Nonparametric Density Estimation on A Graph: Learning Framework, Fast Approximation and Application in Image Segmentation

Size: px
Start display at page:

Download "Nonparametric Density Estimation on A Graph: Learning Framework, Fast Approximation and Application in Image Segmentation"

From this document you will learn the answers to the following questions:

  • What is the ernel center located at a tree node?

  • What is the densty estmator defned wth respect to?

Transcription

1 Nonparametrc Densty Estmaton on A Graph: Learnng Framewor, Fast Approxmaton and Applcaton n Image Segmentaton Zhdng Yu Oscar C. Au Ketan Tang Chunjng Xu Dept. of Electronc and Computer Engneerng Hong Kong Unversty of Scence and Technology {zdyu, eeau, tt}@ust.h Shenzhen Inst. of Advanced Technology Chnese Academy of Scences cj.xu@sat.ac.cn Abstract We present a novel framewor for tree-structure embedded densty estmaton and ts fast approxmaton for mode seeng. The proposed method could fnd dverse applcatons n computer vson and feature space analyss. Gven any undrected, connected and weghted graph, the densty functon s defned as a jont representaton of the feature space and the dstance doman on the graph s spannng tree. Snce the dstance doman of a tree s a constraned one, mode seeng can not be drectly acheved by tradtonal mean shft n both doman. we address ths problem by ntroducng node shftng wth force competton and ts fast approxmaton. Our wor s closely related to the prevous lterature of nonparametrc methods. One shall see, however, that the new formulaton of ths problem can lead to many advantages and new characterstcs n ts applcaton, as wll be llustrated later n ths paper. 1. Introducton Nonparametrc densty estmaton provdes a versatle tool for feature space analyss such as clusterng and local maxma detecton. The ratonale behnd, as ponted out by Comancu et al., s that feature space can be regarded as the emprcal probablty densty functon pdf) of the represented parameter. Fndng local estmated densty maxma or mode seeng) results n the computatonal module of mean shftv [1], an old pattern recognton technque. The robust nature of mean shft leads to wde applcatons n low level computer vson, ncludng edge preserved smoothng, mage segmentaton and object tracng. Recent wors tres to mprove ts performance by ntroducng asymmetrc based ernels n specfc tass, or sees to reduce ts complexty wth fast algorthms. Ths wor has been supported n part by the Research Grants Councl RGC) of the Hong Kong Specal Admnstratve Regon, Chna. GRF ) Fgure 1. Example of data clusterng usng the proposed mode seeng algorthm wth h 1 = 180 and =40. We nvestgate the problem of tree-structure embedded densty estmaton, provdng a novel angle loong nto ths problem. Our method ntroduces metrcs learned from a spannng tree nto mode seeng. In partcular, we adopt mnmum spannng tree MST) to learn compact structures n the feature space or on a connected graph. On one hand, the ncluson of MST helps to fnd manfold structures for feature space analyss and data clusterng. On the other hand, the graph-based attrbute wors compatbly wth regonal level mage operatons n computer vson. A wde range of computer vson problems n prncple requres regonal support, where relaton between mage regons are typcally depcted wth a weghted graph and graph-based methods have consequently become a powerful tool. Such characterstc offers several ntutonally reasonable advantages. Frst, regon-wse operaton allows one to nvestgate and desgn more versatle and powerful features, as a regon often contans much more nformaton than a sngle pxel. Second, adoptng regon as basc processng unt can largely allevate the computatonal burden. In the paper, we only llustrate the applcatons of our method n data clusterng and regon-based mage segmentaton, due to the lmt of page length. Fgure 1 shows one example of data clusterng usng our proposed method. The potental applcaton of ths algorthm, however, s consderable, as mode seeng has dverse applcatons. Ths paper s organzed as follows: In Secton, we brefly ntroduce the bacground and closely related wors. 01

2 Readers already famlar wth nonparametrc densty estmaton and mean shft may jump to Secton 3, where we descrbe the proposed method and dscuss ts mportant propertes. Some expermental results regardng clusterng and applcaton of our method n mage segmentaton are llustrated n Secton 4, showng that the method s an effectve one. Fnally, conclusons are made n the last Secton.. Bacground and related wors Gven a set of ndependent and dentcally dstrbuted data ponts, nonparametrc densty estmaton sees to approxmate ts pdf. Instead of representng the pdf by a sngle parametrc model or a mxture model, the method fnds a small number of nearest or most smlar) tranng nstances and nterpolate from them. To obtan smooth pdf estmaton, gaussan ernel s commonly utlzed as the ernel densty estmator, also nown as Parzen wndow. The paradgm of densty estmaton and clusterng ncludes a famly of mode seeng algorthms wth Parzen densty estmaton. More recently, several wors have explored the mprovement of tradtonal mean shft algorthm. In [], the author ntroduced asymmetrc ernel to mean shft object tracng. The scale and orentaton of the ernel s automatcally and adaptvely selected, dependng on the observatons at each teraton. In [3], A new mode seeng algorthm called the medod shft was proposed. The purpose of medod shft s to extend mode seeng to general metrc spaces. The method, however, requres huge computatonal load and tends to result n over-fragmentaton. It essentally becomes a fnte pont searchng problem and s qute dfferent from our method n terms of both purpose and algorthmc process. In [4], the authors proposed the quc shft algorthm whch s consderably faster than mean shft and medod shft. Ther emphass tends to concentrate on algorthm acceleraton whle preservng ts performance. The GPU mplementaton of quc shft was dscussed n [5] to further speed up the algorthm from the hardware perspectve. There has also been other wors tryng to mprove the effcency of mode seeng [8]. Consderng the nearest neghbor property of MST, our method to some extent are related to prevous wors that generalze mean shft to non-lnear manfolds [9], or ntroduce nonlnear ernelzed or manfold metrcs [3, 4]. Our method can acheve some smlar goals but the dea remans very dfferent. We also notce there exst a great many wors concernng MST based graph segmentatons [10]. Even though our method have also utlzed MST, we generally thn t belongs to the famly of mode seeng methods where the algorthm characterstcs are qute dfferent from many graph based segmentaton methods. Hence these methods may not fall wthn the scope of comparson n ths paper. In fact our wor presents a general framewor of embeddng tree structures nto the mode seeng process. Therefore t s straght forward for one to plug n many other trees and brng n addtonal algorthm characterstcs. 3. Graph-based densty estmaton We propose to perform densty estmaton on a jont doman represented by the node feature space and the dstance space defned by the mnmum spannng tree of that graph. There are several advantages operatng on an MST-based structure. Frst, tree-based structure helps to unquely defne dstances for any node par, as a tree does not have crcles. Of course, one could drectly defne the parwse node dstances n the Eucldean space, resultng n the tradtonal mean shft. But ths bascally dscards the structural nformaton preserved by a graph. In applcatons such as mage segmentaton, spatal nformaton preserved by a graph can be very mportant. Second, an MST s the connected graph structure where all nodes are connected wth least edges numbers and weghts. In other words, an MST can be regarded as a compact structure that preserves mportant nformaton about the cluster structure n a feature space. Although the ntroducton of a tree structure n practce could possbly be problematc - as t faces the rs of large tree structure varaton nduced by nose ponts, especally for those mportant tree roots - one shall see, the proposed method wors pretty well and robustly n real mage segmentaton tests. In addton, such formulaton helps to mprove mode seeng performancesfor many manfoldshaped clusters. There are several exstng methods extractng an MST. In ths paper, we adopt the Krusal s Algorthm to obtan the MST structure from the graph. We then defne the densty functon and descrbe ts mode seeng process n the followng part of ths secton Proposed densty estmator Gven N samples represented by the set V = {v = 1,...,N,v R d } and the undrected weghted graph G =V, E), the mnmum spannng tree S =V, E S ) s a connected graph of G wth E S E, E S = N 1. For any node par, j) where j, there exsts a unque path E j such that E j E S, and j s connected by E j and deletng any element of the set results n the dsconnecton of and j. In addton, we defne E j to be,f = j. Property 3.1 For any gven node par, j), the set of connectng edges E j s unque. The above attrbute comes drectly from the tree structure. The proof s smple: f there s more than one E j then there exsts at least one crcle, whch contradcts wth the proposton. The unque dstance defnton on an MST facltates the defnton of densty for a gven locaton. We propose to use a jont representaton of the MST dstance space or MST space for short) and the feature space 0

