust to outlers, therefore, the correct clusterng structure s recovered. Table reports the average dstance of data ponts from the correspondng cluster centrods. We can see that the dstance of standard algorthm for outlers s a lttle smaller than 27
5 a) Orgnal data wth outlers 5 b) Clusterng results wth standard 5 nn c) Clusterng results wth our three methods Fgure : Clusterng performance on synthetc data. Blue ponts and red ponts are normal data drawn from two gaussan dstrbutons. Black ponts are outlers. Magenta ponts are computed cluster centrods. Accuracy _L OrthNMTF Sem Conv R k means NMI _L OrthNMTF Sem Conv R k means purty _L OrthNMTF Sem Conv R k means Objectve value 2.5 x _L DIGIT HumanEvaYouTube KTH UCF a) Accuracy.2. DIGIT HumanEvaYouTube KTH UCF b) NMI.2. DIGIT HumanEvaYouTube KTH UCF c) purty Number of teratons d) DIGIT Fgure 2: a)-c): Clusterng performance comparson d): objectve value versus number of teratons. Table 2: Computatonal tme n seconds) comparson. Averaged over repettons. L OrthNMTF Sem Conv R K-means DIGIT HumanEva YouTube KTH UCF our methods, snce t pays more attenton to outlers. The dstances of our methods for normal data and for all data are much more smaller than standard. 4.3 Data Set Descrptons In ths secton, we wll present emprcal results to evaluate the proposed and NMTF approaches. Fve real world data sets are used to evaluate the effectveness of our method. Dgt data set s a publc data set hosted n UCI Machne Learnng Repostory. Ths data set conssts of handwrtten dgts from to 9, and each dgt s a class. HumanEva data set 2 contans samples from two subjects wth 5 samples per subject. There are fve types of motons: boxng, walkng, throw-catch, joggng and gesturng. YouTube data set 3 contans human actvtes: soccer lujg/youtube Acton dataset.html jugglng, swngng, tenns swngng, to name a few. Fgure 3 show some sample mages wth boundng acton labels n YouTube data set. UCF data set 4 s an extenson to the YouTube data sets. It contans 5 actons consstng of realstc vdeos n YouTube. Ths data set s very hard to recognze due to large varatons n camera poston, pose, llumnaton condtons, vewpont, and cluttered background. KTH data set 5 contans sx type of human actons: walkng, joggng, runnng, boxng, hand wavng and hand clappng. These actons are performed by 25 subjects n four dfferent scenaros: outdoor, outdoor wth scale varatons, outdoor wth dfferent clothes, and ndoors. All samples are taken n homogeneous background, so that t s more close to natural envronments. Ths settng add nose n the sample, thus, t s more dffcult to recognze the actons
6 Fgure 3: Sample mages from the webste of the YouTube data set. 4.4 Clusterng erformance and Convergence Speed Comparson We use accuracy, NMI normalzed mutual nformaton), and urty as measurements of the clusterng performance. Snce all the methods are senstve to ntalzatons, we repeat each algorthm tmes, and the clusterng results wth the mnmum objectve value s recorded n Fgure 2. We can see that: L and acheve the best performance on most data sets n terms of both accuracy, NMI and purty. The performance of OrthNMTF and R s also pretty good on some data sets. Fgure 2 d) shows the convergence of the proposed algorthms. For space lmtaton, only one of the fve data sets are shown n the paper. We can see that all the three algorthms converge fast, usually n about 5 teratons. Table 2 shows the comparson of computatonal tme. The algorthms are run on a Dell desktop wth double 7 Cores and 6GB memory. For and L, we just use random ntalzaton n the experments. For all of the other methods, K-means result s used as ntalzaton snce t s suggested by prevous research. The tme consumpton of K-means ntalzaton s not consdered for all of these methods. We can see that the L algorthm converges very fast, whch s often comparable or even faster than K- means. Thus, our algorthm s scalable to large data. The computatonal tme of and s also much less than other compared and NMTF methods. 