nips nips2004 nips2004-161 knowledge-graph by maker-knowledge-mining

161 nips-2004-Self-Tuning Spectral Clustering

Source: pdf

Author: Lihi Zelnik-manor, Pietro Perona

Abstract: We study a number of open issues in spectral clustering: (i) Selecting the appropriate scale of analysis, (ii) Handling multi-scale data, (iii) Clustering with irregular background clutter, and, (iv) Finding automatically the number of groups. We ﬁrst propose that a ‘local’ scale should be used to compute the afﬁnity between each pair of points. This local scaling leads to better clustering especially when the data includes multiple scales and when the clusters are placed within a cluttered background. We further suggest exploiting the structure of the eigenvectors to infer automatically the number of groups. This leads to a new algorithm in which the ﬁnal randomly initialized k-means stage is eliminated. 1

Reference: text

Summary: the most important sentenses genereted by tfidf model

sentIndex sentText sentNum sentScore

1 html Abstract We study a number of open issues in spectral clustering: (i) Selecting the appropriate scale of analysis, (ii) Handling multi-scale data, (iii) Clustering with irregular background clutter, and, (iv) Finding automatically the number of groups. [sent-9, score-0.393]

2 This local scaling leads to better clustering especially when the data includes multiple scales and when the clusters are placed within a cluttered background. [sent-11, score-0.614]

3 We further suggest exploiting the structure of the eigenvectors to infer automatically the number of groups. [sent-12, score-0.359]

4 An alternative clustering approach, which was shown to handle such structured data is spectral clustering. [sent-18, score-0.41]

5 It does not require estimating an explicit model of data distribution, rather a spectral analysis of the matrix of point-to-point similarities. [sent-19, score-0.276]

6 There are still open issues: (i) Selection of the appropriate scale in which the data is to be analyzed, (ii) Clustering data that is distributed according to different scales, (iii) Clustering with irregular background clutter, and, (iv) Estimating automatically the number of groups. [sent-26, score-0.23]

7 , sn } in Rl cluster them into C clusters as follows: −d2 (s ,s ) i j 1. [sent-34, score-0.223]

8 Form the afﬁnity matrix A ∈ Rn×n deﬁned by Aij = exp ( ) for i = j σ2 and Aii = 0, where d(si , sj ) is some distance function, often just the Euclidean σ = 0. [sent-35, score-0.202]

9 03125 σ=1 Figure 1: Spectral clustering without local scaling (using the NJW algorithm. [sent-41, score-0.409]

10 ) Top row: When the data incorporates multiple scales standard spectral clustering fails. [sent-42, score-0.48]

11 This highlights the high impact σ has on the clustering quality. [sent-45, score-0.287]

12 Deﬁne D to be a diagonal matrix with Dii = malized afﬁnity matrix L = D−1/2 AD−1/2 . [sent-51, score-0.181]

13 , xC , the C largest eigenvectors of L, and form the matrix X = [x1 , . [sent-58, score-0.33]

14 Treat each row of Y as a point in RC and cluster via k-means. [sent-65, score-0.196]

15 Assign the original point si to cluster c if and only if the corresponding row i of the matrix Y was assigned to cluster c. [sent-67, score-0.508]

16 In Section 2 we analyze the effect of σ on the clustering and suggest a method for setting it automatically. [sent-68, score-0.284]

17 In Section 3 we suggest a scheme for ﬁnding automatically the number of groups C. [sent-70, score-0.365]

18 Our new spectral clustering algorithm is summarized in Section 4. [sent-71, score-0.41]

19 2 Local Scaling As was suggested by [6] the scaling parameter is some measure of when two points are considered similar. [sent-73, score-0.176]

20 [5] suggested selecting σ automatically by running their clustering algorithm repeatedly for a number of values of σ and selecting the one which provides least distorted clusters of the rows of Y . [sent-77, score-0.822]

21 Moreover, when the input data includes clusters with different local statistics there may not be a singe value of σ that works well for all the data. [sent-80, score-0.214]