3 to defne the densty estmator. Consder the smplest case where the MST space ernel center s located exactly at a tree node v j, then the densty estmator can be wrtten as follows: dvj, v ) ) v v fv) =c 0, 1) 1 where dv j, v )= v 1,v ) E j v 1 v s the cumulatve weght of edges that connects the two nodes, v s the feature space ernel center, h 1 and are the bandwdth parameters controllng the wndow sze and c 0 s a constant term determned by the sample sze and bandwdth. x) s the profle of a normal ernel: x) =exp 1 x). To defne a densty estmator for any locaton on the MST space, we have to frst defne the branch of an MST node. Here by sayng any locaton we actually allow the MST space ernel center to be located on an MST edge between neghborng nodes. In other words, the ernel can shft on the constraned space defned by MST. Suppose v negh s a neghborng node of v, we have the followng defnton: Defnton 3.1 The branch of a gven tree node v wth respect to ts connected edge v, v negh ) s a set of nodes and edges B = V B, E B ), such that V B = {v j j, v, v negh ) E j }, E B = {v, v j ) j, v, v negh ) E j }. The branch of a node s an nduced subgraph rooted at v, and descendng from ts referenced connected edge. There exst at least one correspondng MST edge - denoted as e ref - where the MST space ernel center s located on. If the center s located exactly on a tree node, then one may choose any edge connectng ths node to one of ts neghborng nodes as e ref. Suppose that the two nodes connected by e ref are respectvely v ref1 and v ref, and that the dstances from the ernel center to v ref1 and v ref are respectvely x 1 and x x 1 +x = dv ref1 v ref )= v ref1 v ref ), then the densty estmator defned wth respect to v ref1 can be wrtten as: ˆf eref,vref1 v,x 1 )= dvref1, v ) x 1 ) c 0,v V ref1 c 0,v / V ref1 1 1 ) v v + dvref1, v )+x 1 ) ) v v. 3) where V ref1 s the set of branch nodes wth respect to v ref1 and e ref. Smlarly, we can defne the densty estmator wth respect to v ref : ˆf eref,vref v,x )= dvref, v ) x ) c 0,v V ref c 0,v / V ref 1 1 ) v v + dvref, v )+x ) ) v v. where V ref s defned n a smlar way. Assocated wth the above densty estmator are some good propertes that facltates the mode seeng process: Property 3. ˆf eref,vref1 = ˆf eref,vref, e ref E The above equalty holds n the sense that V ref1 V ref = V and V ref1 V ref =, whch ndcates {v v V ref1 } = {v v / V ref }. In addton, snce dv ref1, v ) x 1 = dv ref1, v ref )+dv ref, v ) x 1 = dv ref, v )+x when v V ref1, we obtan the followng equalty:,v V ref1 =,v / V ref 1 dvref1, v ) x 1 ) ) v v 1 4) dvref, v )+x ) ) v v. The equalty relaton between the second term of 3) and the frst term of 4) can be proved smlarly. Property 3. states that the estmated densty does not depend on the choce of reference pont. Property 3.3 If e ref1 and e ref are two edges that connects the same node v ref, ˆferef1,vref v, 0) = ˆf eref,vref v, 0), v ref V. Property 3.3 states that the estmated densty does not depend on the choce of reference edge when the MST space ernel s located on a tree node. Here we consder the specal stuaton where the MST space ernel s shftng from one edge to another. When the ernel s located on v ref,the densty estmator degenerates to 1), as x =0. The same condton also holds when we defne the densty estmator wth respect to any other edge connectng to v ref,whch ndcates the above property. Property 3.4 The ernel defned on the MST dstance space s contnuous and s pecewse dfferentable. 03

4 Accordng to the defnton of densty estmator, one s easy to verfy the pecewse contnuty and dfferentablty gven the MST space ernel s located on the same edge. Together wth Property 3.3, we can obtan Property 3.4. The above property also nfers the contnuty and pecewse dfferentablty of the densty estmator snce t s a lnear combnaton of contnuous and pecewse dfferentable ernels. 3.. Mode seeng wth force competton We see the mode by maxmzng the densty estmator wth respect to v and x smultaneously. The step s to pecewsely estmate the densty gradent, whch s smlar to mean shft. Tang the dervatve of the densty estmator wth respect to v, one get the estmated densty gradent: where K jont, s the product of the feature space ernel and the negatve dervatve of the MST space ernel profle: dv ref,v ) x) ) h v v K jont, = h 1 f v V ref dv ref,v )+x) ) h v v 1 otherwse Equaton 7) can be further rewrtten as: ˆf eref,vref v,x) = [ x c 0 ][ K jont, K jont, dv ref, v ) 1 8),v V ref ) K jont, dv ref, v ) / ] K jont, x,v / V ref ˆf eref,vref v,x) The last term of 8) results n the dsplacement of the MST space ernel, whch s the so called force competton. Force v = c 0 v v competton can also be regarded as a specal case of unvarate mean shft wth v v v)k g h ref representng the orgn. One = c [ 0 v v ] [ K g K g could magne t as a tug of war where data ponts weghted v v ] by K jont are tuggng along each sde of v ref. The shftng step sze, however, should be chosen carefully snce h v h K g v v v ˆf eref,vref s only pecewse dfferentable. Suppose we use 5) the ms to denote the last term of 8), the dsplacement of the MST space ernel s defned as: where gx) = x), K s the MST space ernel functon: { dvref, v K = ) x) / 1) f v V ref1 dv ref, v )+x) / 1 ) otherwse The second term n 5) s the well nown mean shft vector for the feature space ernel center v: mv) = K g v v h v K g v v v. 6) [1] has already developed a sound theoretcal bass for mean shft algorthm concernng ts physcal meanng, convergence analyss and relaton to other feature space analyss methods. Here we wll not extend the dscusson. Now consder the second varable. Tang the dervatve of ˆf eref,vref v,x) wth respect to x, wehave: ˆf eref,vref v,x) = x c 0 dv ref, v ) x)k jont, 1,v V ref + c 0 dv ref, v ) x)k jont,, 1,v / V ref 7) mx) =max x, mn e ref x, ms)) 9) The above term generantees that the MST space ernel s always shfted along the same reference edge. Here we see to provde more ntuton by dscussng some propertes of the densty gradent estmaton: Property 3.5 The estmaton of densty gradent does not depend on the choce of reference node v ref. Snce the densty estmator s pecewse dfferentable on the edge, accordng to Property 3. we can verfy the above property. The estmated densty gradent, however, does depend on the choce of reference edge when the MST space ernel reaches a tree node wth more than two connectng edges. Dfference n the choce of the reference edge results n the followng nequalty: V vref,eref1 V vref,eref V, where V vref,eref1 s the branch node set wth respect to node v ref and ts connectng edge e ref, and smlar for V vref,eref. Such nequalty leads to the sudden jump of estmated densty gradent at some tree nodes. Theorem 3.1 Gven any node v ref where the MST space ernel s located and there are more than two connectng edges, the number of reference edge e ref wth postve MST space ernel dsplacement s no more than 1. 04