5 Conclusons We proposed three new and NMTF models whch are robust to outlers to mprove the human acton clusterng tasks. Effcent algorthms are derved, whch converge as fast as the standard K-means algorthm, and thus are scalable to large-scale data sets. Expermental results on both synthetc and real world data sets show that our methods outperform other exstng and NMTF methods n most cases, and n the meanwhle, take much less computatonal tme. References [Dng et al., 25] C. Dng,. He, and H.D. Smon. On the equvalence of nonnegatve matrx factorzaton and spectral clusterng. In SDM, pages 66 6, 25. [Dng et al., 26] Chrs Dng, Tao L, We eng, and Haesun ark. Orthogonal nonnegatve matrx t-factorzatons for clusterng. In roceedngs of the 2th ACM SIGKDD nternatonal conference on Knowledge dscovery and data mnng, pages ACM, 26. [Dng et al., 2] Chrs Dng, Tao L, and Mchael I Jordan. Convex and sem-nonnegatve matrx factorzatons. attern Analyss and Machne Intellgence, IEEE Transactons on, 32):45 55, 2. [Gao et al., 25] Hongchang Gao, Fepng Ne, Wedong Ca, and Heng Huang. Robust capped norm nonnegatve matrx factorzaton. 24th ACM Internatonal Conference on Informaton and Knowledge Management CIKM 25), pages 87 88, 25. [Gu and Zhou, 29] Quanquan Gu and Je Zhou. Neghborhood preservng nonnegatve matrx factorzaton. In roceedngs of the Brtsh machne vson conference, 29. [Gu et al., 2] Quanquan Gu, Chrs Dng, and Jawe Han. On trval soluton and scale transfer problems n graph regularzed nmf. In roceedngs of the nternatonal jont conference on artfcal ntellgence, 2. [Huang et al., 23] Jn Huang, Fepng Ne, Heng Huang, Ycheng Tu, and Yu Le. Socal trust predcton usng heterogeneous networks. ACM Transactons on Knowledge Dscovery from Data TKDD), 74):7: 7:2,
7 [Kong et al., 2] Deguang Kong, Chrs Dng, and Heng Huang. Robust nonnegatve matrx factorzaton usng l2,-norm. In ACM Conference on Informaton and Knowledge Management CIKM 2), 2. [Lee and Seung, 999] Danel D Lee and H Sebastan Seung. Learnng the parts of objects by non-negatve matrx factorzaton. Nature, 46755):788 79, 999. [Lee and Seung, 2] D.D. Lee and H.S. Seung. Algorthms for nonnegatve matrx factorzaton. In NIS, pages , 2. [L and Dng, 26] Tao L and Chrs Dng. The relatonshps among varous nonnegatve matrx factorzaton methods for clusterng. In Data Mnng, 26. ICDM 6. Sxth Internatonal Conference on, pages IEEE, 26. [Ne et al., 2] F. Ne, H. Huang,. Ca, and C. Dng. Effcent and robust feature selecton va jont l2, -norms mnmzaton. Advances n Neural Informaton rocessng Systems, 23:83 82, 2. [Wang et al., 2] Hua Wang, Heng Huang, and Chrs Dng. Cross-language web page classfcaton va jont nonnegatve matrx tr-factorzaton based dyadc knowledge transfer. In Annual ACM SIGIR Conference 2, pages , 2. [Wang et al., 22] Hua Wang, Heng Huang, Chrs Dng, and Fepng Ne. redctng proten-proten nteractons from multmodal bologcal data sources va nonnegatve matrx tr-factorzaton. The 6th Internatonal Conference on Research n Computatonal Molecular Bology RECOMB), pages , 22. [Wang et al., 23] De Wang, Fepng Ne, Heng Huang, Jngwen Yan, Shannon L Rsacher, Andrew J Saykn, and L Shen. Structural bran network constraned neuromagng marker dentfcaton for predctng cogntve functons. In Informaton rocessng n Medcal Imagng, pages Sprnger Berln Hedelberg, 23. [Wang et al., 24a] De Wang, Fepng Ne, and Heng Huang. Unsupervsed feature selecton va unfed trace rato formulaton and k-means clusterng track). In Machne Learnng and Knowledge Dscovery n Databases, 24. [Wang et al., 24b] Hua Wang, Heng Huang, and Chrs Dng. Correlated proten functon predcton va maxmzaton of data-knowledge consstency. The 8th Internatonal Conference on Research n Computatonal Molecular Bology RECOMB 24), pages 3 325, 24. [Wang et al., 25] Hua Wang, Fepng Ne, and Heng Huang. Large-scale cross-language web page classfcaton va dual knowledge transfer usng fast nonnegatve matrx tr-factorzaton. ACM Transactons on Knowledge Dscovery from Data TKDD), ), 25. 2
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 informationSubspace 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 informationOutline. Type of Machine Learning. Examples of Application. Unsupervised Learning
Outlne Artfcal Intellgence and ts applcatons Lecture 8 Unsupervsed Learnng Professor Danel Yeung danyeung@eee.org Dr. Patrck Chan patrckchan@eee.org South Chna Unversty of Technology, Chna Introducton
More informationParallelism 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 informationUnsupervised Learning