5 Proof: Wthout loss of generalty, suppose the MST space ernel s located on node v ref wth three connectng edges e ref1, e ref and e ref3,andd eref1 >D eref >D eref3, where D eref s defned as follows: D eref = dvref, v ) ),v V vref,eref 1 v v dv ref, v ). The force competton term ms vref,eref equals to the estmated densty gradent wth respect to v ref and e ref tmes a postve scalar: ms vref,eref1 = c ˆf eref,vref v,x) x x=0 = D ref1 D ref D ref3. Smlarly, we have ms vref,eref = D ref D ref1 D ref3 and ms vref,eref3 = D ref3 D ref1 D ref. Snce D eref1 >D eref >D eref3 and D eref > 0, ms vref,eref and ms vref,eref3 can not possbly be larger than 0. The only postve ms vref,eref comes when D ref1 >D ref + D ref3 and the above proof can be easly extended to nodes wth multple edges. Thus we have proved the above Theorem Algorthmc descrpton Theorem 3.1 states that when the MST space ernel s located on any tree node, ether ths node s a local maxma, or there s only one edge to whch shftng the ernel results n the ncrease of the densty. The conveyed ntuton here s mportant: each tme the MST space ernel s shftng from one edge to another, one does not face the problem of multple selectable paths snce there s at most one edge that ncreases the estmated densty. Such property leads to the bass of our mplemented algorthm and ts fast approxmaton method. The mode seeng algorthm s a step sze controlled gradent ascent: 1. For each data pont v, =1,,..., N, ntalze the ts feature space ernel poston as the data pont tself. Select v as v ref and ntalze the MST space ernel on the reference node.. Compute the MST space ernel shft wth the followng rules: If the MST space s exactly located on any tree node, calculate m j x) x=0 wth respect to all ts connectng edges e j. If There exsts one postve m j, select the correspondng edge e j as the reference edge e ref. mx) =m j as the MST space ernel shft. Else mx) =0. Else calculate mx) wth respect to v ref and e ref. 3. Calculate the step control factor α: If mx) =0, α =1. Else α = mx) / ms. 4. Compute the feature space ernel shft and scale t wth α: m v) =αmv). 5. Smultaneously shft the MST space ernel and the feature space ernel wth respect to the ernel shfts calculated n Step and Step 4. The MST space ernel s shfted wth the followng rule: If the MST space ernel s exactly located on a node If mx) = e ref, shft the MST space ernel to the neghborng node connected by e ref and select the neghborng node as the new reference node. Elsef mx) =0, the MST space ernel stays on the current node. Else update the ernel poston on the edge: x = mx). Elsef the MST space ernel s located on an edge If mx) == x, shft the MST space ernel to the reference node. Elsef mx) =e ref x, shft the MST space ernel to the neghborng node connected by e ref and select the neghborng node as the new reference node. Else update the ernel poston on the edge: x = mx)+x. 6. Repeat Step to Step 5 untl convergence Fast approxmaton Due to the pecewse dfferentablty and step control, the above algorthm gves the best mode seeng performance but requres more teratons before convergence. In addton, the algorthm contans numerous f-then-else condtons, whch s not frendly to hardware mplementaton. Here we also propose a fast approxmaton to the orgnal algorthm by teratvely shftng the MST space ernel and the feature space ernel. The method s straght forward: 1. For each data pont, ntalze the MST space ernel and the feature space ernel.. Shft the feature space ernel accordng to 6). 05

6 3. If there exst neghborng nodes that ncrease the estmated densty, shft the MST space ernel to the nearest one. Otherwse, stop shftng. 4. Repeat Step and 3 untl convergence. In all of the followng experments, we only mplement the above fast algorthm. 4. Expermental results We show three sets of experments usng our proposed algorthm. The frst set of experments demonstrates the performance of the method n the tas of data clusterng. Fgure a) shows a character shaped dstrbuton contanng 934 data ponts and ts clusterng result. The bandwdth parameters h 1 and were respectvely set to 150 and 40 for ths experment. Fgure b) shows the mxture of 4 gaussan dstrbutons wth a total of 1500 data ponts. Here we set h 1 to 700 and to 150. From the two experments one could observe that the method wors reasonably well for both arbtrarly shaped and regularly shaped cluster of data. The real challenge comes when we want to cluster the spral-le data dstrbuton wth hghly nonlnear cluster separaton boundares. The example of spral-le data gven n [3] was reproduced wth the Matlab code ndly avalable at new medod.htm. In ths experment h 1 and are respectvely set to 150 and 300. Note that we have acheved the clusterng performance that approxmates the one gven n [3] wthout usng any non-eucldean metrc, whle mean shft or Eucldean medod shft usually wll fal on such tas Fgure 3. Clusterng wth spral-le cluster of data usng the proposed method The second set of experments address the problem dscontnuty preserved smoothng wth superpxelzed mages. As dscussed n prevous secton, regon-wse operaton sgnfcantly reduces the requred computaton power, thus greatly accelerates the mage smoothng and segmentaton process. The ntroducton of MST space ernel wors n compatble wth the regon adjacency graph and n addton, further mproves the smoothng and segmentaton performance. Fgure 4 shows the mages and ther smoothng results usng dfferent methods n the RGB color space. The mages are frst superpxelzed usng normalzed cut[6, 7]. The correspondng Matlab code s ndly provded at mor/research/superpxels/. We set the number of coarse superpxels N sp to 00, the number of fne superpxels N sp to 400 and the number of egenvectors N ev to 40. Each superpxel s then represented by the mean RGB value and the whole mage s mapped to an undrected, weghted regon adjacency graph where edges corresponds to the eght-connectvtes of two regons and edge weghts are defned as the Eucldean dstances between the regon means. We extract the mnmum spannng tree from the regon adjacency graph usng Krusal s Algorthm and perform mode seeng usng our proposed method. Here we fxed h 1 as 30 and as 50 for all the test mages. The obtaned results are llustrated n the second column of fgure 4. To demonstrate the mprovement of algorthm performance by ntroducng the MST space ernel, we compare the results wth medod shft smoothng where each super pxel s represented by the 5D jont representaton of the RGB mean and spatal coordnate mean. The dstance matrx s obtaned by calculatng the Eucldean dstances between each par of super pxels and the parameter Sgma s set to 000. We also compare our results wth quc shft whch s a fast mode seeng algorthm. We run the quc shft algorthm wth the VLFeat Matlab pacage whch s publcly avalable at The parameters rato, ernelsze and maxdst are respectvely set to 0.3, 1 and 30. The results llustrated n fgure 4 ndcates the advantage of usng our proposed method for mage smoothng. We llustrate the potental applcaton of mage segmentaton usng our method n the last set of experments. Note that the segmentaton performance depends largely on the defned feature. Wth superpxelzed mages, the defnton of mage feature becomes much more versatle than pxel based methods. Such framewor allows one to mprove the segmentaton performance by defnng the feature n a sophstcated way, usng textons, texture detectors or other regon statstcs. For smplcty we only adopt regon color hstogram n ths paper. Each regon s represented by a 4-D concatenated hstogram wth each RGB channel returnng a hstogram of 8 bns. We then use prncpal component analyss PCA) to perform dmensonalty reducton on the obtaned hstograms. The percentage of preserved varance for PCA s set to 0.9, a typcal rule of thumb value for PCA. For most of the mages, the reduced dmenson after performng PCA often les n between 4-8, whch s much smaller than the orgnal dmenson number. By runnng PCA we reduces the computatonal complexty and effec- 06

7 a) Fgure. Data clusterng usng the proposed method. a) Clusterng wth lnearly separable data. b) Clusterng wth mxture of gaussans b) Fgure 4. Dscontnuty preserved smoothng wth superpxelzed mages: The frst column contans the orgnal mages. The second column corresponds to the smoothng results usng the proposed method. The second column contans the smoothng results usng medod shft. The last column are the results obtaned by quc shft. tvely avods from sufferng the curse of dmensonalty. The segmentaton results are shown n fgure 5. One could observe that the proposed method s effectve and produces reasonably good segmentatons. 5. Concluson In ths paper, by ntroducng the MST space ernel, we have proposed a novel mode seeng method that can mprove mode seeng performance on manfold-structured data and can wor compatbly wth regon-wse mage pro- 07

8 Fgure 5. Image segmentaton experments wth regon hstogram cesng operatons. We acheved good algorthm performance n clusterng data wth hghly nonlnear separaton boundares wthout usng any manfold dstance or some other non Eucldean metrcs, whch s of consderable challenge. The advantage of usng the proposed method for mage smoothng and segmentaton s also supported by our experments. References [1] D. Comancu and P. Meer. Mean shft: A robust approach toward feature space analyss. IEEE Trans. Pattern Anal. Mach. Intell., 45): , 00. [] A. Ylmaz, Object tracng by Asymmetrc ernel mean shft wth automatc scale and orentaton selecton. In CVPR, 007. [3] Y. A. Sheh, E. A. Khan and T. Kanade. Modeseeng by Medodshfts. In ICCV, 007. [5] A. Vedald and S. Soatto. Really quc shft: Image segmentaton on a GPU. In Worshop on Computer Vson usng GPUs, held wth ECCV, 010. [6] J. Sh and J. Mal. Normalzed cuts and mage segmentaton. IEEE Trans. Pattern Anal. Mach. Intell., 8): , 000. [7] X. Ren and J. Mal. NLearnng a classfcaton model for segmentaton. In ICCV, 003. [8] K. Zhang, J. T. Kwo and M. Tang. Accelerated convergence usng dynamc mean shft. In ECCV, 006. [9] R. Subbarao and P. Meer. Nonlnear mean shft for clusterng over analytc manfolds. In CVPR, 006. [10] O. J. Morrs, M.de J. Lee, and A.G. Constantndes. Graph theory for mage analyss: An approach based on the shortest spannng tree, In IEE Proc. F., Communcatons. Radar & Sgnal Processng, 133:146-15, [4] A. Vedald and S. Soatto. Quc shft and ernel methods for mode seeng. In ECCV,

What is Candidate Sampling

What is Candidate Sampling What s Canddate Samplng Say we have a multclass or mult label problem where each tranng example ( x, T ) conssts of a context x a small (mult)set of target classes T out of a large unverse L of possble

More information

8.5 UNITARY AND HERMITIAN MATRICES. The conjugate transpose of a complex matrix A, denoted by A*, is given by

8.5 UNITARY AND HERMITIAN MATRICES. The conjugate transpose of a complex matrix A, denoted by A*, is given by 6 CHAPTER 8 COMPLEX VECTOR SPACES 5. Fnd the kernel of the lnear transformaton gven n Exercse 5. In Exercses 55 and 56, fnd the mage of v, for the ndcated composton, where and are gven by the followng

More information

v a 1 b 1 i, a 2 b 2 i,..., a n b n i.

v a 1 b 1 i, a 2 b 2 i,..., a n b n i. SECTION 8.4 COMPLEX VECTOR SPACES AND INNER PRODUCTS 455 8.4 COMPLEX VECTOR SPACES AND INNER PRODUCTS All the vector spaces we have studed thus far n the text are real vector spaces snce the scalars are

More information

The Development of Web Log Mining Based on Improve-K-Means Clustering Analysis

The Development of Web Log Mining Based on Improve-K-Means Clustering Analysis The Development of Web Log Mnng Based on Improve-K-Means Clusterng Analyss TngZhong Wang * College of Informaton Technology, Luoyang Normal Unversty, Luoyang, 471022, Chna wangtngzhong2@sna.cn Abstract.

More information

Conversion between the vector and raster data structures using Fuzzy Geographical Entities

Conversion between the vector and raster data structures using Fuzzy Geographical Entities Converson between the vector and raster data structures usng Fuzzy Geographcal Enttes Cdála Fonte Department of Mathematcs Faculty of Scences and Technology Unversty of Combra, Apartado 38, 3 454 Combra,

More information

Luby s Alg. for Maximal Independent Sets using Pairwise Independence

Luby s Alg. for Maximal Independent Sets using Pairwise Independence Lecture Notes for Randomzed Algorthms Luby s Alg. for Maxmal Independent Sets usng Parwse Independence Last Updated by Erc Vgoda on February, 006 8. Maxmal Independent Sets For a graph G = (V, E), an ndependent

More information

Face Verification Problem. Face Recognition Problem. Application: Access Control. Biometric Authentication. Face Verification (1:1 matching)

Face Verification Problem. Face Recognition Problem. Application: Access Control. Biometric Authentication. Face Verification (1:1 matching) Face Recognton Problem Face Verfcaton Problem Face Verfcaton (1:1 matchng) Querymage face query Face Recognton (1:N matchng) database Applcaton: Access Control www.vsage.com www.vsoncs.com Bometrc Authentcaton

More information

Recurrence. 1 Definitions and main statements

Recurrence. 1 Definitions and main statements Recurrence 1 Defntons and man statements Let X n, n = 0, 1, 2,... be a MC wth the state space S = (1, 2,...), transton probabltes p j = P {X n+1 = j X n = }, and the transton matrx P = (p j ),j S def.

More information

Module 2 LOSSLESS IMAGE COMPRESSION SYSTEMS. Version 2 ECE IIT, Kharagpur

Module 2 LOSSLESS IMAGE COMPRESSION SYSTEMS. Version 2 ECE IIT, Kharagpur Module LOSSLESS IMAGE COMPRESSION SYSTEMS Lesson 3 Lossless Compresson: Huffman Codng Instructonal Objectves At the end of ths lesson, the students should be able to:. Defne and measure source entropy..

More information

Vision Mouse. Saurabh Sarkar a* University of Cincinnati, Cincinnati, USA ABSTRACT 1. INTRODUCTION

Vision Mouse. Saurabh Sarkar a* University of Cincinnati, Cincinnati, USA ABSTRACT 1. INTRODUCTION Vson Mouse Saurabh Sarkar a* a Unversty of Cncnnat, Cncnnat, USA ABSTRACT The report dscusses a vson based approach towards trackng of eyes and fngers. The report descrbes the process of locatng the possble

More information

How To Calculate The Accountng Perod Of Nequalty

How To Calculate The Accountng Perod Of Nequalty Inequalty and The Accountng Perod Quentn Wodon and Shlomo Ytzha World Ban and Hebrew Unversty September Abstract Income nequalty typcally declnes wth the length of tme taen nto account for measurement.

More information

benefit is 2, paid if the policyholder dies within the year, and probability of death within the year is ).

benefit is 2, paid if the policyholder dies within the year, and probability of death within the year is ). REVIEW OF RISK MANAGEMENT CONCEPTS LOSS DISTRIBUTIONS AND INSURANCE Loss and nsurance: When someone s subject to the rsk of ncurrng a fnancal loss, the loss s generally modeled usng a random varable or

More information

L10: Linear discriminants analysis

L10: Linear discriminants analysis L0: Lnear dscrmnants analyss Lnear dscrmnant analyss, two classes Lnear dscrmnant analyss, C classes LDA vs. PCA Lmtatons of LDA Varants of LDA Other dmensonalty reducton methods CSCE 666 Pattern Analyss

More information

A machine vision approach for detecting and inspecting circular parts

A machine vision approach for detecting and inspecting circular parts A machne vson approach for detectng and nspectng crcular parts Du-Mng Tsa Machne Vson Lab. Department of Industral Engneerng and Management Yuan-Ze Unversty, Chung-L, Tawan, R.O.C. E-mal: edmtsa@saturn.yzu.edu.tw

More information

Descriptive Models. Cluster Analysis. Example. General Applications of Clustering. Examples of Clustering Applications

Descriptive Models. Cluster Analysis. Example. General Applications of Clustering. Examples of Clustering Applications CMSC828G Prncples of Data Mnng Lecture #9 Today s Readng: HMS, chapter 9 Today s Lecture: Descrptve Modelng Clusterng Algorthms Descrptve Models model presents the man features of the data, a global summary

More information

An Interest-Oriented Network Evolution Mechanism for Online Communities

An Interest-Oriented Network Evolution Mechanism for Online Communities An Interest-Orented Network Evoluton Mechansm for Onlne Communtes Cahong Sun and Xaopng Yang School of Informaton, Renmn Unversty of Chna, Bejng 100872, P.R. Chna {chsun,yang}@ruc.edu.cn Abstract. Onlne

More information

1 Example 1: Axis-aligned rectangles

1 Example 1: Axis-aligned rectangles COS 511: Theoretcal Machne Learnng Lecturer: Rob Schapre Lecture # 6 Scrbe: Aaron Schld February 21, 2013 Last class, we dscussed an analogue for Occam s Razor for nfnte hypothess spaces that, n conjuncton

More information

A Novel Methodology of Working Capital Management for Large. Public Constructions by Using Fuzzy S-curve Regression

A Novel Methodology of Working Capital Management for Large. Public Constructions by Using Fuzzy S-curve Regression Novel Methodology of Workng Captal Management for Large Publc Constructons by Usng Fuzzy S-curve Regresson Cheng-Wu Chen, Morrs H. L. Wang and Tng-Ya Hseh Department of Cvl Engneerng, Natonal Central Unversty,

More information

Single and multiple stage classifiers implementing logistic discrimination

Single and multiple stage classifiers implementing logistic discrimination Sngle and multple stage classfers mplementng logstc dscrmnaton Hélo Radke Bttencourt 1 Dens Alter de Olvera Moraes 2 Vctor Haertel 2 1 Pontfíca Unversdade Católca do Ro Grande do Sul - PUCRS Av. Ipranga,