Pattern Recognton Lecture 8 Outlne Introducton Unsupervsed Learnng Parametrc VS Non-Parametrc Approach Mxture of Denstes Maxmum-Lkelhood Estmates Clusterng Prof. Danel Yeung School of Computer Scence and
More informationThe Research of Support Vector Machine in Agricultural Data Classification
The Research of Support Vector Machne n Agrcultural Data Classfcaton Le Sh, Qguo Duan, Xnmng Ma, Me Weng College of Informaton and Management Scence, HeNan Agrcultural Unversty, Zhengzhou 45000 Chna Zhengzhou
More informationDiscriminative Dictionary Learning with Pairwise Constraints
Dscrmnatve Dctonary Learnng wth Parwse Constrants Humn Guo Zhuoln Jang LARRY S. DAVIS UNIVERSITY OF MARYLAND Nov. 6 th, Outlne Introducton/motvaton Dctonary Learnng Dscrmnatve Dctonary Learnng wth Parwse
More informationPositive 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 informationRobust Dictionary Learning with Capped l 1 -Norm
Proceedngs of the Twenty-Fourth Internatonal Jont Conference on Artfcal Intellgence (IJCAI 205) Robust Dctonary Learnng wth Capped l -Norm Wenhao Jang, Fepng Ne, Heng Huang Unversty of Texas at Arlngton
More informationNUMERICAL SOLVING OPTIMAL CONTROL PROBLEMS BY THE METHOD OF VARIATIONS
ARPN Journal of Engneerng and Appled Scences 006-017 Asan Research Publshng Network (ARPN). All rghts reserved. NUMERICAL SOLVING OPTIMAL CONTROL PROBLEMS BY THE METHOD OF VARIATIONS Igor Grgoryev, Svetlana
More informationLearning a Class-Specific Dictionary for Facial Expression Recognition
BULGARIAN ACADEMY OF SCIENCES CYBERNETICS AND INFORMATION TECHNOLOGIES Volume 16, No 4 Sofa 016 Prnt ISSN: 1311-970; Onlne ISSN: 1314-4081 DOI: 10.1515/cat-016-0067 Learnng a Class-Specfc Dctonary for
More informationRange images. Range image registration. Examples of sampling patterns. Range images and range surfaces
Range mages For many structured lght scanners, the range data forms a hghly regular pattern known as a range mage. he samplng pattern s determned by the specfc scanner. Range mage regstraton 1 Examples
More informationA Binarization Algorithm specialized on Document Images and Photos
A Bnarzaton Algorthm specalzed on Document mages and Photos Ergna Kavalleratou Dept. of nformaton and Communcaton Systems Engneerng Unversty of the Aegean kavalleratou@aegean.gr Abstract n ths paper, a
More informationClassifier 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 informationHierarchical clustering for gene expression data analysis
Herarchcal clusterng for gene expresson data analyss Gorgo Valentn e-mal: valentn@ds.unm.t Clusterng of Mcroarray Data. Clusterng of gene expresson profles (rows) => dscovery of co-regulated and functonally
More informationTsinghua University at TAC 2009: Summarizing Multi-documents by Information Distance
Tsnghua Unversty at TAC 2009: Summarzng Mult-documents by Informaton Dstance Chong Long, Mnle Huang, Xaoyan Zhu State Key Laboratory of Intellgent Technology and Systems, Tsnghua Natonal Laboratory for
More informationNew l 1 -Norm Relaxations and Optimizations for Graph Clustering
Proceedngs of the Thrteth AAAI Conference on Artfcal Intellgence (AAAI-6) New l -Norm Relaxatons and Optmzatons for Graph Clusterng Fepng Ne, Hua Wang, Cheng Deng 3, Xnbo Gao 3, Xuelong L 4, Heng Huang
More informationHigh-Boost Mesh Filtering for 3-D Shape Enhancement
Hgh-Boost Mesh Flterng for 3-D Shape Enhancement Hrokazu Yagou Λ Alexander Belyaev y Damng We z Λ y z ; ; Shape Modelng Laboratory, Unversty of Azu, Azu-Wakamatsu 965-8580 Japan y Computer Graphcs Group,
More informationSupport 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 information6.854 Advanced Algorithms Petar Maymounkov Problem Set 11 (November 23, 2005) With: Benjamin Rossman, Oren Weimann, and Pouya Kheradpour
6.854 Advanced Algorthms Petar Maymounkov Problem Set 11 (November 23, 2005) Wth: Benjamn Rossman, Oren Wemann, and Pouya Kheradpour Problem 1. We reduce vertex cover to MAX-SAT wth weghts, such that the
More informationHybridization of Expectation-Maximization and K-Means Algorithms for Better Clustering Performance
BULGARIAN ACADEMY OF SCIENCES CYBERNETICS AND INFORMATION TECHNOLOGIES Volume 16, No 2 Sofa 2016 Prnt ISSN: 1311-9702; Onlne ISSN: 1314-4081 DOI: 10.1515/cat-2016-0017 Hybrdzaton of Expectaton-Maxmzaton
More informationAn Application of the Dulmage-Mendelsohn Decomposition to Sparse Null Space Bases of Full Row Rank Matrices
Internatonal Mathematcal Forum, Vol 7, 2012, no 52, 2549-2554 An Applcaton of the Dulmage-Mendelsohn Decomposton to Sparse Null Space Bases of Full Row Rank Matrces Mostafa Khorramzadeh Department of Mathematcal
More informationClassification / Regression Support Vector Machines
Classfcaton / Regresson Support Vector Machnes Jeff Howbert Introducton to Machne Learnng Wnter 04 Topcs SVM classfers for lnearly separable classes SVM classfers for non-lnearly separable classes SVM
More informationGSLM 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 informationCS 534: Computer Vision Model Fitting
CS 534: Computer Vson Model Fttng Sprng 004 Ahmed Elgammal Dept of Computer Scence CS 534 Model Fttng - 1 Outlnes Model fttng s mportant Least-squares fttng Maxmum lkelhood estmaton MAP estmaton Robust
More informationProblem 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 informationA Robust Method for Estimating the Fundamental Matrix
Proc. VIIth Dgtal Image Computng: Technques and Applcatons, Sun C., Talbot H., Ourseln S. and Adraansen T. (Eds.), 0- Dec. 003, Sydney A Robust Method for Estmatng the Fundamental Matrx C.L. Feng and Y.S.
More informationDetermining 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 informationSupport 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 information5 The Primal-Dual Method
5 The Prmal-Dual Method Orgnally desgned as a method for solvng lnear programs, where t reduces weghted optmzaton problems to smpler combnatoral ones, the prmal-dual method (PDM) has receved much attenton
More informationA 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 informationModule Management Tool in Software Development Organizations
Journal of Computer Scence (5): 8-, 7 ISSN 59-66 7 Scence Publcatons Management Tool n Software Development Organzatons Ahmad A. Al-Rababah and Mohammad A. Al-Rababah Faculty of IT, Al-Ahlyyah Amman Unversty,
More informationMulti-Source Multi-View Clustering via Discrepancy Penalty
Mult-Source Mult-Vew Clusterng va Dscrepancy Penalty Wexang Shao, Jawe Zhang, Lfang He, Phlp S. Yu Unversty of Illnos at Chcago Emal: wshao4@uc.edu, jzhan9@uc.edu, psyu@uc.edu Shenzhen Unversty, Chna Emal:
More informationEfficient Text Classification by Weighted Proximal SVM *
Effcent ext Classfcaton by Weghted Proxmal SVM * Dong Zhuang 1, Benyu Zhang, Qang Yang 3, Jun Yan 4, Zheng Chen, Yng Chen 1 1 Computer Scence and Engneerng, Bejng Insttute of echnology, Bejng 100081, Chna
More informationSmoothing Spline ANOVA for variable screening
Smoothng Splne ANOVA for varable screenng a useful tool for metamodels tranng and mult-objectve optmzaton L. Rcco, E. Rgon, A. Turco Outlne RSM Introducton Possble couplng Test case MOO MOO wth Game Theory
More informationGraph-based Clustering
Graphbased Clusterng Transform the data nto a graph representaton ertces are the data ponts to be clustered Edges are eghted based on smlarty beteen data ponts Graph parttonng Þ Each connected component
More informationOverlapping Clustering with Sparseness Constraints
2012 IEEE 12th Internatonal Conference on Data Mnng Workshops Overlappng Clusterng wth Sparseness Constrants Habng Lu OMIS, Santa Clara Unversty hlu@scu.edu Yuan Hong MSIS, Rutgers Unversty yhong@cmc.rutgers.edu
More informationAdaptive Transfer Learning
Adaptve Transfer Learnng Bn Cao, Snno Jaln Pan, Yu Zhang, Dt-Yan Yeung, Qang Yang Hong Kong Unversty of Scence and Technology Clear Water Bay, Kowloon, Hong Kong {caobn,snnopan,zhangyu,dyyeung,qyang}@cse.ust.hk
More informationHermite Splines in Lie Groups as Products of Geodesics
Hermte Splnes n Le Groups as Products of Geodescs Ethan Eade Updated May 28, 2017 1 Introducton 1.1 Goal Ths document defnes a curve n the Le group G parametrzed by tme and by structural parameters n the
More informationBackpropagation: In Search of Performance Parameters
Bacpropagaton: In Search of Performance Parameters ANIL KUMAR ENUMULAPALLY, LINGGUO BU, and KHOSROW KAIKHAH, Ph.D. Computer Scence Department Texas State Unversty-San Marcos San Marcos, TX-78666 USA ae049@txstate.edu,