More information

An Enhanced Super-Resolution System with Improved Image Registration, Automatic Image Selection, and Image Enhancement

An Enhanced Super-Resolution System with Improved Image Registration, Automatic Image Selection, and Image Enhancement An Enhanced Super-Resoluton System wth Improved Image Regstraton, Automatc Image Selecton, and Image Enhancement Yu-Chuan Kuo ( ), Chen-Yu Chen ( ), and Chou-Shann Fuh ( ) Department of Computer Scence

More information

Detecting Global Motion Patterns in Complex Videos

Detecting Global Motion Patterns in Complex Videos Detectng Global Moton Patterns n Complex Vdeos Mn Hu, Saad Al, Mubarak Shah Computer Vson Lab, Unversty of Central Florda {mhu,sal,shah}@eecs.ucf.edu Abstract Learnng domnant moton patterns or actvtes

More information

Linear Circuits Analysis. Superposition, Thevenin /Norton Equivalent circuits

Linear Circuits Analysis. Superposition, Thevenin /Norton Equivalent circuits Lnear Crcuts Analyss. Superposton, Theenn /Norton Equalent crcuts So far we hae explored tmendependent (resste) elements that are also lnear. A tmendependent elements s one for whch we can plot an / cure.

More information

Support Vector Machines

Support Vector Machines Support Vector Machnes Max Wellng Department of Computer Scence Unversty of Toronto 10 Kng s College Road Toronto, M5S 3G5 Canada wellng@cs.toronto.edu Abstract Ths s a note to explan support vector machnes.

More information

Can Auto Liability Insurance Purchases Signal Risk Attitude?

Can Auto Liability Insurance Purchases Signal Risk Attitude? Internatonal Journal of Busness and Economcs, 2011, Vol. 10, No. 2, 159-164 Can Auto Lablty Insurance Purchases Sgnal Rsk Atttude? Chu-Shu L Department of Internatonal Busness, Asa Unversty, Tawan Sheng-Chang

More information

A hybrid global optimization algorithm based on parallel chaos optimization and outlook algorithm

A hybrid global optimization algorithm based on parallel chaos optimization and outlook algorithm Avalable onlne www.ocpr.com Journal of Chemcal and Pharmaceutcal Research, 2014, 6(7):1884-1889 Research Artcle ISSN : 0975-7384 CODEN(USA) : JCPRC5 A hybrd global optmzaton algorthm based on parallel

More information

The OC Curve of Attribute Acceptance Plans

The OC Curve of Attribute Acceptance Plans The OC Curve of Attrbute Acceptance Plans The Operatng Characterstc (OC) curve descrbes the probablty of acceptng a lot as a functon of the lot s qualty. Fgure 1 shows a typcal OC Curve. 10 8 6 4 1 3 4

More information

An Alternative Way to Measure Private Equity Performance

An Alternative Way to Measure Private Equity Performance An Alternatve Way to Measure Prvate Equty Performance Peter Todd Parlux Investment Technology LLC Summary Internal Rate of Return (IRR) s probably the most common way to measure the performance of prvate

More information

Feature selection for intrusion detection. Slobodan Petrović NISlab, Gjøvik University College

Feature selection for intrusion detection. Slobodan Petrović NISlab, Gjøvik University College Feature selecton for ntruson detecton Slobodan Petrovć NISlab, Gjøvk Unversty College Contents The feature selecton problem Intruson detecton Traffc features relevant for IDS The CFS measure The mrmr measure

More information

Ring structure of splines on triangulations

Ring structure of splines on triangulations www.oeaw.ac.at Rng structure of splnes on trangulatons N. Vllamzar RICAM-Report 2014-48 www.rcam.oeaw.ac.at RING STRUCTURE OF SPLINES ON TRIANGULATIONS NELLY VILLAMIZAR Introducton For a trangulated regon

More information

THE DISTRIBUTION OF LOAN PORTFOLIO VALUE * Oldrich Alfons Vasicek

THE DISTRIBUTION OF LOAN PORTFOLIO VALUE * Oldrich Alfons Vasicek HE DISRIBUION OF LOAN PORFOLIO VALUE * Oldrch Alfons Vascek he amount of captal necessary to support a portfolo of debt securtes depends on the probablty dstrbuton of the portfolo loss. Consder a portfolo

More information

Forecasting the Direction and Strength of Stock Market Movement

Forecasting the Direction and Strength of Stock Market Movement Forecastng the Drecton and Strength of Stock Market Movement Jngwe Chen Mng Chen Nan Ye cjngwe@stanford.edu mchen5@stanford.edu nanye@stanford.edu Abstract - Stock market s one of the most complcated systems

More information

Properties of Indoor Received Signal Strength for WLAN Location Fingerprinting

Properties of Indoor Received Signal Strength for WLAN Location Fingerprinting Propertes of Indoor Receved Sgnal Strength for WLAN Locaton Fngerprntng Kamol Kaemarungs and Prashant Krshnamurthy Telecommuncatons Program, School of Informaton Scences, Unversty of Pttsburgh E-mal: kakst2,prashk@ptt.edu

More information

CS 2750 Machine Learning. Lecture 3. Density estimation. CS 2750 Machine Learning. Announcements

CS 2750 Machine Learning. Lecture 3. Density estimation. CS 2750 Machine Learning. Announcements Lecture 3 Densty estmaton Mlos Hauskrecht mlos@cs.ptt.edu 5329 Sennott Square Next lecture: Matlab tutoral Announcements Rules for attendng the class: Regstered for credt Regstered for audt (only f there

More information

+ + + - - This circuit than can be reduced to a planar circuit

+ + + - - This circuit than can be reduced to a planar circuit MeshCurrent Method The meshcurrent s analog of the nodeoltage method. We sole for a new set of arables, mesh currents, that automatcally satsfy KCLs. As such, meshcurrent method reduces crcut soluton to

More information

The Greedy Method. Introduction. 0/1 Knapsack Problem

The Greedy Method. Introduction. 0/1 Knapsack Problem The Greedy Method Introducton We have completed data structures. We now are gong to look at algorthm desgn methods. Often we are lookng at optmzaton problems whose performance s exponental. For an optmzaton

More information

J. Parallel Distrib. Comput.

J. Parallel Distrib. Comput. J. Parallel Dstrb. Comput. 71 (2011) 62 76 Contents lsts avalable at ScenceDrect J. Parallel Dstrb. Comput. journal homepage: www.elsever.com/locate/jpdc Optmzng server placement n dstrbuted systems n

More information

Calculation of Sampling Weights

Calculation of Sampling Weights Perre Foy Statstcs Canada 4 Calculaton of Samplng Weghts 4.1 OVERVIEW The basc sample desgn used n TIMSS Populatons 1 and 2 was a two-stage stratfed cluster desgn. 1 The frst stage conssted of a sample

More information

DEFINING %COMPLETE IN MICROSOFT PROJECT

DEFINING %COMPLETE IN MICROSOFT PROJECT CelersSystems DEFINING %COMPLETE IN MICROSOFT PROJECT PREPARED BY James E Aksel, PMP, PMI-SP, MVP For Addtonal Informaton about Earned Value Management Systems and reportng, please contact: CelersSystems,

More information

PERRON FROBENIUS THEOREM

PERRON FROBENIUS THEOREM PERRON FROBENIUS THEOREM R. CLARK ROBINSON Defnton. A n n matrx M wth real entres m, s called a stochastc matrx provded () all the entres m satsfy 0 m, () each of the columns sum to one, m = for all, ()

More information

A Simple Approach to Clustering in Excel

A Simple Approach to Clustering in Excel A Smple Approach to Clusterng n Excel Aravnd H Center for Computatonal Engneerng and Networng Amrta Vshwa Vdyapeetham, Combatore, Inda C Rajgopal Center for Computatonal Engneerng and Networng Amrta Vshwa

More information

A Fast Incremental Spectral Clustering for Large Data Sets

A Fast Incremental Spectral Clustering for Large Data Sets 2011 12th Internatonal Conference on Parallel and Dstrbuted Computng, Applcatons and Technologes A Fast Incremental Spectral Clusterng for Large Data Sets Tengteng Kong 1,YeTan 1, Hong Shen 1,2 1 School

More information

Out-of-Sample Extensions for LLE, Isomap, MDS, Eigenmaps, and Spectral Clustering

Out-of-Sample Extensions for LLE, Isomap, MDS, Eigenmaps, and Spectral Clustering Out-of-Sample Extensons for LLE, Isomap, MDS, Egenmaps, and Spectral Clusterng Yoshua Bengo, Jean-Franços Paement, Pascal Vncent Olver Delalleau, Ncolas Le Roux and Mare Oumet Département d Informatque

More information

where the coordinates are related to those in the old frame as follows.

where the coordinates are related to those in the old frame as follows. Chapter 2 - Cartesan Vectors and Tensors: Ther Algebra Defnton of a vector Examples of vectors Scalar multplcaton Addton of vectors coplanar vectors Unt vectors A bass of non-coplanar vectors Scalar product

More information

8 Algorithm for Binary Searching in Trees

8 Algorithm for Binary Searching in Trees 8 Algorthm for Bnary Searchng n Trees In ths secton we present our algorthm for bnary searchng n trees. A crucal observaton employed by the algorthm s that ths problem can be effcently solved when the

More information

BERNSTEIN POLYNOMIALS

BERNSTEIN POLYNOMIALS On-Lne Geometrc Modelng Notes BERNSTEIN POLYNOMIALS Kenneth I. Joy Vsualzaton and Graphcs Research Group Department of Computer Scence Unversty of Calforna, Davs Overvew Polynomals are ncredbly useful

More information

Logistic Regression. Lecture 4: More classifiers and classes. Logistic regression. Adaboost. Optimization. Multiple class classification

Logistic Regression. Lecture 4: More classifiers and classes. Logistic regression. Adaboost. Optimization. Multiple class classification Lecture 4: More classfers and classes C4B Machne Learnng Hlary 20 A. Zsserman Logstc regresson Loss functons revsted Adaboost Loss functons revsted Optmzaton Multple class classfcaton Logstc Regresson

More information

Forecasting the Demand of Emergency Supplies: Based on the CBR Theory and BP Neural Network

Forecasting the Demand of Emergency Supplies: Based on the CBR Theory and BP Neural Network 700 Proceedngs of the 8th Internatonal Conference on Innovaton & Management Forecastng the Demand of Emergency Supples: Based on the CBR Theory and BP Neural Network Fu Deqang, Lu Yun, L Changbng School

More information

NPAR TESTS. One-Sample Chi-Square Test. Cell Specification. Observed Frequencies 1O i 6. Expected Frequencies 1EXP i 6

NPAR TESTS. One-Sample Chi-Square Test. Cell Specification. Observed Frequencies 1O i 6. Expected Frequencies 1EXP i 6 PAR TESTS If a WEIGHT varable s specfed, t s used to replcate a case as many tmes as ndcated by the weght value rounded to the nearest nteger. If the workspace requrements are exceeded and samplng has

More information

Traffic State Estimation in the Traffic Management Center of Berlin

Traffic State Estimation in the Traffic Management Center of Berlin Traffc State Estmaton n the Traffc Management Center of Berln Authors: Peter Vortsch, PTV AG, Stumpfstrasse, D-763 Karlsruhe, Germany phone ++49/72/965/35, emal peter.vortsch@ptv.de Peter Möhl, PTV AG,

More information

Logical Development Of Vogel s Approximation Method (LD-VAM): An Approach To Find Basic Feasible Solution Of Transportation Problem

Logical Development Of Vogel s Approximation Method (LD-VAM): An Approach To Find Basic Feasible Solution Of Transportation Problem INTERNATIONAL JOURNAL OF SCIENTIFIC & TECHNOLOGY RESEARCH VOLUME, ISSUE, FEBRUARY ISSN 77-866 Logcal Development Of Vogel s Approxmaton Method (LD- An Approach To Fnd Basc Feasble Soluton Of Transportaton

More information

How Sets of Coherent Probabilities May Serve as Models for Degrees of Incoherence

How Sets of Coherent Probabilities May Serve as Models for Degrees of Incoherence 1 st Internatonal Symposum on Imprecse Probabltes and Ther Applcatons, Ghent, Belgum, 29 June 2 July 1999 How Sets of Coherent Probabltes May Serve as Models for Degrees of Incoherence Mar J. Schervsh

More information

"Research Note" APPLICATION OF CHARGE SIMULATION METHOD TO ELECTRIC FIELD CALCULATION IN THE POWER CABLES *

Research Note APPLICATION OF CHARGE SIMULATION METHOD TO ELECTRIC FIELD CALCULATION IN THE POWER CABLES * Iranan Journal of Scence & Technology, Transacton B, Engneerng, ol. 30, No. B6, 789-794 rnted n The Islamc Republc of Iran, 006 Shraz Unversty "Research Note" ALICATION OF CHARGE SIMULATION METHOD TO ELECTRIC

More information

Risk-based Fatigue Estimate of Deep Water Risers -- Course Project for EM388F: Fracture Mechanics, Spring 2008

Risk-based Fatigue Estimate of Deep Water Risers -- Course Project for EM388F: Fracture Mechanics, Spring 2008 Rsk-based Fatgue Estmate of Deep Water Rsers -- Course Project for EM388F: Fracture Mechancs, Sprng 2008 Chen Sh Department of Cvl, Archtectural, and Envronmental Engneerng The Unversty of Texas at Austn

More information

SPEE Recommended Evaluation Practice #6 Definition of Decline Curve Parameters Background:

SPEE Recommended Evaluation Practice #6 Definition of Decline Curve Parameters Background: SPEE Recommended Evaluaton Practce #6 efnton of eclne Curve Parameters Background: The producton hstores of ol and gas wells can be analyzed to estmate reserves and future ol and gas producton rates and

More information

Efficient Project Portfolio as a tool for Enterprise Risk Management

Efficient Project Portfolio as a tool for Enterprise Risk Management Effcent Proect Portfolo as a tool for Enterprse Rsk Management Valentn O. Nkonov Ural State Techncal Unversty Growth Traectory Consultng Company January 5, 27 Effcent Proect Portfolo as a tool for Enterprse

More information

An Algorithm for Data-Driven Bandwidth Selection

An Algorithm for Data-Driven Bandwidth Selection IEEE TRANSACTIONS ON PATTERN ANALYSIS AND MACHINE INTELLIGENCE, VOL. 25, NO. 2, FEBRUARY 2003 An Algorthm for Data-Drven Bandwdth Selecton Dorn Comancu, Member, IEEE Abstract The analyss of a feature space

More information

A Secure Password-Authenticated Key Agreement Using Smart Cards

A Secure Password-Authenticated Key Agreement Using Smart Cards A Secure Password-Authentcated Key Agreement Usng Smart Cards Ka Chan 1, Wen-Chung Kuo 2 and Jn-Chou Cheng 3 1 Department of Computer and Informaton Scence, R.O.C. Mltary Academy, Kaohsung 83059, Tawan,

More information

A Performance Analysis of View Maintenance Techniques for Data Warehouses

A Performance Analysis of View Maintenance Techniques for Data Warehouses A Performance Analyss of Vew Mantenance Technques for Data Warehouses Xng Wang Dell Computer Corporaton Round Roc, Texas Le Gruenwald The nversty of Olahoma School of Computer Scence orman, OK 739 Guangtao

More information

Project Networks With Mixed-Time Constraints

Project Networks With Mixed-Time Constraints Project Networs Wth Mxed-Tme Constrants L Caccetta and B Wattananon Western Australan Centre of Excellence n Industral Optmsaton (WACEIO) Curtn Unversty of Technology GPO Box U1987 Perth Western Australa

More information

A Multi-mode Image Tracking System Based on Distributed Fusion

A Multi-mode Image Tracking System Based on Distributed Fusion A Mult-mode Image Tracng System Based on Dstrbuted Fuson Ln zheng Chongzhao Han Dongguang Zuo Hongsen Yan School of Electroncs & nformaton engneerng, X an Jaotong Unversty X an, Shaanx, Chna Lnzheng@malst.xjtu.edu.cn

More information

Calculating the high frequency transmission line parameters of power cables

Calculating the high frequency transmission line parameters of power cables < ' Calculatng the hgh frequency transmsson lne parameters of power cables Authors: Dr. John Dcknson, Laboratory Servces Manager, N 0 RW E B Communcatons Mr. Peter J. Ncholson, Project Assgnment Manager,

More information

RESEARCH ON DUAL-SHAKER SINE VIBRATION CONTROL. Yaoqi FENG 1, Hanping QIU 1. China Academy of Space Technology (CAST) yaoqi.feng@yahoo.