More informationSupport Vector Machines. CS534 - Machine Learning
Support Vector Machnes CS534 - Machne Learnng Perceptron Revsted: Lnear Separators Bnar classfcaton can be veed as the task of separatng classes n feature space: b > 0 b 0 b < 0 f() sgn( b) Lnear Separators
More informationAngle-Independent 3D Reconstruction. Ji Zhang Mireille Boutin Daniel Aliaga
Angle-Independent 3D Reconstructon J Zhang Mrelle Boutn Danel Alaga Goal: Structure from Moton To reconstruct the 3D geometry of a scene from a set of pctures (e.g. a move of the scene pont reconstructon
More informationCS434a/541a: Pattern Recognition Prof. Olga Veksler. Lecture 15
CS434a/541a: Pattern Recognton Prof. Olga Veksler Lecture 15 Today New Topc: Unsupervsed Learnng Supervsed vs. unsupervsed learnng Unsupervsed learnng Net Tme: parametrc unsupervsed learnng Today: nonparametrc
More informationMULTISPECTRAL 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 informationComparison of Heuristics for Scheduling Independent Tasks on Heterogeneous Distributed Environments
Comparson of Heurstcs for Schedulng Independent Tasks on Heterogeneous Dstrbuted Envronments Hesam Izakan¹, Ath Abraham², Senor Member, IEEE, Václav Snášel³ ¹ Islamc Azad Unversty, Ramsar Branch, Ramsar,
More informationSum of Linear and Fractional Multiobjective Programming Problem under Fuzzy Rules Constraints
Australan Journal of Basc and Appled Scences, 2(4): 1204-1208, 2008 ISSN 1991-8178 Sum of Lnear and Fractonal Multobjectve Programmng Problem under Fuzzy Rules Constrants 1 2 Sanjay Jan and Kalash Lachhwan
More informationAn 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 informationA Saturation Binary Neural Network for Crossbar Switching Problem
A Saturaton Bnary Neural Network for Crossbar Swtchng Problem Cu Zhang 1, L-Qng Zhao 2, and Rong-Long Wang 2 1 Department of Autocontrol, Laonng Insttute of Scence and Technology, Benx, Chna bxlkyzhangcu@163.com
More informationEcient Computation of the Most Probable Motion from Fuzzy. Moshe Ben-Ezra Shmuel Peleg Michael Werman. The Hebrew University of Jerusalem
Ecent Computaton of the Most Probable Moton from Fuzzy Correspondences Moshe Ben-Ezra Shmuel Peleg Mchael Werman Insttute of Computer Scence The Hebrew Unversty of Jerusalem 91904 Jerusalem, Israel Emal:
More information12/2/2009. Announcements. Parametric / Non-parametric. Case-Based Reasoning. Nearest-Neighbor on Images. Nearest-Neighbor Classification
Introducton to Artfcal Intellgence V22.0472-001 Fall 2009 Lecture 24: Nearest-Neghbors & Support Vector Machnes Rob Fergus Dept of Computer Scence, Courant Insttute, NYU Sldes from Danel Yeung, John DeNero
More informationCLUSTERING that discovers the relationship among data
Ramp-based Twn Support Vector Clusterng Zhen Wang, Xu Chen, Chun-Na L, and Yuan-Ha Shao arxv:82.0370v [cs.lg] 0 Dec 208 Abstract Tradtonal plane-based clusterng methods measure the cost of wthn-cluster
More informationOutline. Self-Organizing Maps (SOM) US Hebbian Learning, Cntd. The learning rule is Hebbian like:
Self-Organzng Maps (SOM) Turgay İBRİKÇİ, PhD. Outlne Introducton Structures of SOM SOM Archtecture Neghborhoods SOM Algorthm Examples Summary 1 2 Unsupervsed Hebban Learnng US Hebban Learnng, Cntd 3 A
More informationSolving two-person zero-sum game by Matlab
Appled Mechancs and Materals Onlne: 2011-02-02 ISSN: 1662-7482, Vols. 50-51, pp 262-265 do:10.4028/www.scentfc.net/amm.50-51.262 2011 Trans Tech Publcatons, Swtzerland Solvng two-person zero-sum game by
More informationLecture 4: Principal components
/3/6 Lecture 4: Prncpal components 3..6 Multvarate lnear regresson MLR s optmal for the estmaton data...but poor for handlng collnear data Covarance matrx s not nvertble (large condton number) Robustness
More informationAn Iterative Solution Approach to Process Plant Layout using Mixed Integer Optimisation
17 th European Symposum on Computer Aded Process Engneerng ESCAPE17 V. Plesu and P.S. Agach (Edtors) 2007 Elsever B.V. All rghts reserved. 1 An Iteratve Soluton Approach to Process Plant Layout usng Mxed
More informationFixing Max-Product: Convergent Message Passing Algorithms for MAP LP-Relaxations
Fxng Max-Product: Convergent Message Passng Algorthms for MAP LP-Relaxatons Amr Globerson Tomm Jaakkola Computer Scence and Artfcal Intellgence Laboratory Massachusetts Insttute of Technology Cambrdge,
More informationR s s f. m y s. SPH3UW Unit 7.3 Spherical Concave Mirrors Page 1 of 12. Notes
SPH3UW Unt 7.3 Sphercal Concave Mrrors Page 1 of 1 Notes Physcs Tool box Concave Mrror If the reflectng surface takes place on the nner surface of the sphercal shape so that the centre of the mrror bulges
More informationEdge 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 informationCluster 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 informationA 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 informationAn Image Compression Algorithm based on Wavelet Transform and LZW
An Image Compresson Algorthm based on Wavelet Transform and LZW Png Luo a, Janyong Yu b School of Chongqng Unversty of Posts and Telecommuncatons, Chongqng, 400065, Chna Abstract a cylpng@63.com, b y27769864@sna.cn
More informationHuman Skeleton Reconstruction for Optical Motion Capture
Journal of Computatonal Informaton Systems 9: 0 (013) 8073 8080 Avalable at http://www.jofcs.com Human Skeleton Reconstructon for Optcal Moton Capture Guanghua TAN, Melan ZHOU, Chunmng GAO College of Informaton
More informationMachine Learning. Topic 6: Clustering
Machne Learnng Topc 6: lusterng lusterng Groupng data nto (hopefully useful) sets. Thngs on the left Thngs on the rght Applcatons of lusterng Hypothess Generaton lusters mght suggest natural groups. Hypothess
More informationWeighted Feature Subset Non-Negative Matrix Factorization and its Applications to Document Understanding
200 IEEE Internatonal Conference on Data Mnng Weghted Feature Subset Non-Negatve Matrx Factorzaton and ts Applcatons to Document Understandng Dngdng Wang Tao L School of Computng and Informaton Scences
More informationMachine 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 informationLearning-Based Top-N Selection Query Evaluation over Relational Databases
Learnng-Based Top-N Selecton Query Evaluaton over Relatonal Databases Lang Zhu *, Wey Meng ** * School of Mathematcs and Computer Scence, Hebe Unversty, Baodng, Hebe 071002, Chna, zhu@mal.hbu.edu.cn **
More informationFuzzy C-Means Initialized by Fixed Threshold Clustering for Improving Image Retrieval
Fuzzy -Means Intalzed by Fxed Threshold lusterng for Improvng Image Retreval NAWARA HANSIRI, SIRIPORN SUPRATID,HOM KIMPAN 3 Faculty of Informaton Technology Rangst Unversty Muang-Ake, Paholyotn Road, Patumtan,
More informationCollaboratively 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 informationRecommended Items Rating Prediction based on RBF Neural Network Optimized by PSO Algorithm
Recommended Items Ratng Predcton based on RBF Neural Network Optmzed by PSO Algorthm Chengfang Tan, Cayn Wang, Yuln L and Xx Q Abstract In order to mtgate the data sparsty and cold-start problems of recommendaton
More informationReal-time Joint Tracking of a Hand Manipulating an Object from RGB-D Input
Real-tme Jont Tracng of a Hand Manpulatng an Object from RGB-D Input Srnath Srdhar 1 Franzsa Mueller 1 Mchael Zollhöfer 1 Dan Casas 1 Antt Oulasvrta 2 Chrstan Theobalt 1 1 Max Planc Insttute for Informatcs
More informationStudy of Data Stream Clustering Based on Bio-inspired Model
, pp.412-418 http://dx.do.org/10.14257/astl.2014.53.86 Study of Data Stream lusterng Based on Bo-nspred Model Yngme L, Mn L, Jngbo Shao, Gaoyang Wang ollege of omputer Scence and Informaton Engneerng,
More informationA Scalable Projective Bundle Adjustment Algorithm using the L Norm
Sxth Indan Conference on Computer Vson, Graphcs & Image Processng A Scalable Projectve Bundle Adjustment Algorthm usng the Norm Kaushk Mtra and Rama Chellappa Dept. of Electrcal and Computer Engneerng
More informationContent 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 informationTerm 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 informationVideo Object Tracking Based On Extended Active Shape Models With Color Information
CGIV'2002: he Frst Frst European Conference Colour on Colour n Graphcs, Imagng, and Vson Vdeo Object rackng Based On Extended Actve Shape Models Wth Color Informaton A. Koschan, S.K. Kang, J.K. Pak, B.