RESEARCH ON DUAL-SHAKER SINE VIBRATION CONTROL. Yaoqi FENG 1, Hanping QIU 1. China Academy of Space Technology (CAST) yaoqi.feng@yahoo. ICSV4 Carns Australa 9- July, 007 RESEARCH ON DUAL-SHAKER SINE VIBRATION CONTROL Yaoq FENG, Hanpng QIU Dynamc Test Laboratory, BISEE Chna Academy of Space Technology (CAST) yaoq.feng@yahoo.com Abstract

More information

Graph Calculus: Scalable Shortest Path Analytics for Large Social Graphs through Core Net

Graph Calculus: Scalable Shortest Path Analytics for Large Social Graphs through Core Net Graph Calculus: Scalable Shortest Path Analytcs for Large Socal Graphs through Core Net Lxn Fu Department of Computer Scence Unversty of North Carolna at Greensboro Greensboro, NC, U.S.A. lfu@uncg.edu

More information

Eye Center Localization on a Facial Image Based on Multi-Block Local Binary Patterns

Eye Center Localization on a Facial Image Based on Multi-Block Local Binary Patterns Eye Center Localzaton on a Facal Image Based on Mult-Bloc Local Bnary Patterns Anatoly tn, Vladmr Khryashchev, Olga Stepanova Yaroslavl State Unversty Yaroslavl, Russa anatolyntnyar@gmal.com, vhr@yandex.ru,

More information

Period and Deadline Selection for Schedulability in Real-Time Systems

Period and Deadline Selection for Schedulability in Real-Time Systems Perod and Deadlne Selecton for Schedulablty n Real-Tme Systems Thdapat Chantem, Xaofeng Wang, M.D. Lemmon, and X. Sharon Hu Department of Computer Scence and Engneerng, Department of Electrcal Engneerng

More information

On the Optimal Control of a Cascade of Hydro-Electric Power Stations

On the Optimal Control of a Cascade of Hydro-Electric Power Stations On the Optmal Control of a Cascade of Hydro-Electrc Power Statons M.C.M. Guedes a, A.F. Rbero a, G.V. Smrnov b and S. Vlela c a Department of Mathematcs, School of Scences, Unversty of Porto, Portugal;

More information

An Analysis of Central Processor Scheduling in Multiprogrammed Computer Systems

An Analysis of Central Processor Scheduling in Multiprogrammed Computer Systems STAN-CS-73-355 I SU-SE-73-013 An Analyss of Central Processor Schedulng n Multprogrammed Computer Systems (Dgest Edton) by Thomas G. Prce October 1972 Techncal Report No. 57 Reproducton n whole or n part

More information

INVESTIGATION OF VEHICULAR USERS FAIRNESS IN CDMA-HDR NETWORKS

INVESTIGATION OF VEHICULAR USERS FAIRNESS IN CDMA-HDR NETWORKS 21 22 September 2007, BULGARIA 119 Proceedngs of the Internatonal Conference on Informaton Technologes (InfoTech-2007) 21 st 22 nd September 2007, Bulgara vol. 2 INVESTIGATION OF VEHICULAR USERS FAIRNESS

More information

Lecture 2: Single Layer Perceptrons Kevin Swingler

Lecture 2: Single Layer Perceptrons Kevin Swingler Lecture 2: Sngle Layer Perceptrons Kevn Sngler kms@cs.str.ac.uk Recap: McCulloch-Ptts Neuron Ths vastly smplfed model of real neurons s also knon as a Threshold Logc Unt: W 2 A Y 3 n W n. A set of synapses

More information

Implementation of Deutsch's Algorithm Using Mathcad

Implementation of Deutsch's Algorithm Using Mathcad Implementaton of Deutsch's Algorthm Usng Mathcad Frank Roux The followng s a Mathcad mplementaton of Davd Deutsch's quantum computer prototype as presented on pages - n "Machnes, Logc and Quantum Physcs"

More information

A DATA MINING APPLICATION IN A STUDENT DATABASE

A DATA MINING APPLICATION IN A STUDENT DATABASE JOURNAL OF AERONAUTICS AND SPACE TECHNOLOGIES JULY 005 VOLUME NUMBER (53-57) A DATA MINING APPLICATION IN A STUDENT DATABASE Şenol Zafer ERDOĞAN Maltepe Ünversty Faculty of Engneerng Büyükbakkalköy-Istanbul

More information

Characterization of Assembly. Variation Analysis Methods. A Thesis. Presented to the. Department of Mechanical Engineering. Brigham Young University

Characterization of Assembly. Variation Analysis Methods. A Thesis. Presented to the. Department of Mechanical Engineering. Brigham Young University Characterzaton of Assembly Varaton Analyss Methods A Thess Presented to the Department of Mechancal Engneerng Brgham Young Unversty In Partal Fulfllment of the Requrements for the Degree Master of Scence

More information

Fisher Markets and Convex Programs

Fisher Markets and Convex Programs Fsher Markets and Convex Programs Nkhl R. Devanur 1 Introducton Convex programmng dualty s usually stated n ts most general form, wth convex objectve functons and convex constrants. (The book by Boyd and

More information

Vehicle Detection and Tracking in Video from Moving Airborne Platform

Vehicle Detection and Tracking in Video from Moving Airborne Platform Journal of Computatonal Informaton Systems 10: 12 (2014) 4965 4972 Avalable at http://www.jofcs.com Vehcle Detecton and Trackng n Vdeo from Movng Arborne Platform Lye ZHANG 1,2,, Hua WANG 3, L LI 2 1 School

More information

Institute of Informatics, Faculty of Business and Management, Brno University of Technology,Czech Republic

Institute of Informatics, Faculty of Business and Management, Brno University of Technology,Czech Republic Lagrange Multplers as Quanttatve Indcators n Economcs Ivan Mezník Insttute of Informatcs, Faculty of Busness and Management, Brno Unversty of TechnologCzech Republc Abstract The quanttatve role of Lagrange

More information

1. Fundamentals of probability theory 2. Emergence of communication traffic 3. Stochastic & Markovian Processes (SP & MP)

1. Fundamentals of probability theory 2. Emergence of communication traffic 3. Stochastic & Markovian Processes (SP & MP) 6.3 / -- Communcaton Networks II (Görg) SS20 -- www.comnets.un-bremen.de Communcaton Networks II Contents. Fundamentals of probablty theory 2. Emergence of communcaton traffc 3. Stochastc & Markovan Processes

More information

Human Tracking by Fast Mean Shift Mode Seeking

Human Tracking by Fast Mean Shift Mode Seeking JOURAL OF MULTIMEDIA, VOL. 1, O. 1, APRIL 2006 1 Human Trackng by Fast Mean Shft Mode Seekng [10 font sze blank 1] [10 font sze blank 2] C. Belezna Advanced Computer Vson GmbH - ACV, Venna, Austra Emal:

More information

When Network Effect Meets Congestion Effect: Leveraging Social Services for Wireless Services

When Network Effect Meets Congestion Effect: Leveraging Social Services for Wireless Services When Network Effect Meets Congeston Effect: Leveragng Socal Servces for Wreless Servces aowen Gong School of Electrcal, Computer and Energy Engeerng Arzona State Unversty Tempe, AZ 8587, USA xgong9@asuedu

More information

Realistic Image Synthesis

Realistic Image Synthesis Realstc Image Synthess - Combned Samplng and Path Tracng - Phlpp Slusallek Karol Myszkowsk Vncent Pegoraro Overvew: Today Combned Samplng (Multple Importance Samplng) Renderng and Measurng Equaton Random

More information

) of the Cell class is created containing information about events associated with the cell. Events are added to the Cell instance

) of the Cell class is created containing information about events associated with the cell. Events are added to the Cell instance Calbraton Method Instances of the Cell class (one nstance for each FMS cell) contan ADC raw data and methods assocated wth each partcular FMS cell. The calbraton method ncludes event selecton (Class Cell

More information

Gender Classification for Real-Time Audience Analysis System

Gender Classification for Real-Time Audience Analysis System Gender Classfcaton for Real-Tme Audence Analyss System Vladmr Khryashchev, Lev Shmaglt, Andrey Shemyakov, Anton Lebedev Yaroslavl State Unversty Yaroslavl, Russa vhr@yandex.ru, shmaglt_lev@yahoo.com, andrey.shemakov@gmal.com,

More information

Generalizing the degree sequence problem

Generalizing the degree sequence problem Mddlebury College March 2009 Arzona State Unversty Dscrete Mathematcs Semnar The degree sequence problem Problem: Gven an nteger sequence d = (d 1,...,d n ) determne f there exsts a graph G wth d as ts

More information

1. Measuring association using correlation and regression

1. Measuring association using correlation and regression How to measure assocaton I: Correlaton. 1. Measurng assocaton usng correlaton and regresson We often would lke to know how one varable, such as a mother's weght, s related to another varable, such as a

More information

SIMPLE LINEAR CORRELATION

SIMPLE LINEAR CORRELATION SIMPLE LINEAR CORRELATION Smple lnear correlaton s a measure of the degree to whch two varables vary together, or a measure of the ntensty of the assocaton between two varables. Correlaton often s abused.

More information

Causal, Explanatory Forecasting. Analysis. Regression Analysis. Simple Linear Regression. Which is Independent? Forecasting

Causal, Explanatory Forecasting. Analysis. Regression Analysis. Simple Linear Regression. Which is Independent? Forecasting Causal, Explanatory Forecastng Assumes cause-and-effect relatonshp between system nputs and ts output Forecastng wth Regresson Analyss Rchard S. Barr Inputs System Cause + Effect Relatonshp The job of

More information

APPLICATION OF PROBE DATA COLLECTED VIA INFRARED BEACONS TO TRAFFIC MANEGEMENT

APPLICATION OF PROBE DATA COLLECTED VIA INFRARED BEACONS TO TRAFFIC MANEGEMENT APPLICATION OF PROBE DATA COLLECTED VIA INFRARED BEACONS TO TRAFFIC MANEGEMENT Toshhko Oda (1), Kochro Iwaoka (2) (1), (2) Infrastructure Systems Busness Unt, Panasonc System Networks Co., Ltd. Saedo-cho

More information

A Suspect Vehicle Tracking System Based on Video

A Suspect Vehicle Tracking System Based on Video 3rd Internatonal Conference on Multmeda Technology ICMT 2013) A Suspect Vehcle Trackng System Based on Vdeo Yad Chen 1, Tuo Wang Abstract. Vdeo survellance systems are wdely used n securty feld. The large

More information

Bypassing Synthesis: PLS for Face Recognition with Pose, Low-Resolution and Sketch

Bypassing Synthesis: PLS for Face Recognition with Pose, Low-Resolution and Sketch Bypassng Synthess: PLS for Face Recognton wth Pose, Low-Resoluton and Setch Abhshe Sharma Insttute of Advanced Computer Scence Unversty of Maryland, USA bhoaal@umacs.umd.edu Davd W Jacobs Insttute of Advanced

More information

Hallucinating Multiple Occluded CCTV Face Images of Different Resolutions

Hallucinating Multiple Occluded CCTV Face Images of Different Resolutions In Proc. IEEE Internatonal Conference on Advanced Vdeo and Sgnal based Survellance (AVSS 05), September 2005 Hallucnatng Multple Occluded CCTV Face Images of Dfferent Resolutons Ku Ja Shaogang Gong Computer

More information

Data Broadcast on a Multi-System Heterogeneous Overlayed Wireless Network *

Data Broadcast on a Multi-System Heterogeneous Overlayed Wireless Network * JOURNAL OF INFORMATION SCIENCE AND ENGINEERING 24, 819-840 (2008) Data Broadcast on a Mult-System Heterogeneous Overlayed Wreless Network * Department of Computer Scence Natonal Chao Tung Unversty Hsnchu,

More information

An Evaluation of the Extended Logistic, Simple Logistic, and Gompertz Models for Forecasting Short Lifecycle Products and Services

An Evaluation of the Extended Logistic, Simple Logistic, and Gompertz Models for Forecasting Short Lifecycle Products and Services An Evaluaton of the Extended Logstc, Smple Logstc, and Gompertz Models for Forecastng Short Lfecycle Products and Servces Charles V. Trappey a,1, Hsn-yng Wu b a Professor (Management Scence), Natonal Chao

More information

Imperial College London

Imperial College London F. Fang 1, C.C. Pan 1, I.M. Navon 2, M.D. Pggott 1, G.J. Gorman 1, P.A. Allson 1 and A.J.H. Goddard 1 1 Appled Modellng and Computaton Group Department of Earth Scence and Engneerng Imperal College London,

More information

Automated information technology for ionosphere monitoring of low-orbit navigation satellite signals

Automated information technology for ionosphere monitoring of low-orbit navigation satellite signals Automated nformaton technology for onosphere montorng of low-orbt navgaton satellte sgnals Alexander Romanov, Sergey Trusov and Alexey Romanov Federal State Untary Enterprse Russan Insttute of Space Devce

More information

FREQUENCY OF OCCURRENCE OF CERTAIN CHEMICAL CLASSES OF GSR FROM VARIOUS AMMUNITION TYPES

FREQUENCY OF OCCURRENCE OF CERTAIN CHEMICAL CLASSES OF GSR FROM VARIOUS AMMUNITION TYPES FREQUENCY OF OCCURRENCE OF CERTAIN CHEMICAL CLASSES OF GSR FROM VARIOUS AMMUNITION TYPES Zuzanna BRO EK-MUCHA, Grzegorz ZADORA, 2 Insttute of Forensc Research, Cracow, Poland 2 Faculty of Chemstry, Jagellonan

More information

Traffic-light a stress test for life insurance provisions

Traffic-light a stress test for life insurance provisions MEMORANDUM Date 006-09-7 Authors Bengt von Bahr, Göran Ronge Traffc-lght a stress test for lfe nsurance provsons Fnansnspetonen P.O. Box 6750 SE-113 85 Stocholm [Sveavägen 167] Tel +46 8 787 80 00 Fax

More information

Bayesian Cluster Ensembles

Bayesian Cluster Ensembles Bayesan Cluster Ensembles Hongjun Wang 1, Hanhua Shan 2 and Arndam Banerjee 2 1 Informaton Research Insttute, Southwest Jaotong Unversty, Chengdu, Schuan, 610031, Chna 2 Department of Computer Scence &

More information

Document Clustering Analysis Based on Hybrid PSO+K-means Algorithm

Document Clustering Analysis Based on Hybrid PSO+K-means Algorithm Document Clusterng Analyss Based on Hybrd PSO+K-means Algorthm Xaohu Cu, Thomas E. Potok Appled Software Engneerng Research Group, Computatonal Scences and Engneerng Dvson, Oak Rdge Natonal Laboratory,

More information

Sketching Sampled Data Streams

Sketching Sampled Data Streams Sketchng Sampled Data Streams Florn Rusu, Aln Dobra CISE Department Unversty of Florda Ganesvlle, FL, USA frusu@cse.ufl.edu adobra@cse.ufl.edu Abstract Samplng s used as a unversal method to reduce the

More information

ANALYZING THE RELATIONSHIPS BETWEEN QUALITY, TIME, AND COST IN PROJECT MANAGEMENT DECISION MAKING

ANALYZING THE RELATIONSHIPS BETWEEN QUALITY, TIME, AND COST IN PROJECT MANAGEMENT DECISION MAKING ANALYZING THE RELATIONSHIPS BETWEEN QUALITY, TIME, AND COST IN PROJECT MANAGEMENT DECISION MAKING Matthew J. Lberatore, Department of Management and Operatons, Vllanova Unversty, Vllanova, PA 19085, 610-519-4390,

More information

Performance Analysis and Coding Strategy of ECOC SVMs

Performance Analysis and Coding Strategy of ECOC SVMs Internatonal Journal of Grd and Dstrbuted Computng Vol.7, No. (04), pp.67-76 http://dx.do.org/0.457/jgdc.04.7..07 Performance Analyss and Codng Strategy of ECOC SVMs Zhgang Yan, and Yuanxuan Yang, School

More information