More informationHuman Action Recognition Using Dynamic Time Warping Algorithm and Reproducing Kernel Hilbert Space for Matrix Manifold
IJCTA, 10(07), 2017, pp 79-85 Internatonal Scence Press Closed Loop Control of Soft Swtched Forward Converter Usng Intellgent Controller 79 Human Acton Recognton Usng Dynamc Tme Warpng Algorthm and Reproducng
More informationDynamic Camera Assignment and Handoff
12 Dynamc Camera Assgnment and Handoff Br Bhanu and Ymng L 12.1 Introducton...338 12.2 Techncal Approach...339 12.2.1 Motvaton and Problem Formulaton...339 12.2.2 Game Theoretc Framework...339 12.2.2.1
More informationRelevance Assignment and Fusion of Multiple Learning Methods Applied to Remote Sensing Image Analysis
Assgnment and Fuson of Multple Learnng Methods Appled to Remote Sensng Image Analyss Peter Bajcsy, We-Wen Feng and Praveen Kumar Natonal Center for Supercomputng Applcaton (NCSA), Unversty of Illnos at
More informationCourse Introduction. Algorithm 8/31/2017. COSC 320 Advanced Data Structures and Algorithms. COSC 320 Advanced Data Structures and Algorithms
Course Introducton Course Topcs Exams, abs, Proects A quc loo at a few algorthms 1 Advanced Data Structures and Algorthms Descrpton: We are gong to dscuss algorthm complexty analyss, algorthm desgn technques
More informationFast Computation of Shortest Path for Visiting Segments in the Plane
Send Orders for Reprnts to reprnts@benthamscence.ae 4 The Open Cybernetcs & Systemcs Journal, 04, 8, 4-9 Open Access Fast Computaton of Shortest Path for Vstng Segments n the Plane Ljuan Wang,, Bo Jang
More informationActive Contours/Snakes
Actve Contours/Snakes Erkut Erdem Acknowledgement: The sldes are adapted from the sldes prepared by K. Grauman of Unversty of Texas at Austn Fttng: Edges vs. boundares Edges useful sgnal to ndcate occludng
More informationEVALUATION OF THE PERFORMANCES OF ARTIFICIAL BEE COLONY AND INVASIVE WEED OPTIMIZATION ALGORITHMS ON THE MODIFIED BENCHMARK FUNCTIONS
Academc Research Internatonal ISS-L: 3-9553, ISS: 3-9944 Vol., o. 3, May 0 EVALUATIO OF THE PERFORMACES OF ARTIFICIAL BEE COLOY AD IVASIVE WEED OPTIMIZATIO ALGORITHMS O THE MODIFIED BECHMARK FUCTIOS Dlay
More informationUnsupervised Learning and Clustering
Unsupervsed Learnng and Clusterng Why consder unlabeled samples?. Collectng and labelng large set of samples s costly Gettng recorded speech s free, labelng s tme consumng 2. Classfer could be desgned
More informationQuality Improvement Algorithm for Tetrahedral Mesh Based on Optimal Delaunay Triangulation
Intellgent Informaton Management, 013, 5, 191-195 Publshed Onlne November 013 (http://www.scrp.org/journal/m) http://dx.do.org/10.36/m.013.5601 Qualty Improvement Algorthm for Tetrahedral Mesh Based on
More informationMachine Learning 9. week
Machne Learnng 9. week Mappng Concept Radal Bass Functons (RBF) RBF Networks 1 Mappng It s probably the best scenaro for the classfcaton of two dataset s to separate them lnearly. As you see n the below
More informationTPL-Aware Displacement-driven Detailed Placement Refinement with Coloring Constraints
TPL-ware Dsplacement-drven Detaled Placement Refnement wth Colorng Constrants Tao Ln Iowa State Unversty tln@astate.edu Chrs Chu Iowa State Unversty cnchu@astate.edu BSTRCT To mnmze the effect of process
More informationLoad-Balanced Anycast Routing
Load-Balanced Anycast Routng Chng-Yu Ln, Jung-Hua Lo, and Sy-Yen Kuo Department of Electrcal Engneerng atonal Tawan Unversty, Tape, Tawan sykuo@cc.ee.ntu.edu.tw Abstract For fault-tolerance and load-balance
More informationHelsinki 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 informationA Robust LS-SVM Regression
PROCEEDIGS OF WORLD ACADEMY OF SCIECE, EGIEERIG AD ECHOLOGY VOLUME 7 AUGUS 5 ISS 37- A Robust LS-SVM Regresson József Valyon, and Gábor Horváth Abstract In comparson to the orgnal SVM, whch nvolves a quadratc
More informationAn Internal Clustering Validation Index for Boolean Data
BULGARIAN ACADEMY OF SCIENCES CYBERNETICS AND INFORMATION TECHNOLOGIES Volume 16, No 6 Specal ssue wth selecton of extended papers from 6th Internatonal Conference on Logstc, Informatcs and Servce Scence
More informationA 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 informationSemi-Supervised Kernel Mean Shift Clustering
IEEE TRANSACTIONS ON PATTERN ANALYSIS AND MACHINE INTELLIGENCE, VOL. XX, NO. XX, JANUARY XXXX 1 Sem-Supervsed Kernel Mean Shft Clusterng Saket Anand, Student Member, IEEE, Sushl Mttal, Member, IEEE, Oncel
More informationSLAM Summer School 2006 Practical 2: SLAM using Monocular Vision
SLAM Summer School 2006 Practcal 2: SLAM usng Monocular Vson Javer Cvera, Unversty of Zaragoza Andrew J. Davson, Imperal College London J.M.M Montel, Unversty of Zaragoza. josemar@unzar.es, jcvera@unzar.es,
More informationA Fast Content-Based Multimedia Retrieval Technique Using Compressed Data
A Fast Content-Based Multmeda Retreval Technque Usng Compressed Data Borko Furht and Pornvt Saksobhavvat NSF Multmeda Laboratory Florda Atlantc Unversty, Boca Raton, Florda 3343 ABSTRACT In ths paper,
More informationUnsupervised Learning and Clustering
Unsupervsed Learnng and Clusterng Supervsed vs. Unsupervsed Learnng Up to now we consdered supervsed learnng scenaro, where we are gven 1. samples 1,, n 2. class labels for all samples 1,, n Ths s also
More informationRobust Mean Shift Tracking with Corrected Background-Weighted Histogram
Robust Mean Shft Trackng wth Corrected Background-Weghted Hstogram Jfeng Nng, Le Zhang, Davd Zhang and Chengke Wu Abstract: The background-weghted hstogram (BWH) algorthm proposed n [] attempts to reduce
More informationSHAPE RECOGNITION METHOD BASED ON THE k-nearest NEIGHBOR RULE
SHAPE RECOGNITION METHOD BASED ON THE k-nearest NEIGHBOR RULE Dorna Purcaru Faculty of Automaton, Computers and Electroncs Unersty of Craoa 13 Al. I. Cuza Street, Craoa RO-1100 ROMANIA E-mal: dpurcaru@electroncs.uc.ro
More informationLecture 5: Multilayer Perceptrons
Lecture 5: Multlayer Perceptrons Roger Grosse 1 Introducton So far, we ve only talked about lnear models: lnear regresson and lnear bnary classfers. We noted that there are functons that can t be represented
More informationFast 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 informationA PATTERN RECOGNITION APPROACH TO IMAGE SEGMENTATION
1 THE PUBLISHING HOUSE PROCEEDINGS OF THE ROMANIAN ACADEMY, Seres A, OF THE ROMANIAN ACADEMY Volume 4, Number 2/2003, pp.000-000 A PATTERN RECOGNITION APPROACH TO IMAGE SEGMENTATION Tudor BARBU Insttute
More information