22 When the data contains multiple scales, even using the optimal σ fails to provide good clustering (see examples at the right of top row). [sent-82, score-0.288]

23 The afﬁnities across clusters are larger than the afﬁnities within the background cluster. [sent-87, score-0.174]

24 Introducing Local Scaling: Instead of selecting a single scaling parameter σ we propose to calculate a local scaling parameter σi for each data point si . [sent-90, score-0.486]

25 The distance from si to sj as ‘seen’ by si is d(si , sj )/σi while the converse is d(sj , si )/σj . [sent-91, score-0.776]

26 The selection of the local scale σi can be done by studying the local statistics of the neighborhood of point si . [sent-93, score-0.379]

27 Figure 2 provides a visualization of the effect of the suggested local scaling. [sent-97, score-0.172]

28 Since the data resides in multiple scales (one cluster is tight and the other is sparse) the standard approach to estimating afﬁnities fails to capture the data structure (see Figure 2. [sent-98, score-0.256]

29 Local scaling automatically ﬁnds the two scales and results in high afﬁnities within clusters and low afﬁnities across clusters (see Figure 2. [sent-100, score-0.517]

30 We tested the power of local scaling by clustering the data set of Figure 1, plus four additional examples. [sent-103, score-0.409]

31 In spite of the multiple scales and the various types of structure, the groups now match the intuitive solution. [sent-108, score-0.238]

32 3 Estimating the Number of Clusters Having deﬁned a scheme to set the scale parameter automatically we are left with one more free parameter: the number of clusters. [sent-109, score-0.217]

33 This parameter is usually set manually and Figure 3: Our clustering results. [sent-110, score-0.247]

34 The ﬁrst 10 eigenvalues of L corresponding to the top row data sets of Figure 3. [sent-134, score-0.304]

35 The suggested scheme turns out to lead to a new spatial clustering algorithm. [sent-137, score-0.406]

36 1 The Intuitive Solution: Analyzing the Eigenvalues One possible approach to try and discover the number of groups is to analyze the eigenvalues of the afﬁnity matrix. [sent-139, score-0.323]

37 1) will be a repeated eigenvalue of magnitude 1 with multiplicity equal to the number of groups C. [sent-141, score-0.396]

38 Examining the eigenvalues of our locally scaled matrix, corresponding to clean data-sets, indeed shows that the multiplicity of eigenvalue 1 equals the number of groups. [sent-143, score-0.392]

39 An alternative approach would be to search for a drop in the magnitude of the eigenvalues (this was pursued to some extent by Polito and Perona in [7]). [sent-145, score-0.194]

40 The eigenvalues of L are the union of the eigenvalues of the sub-matrices corresponding to each cluster. [sent-147, score-0.353]

41 This implies the eigenvalues depend on the structure of the individual clusters and thus no assumptions can be placed on their values. [sent-148, score-0.29]

42 Figure 4 shows the ﬁrst 10 eigenvalues corresponding to the top row examples of Figure 3. [sent-150, score-0.304]

43 It highlights the different patterns of distribution of eigenvalues for different data sets. [sent-151, score-0.195]

44 , C), its eigenvalues and eigenvectors are the union of the eigenvalues and eigenvectors of its blocks padded appropriately with zeros (see [6, 5]). [sent-159, score-0.968]

45 However, as was shown above, the eigenvalue 1 is bound to be a repeated eigenvalue with multiplicity equal to the number of groups C. [sent-161, score-0.434]

46 This, however, implies that even if the eigensolver provided us the rotated set of vectors, ˆ ˆ we are still guaranteed that there exists a rotation R such that each row in the matrix X R has a single non-zero entry. [sent-164, score-0.38]

47 Since the eigenvectors of L are the union of the eigenvectors of its individual blocks (padded with zeros), taking more than the ﬁrst C eigenvectors will result in more than one non-zero entry in some of the rows. [sent-165, score-0.832]

48 Taking fewer eigenvectors we do not have a full basis spanning the subspace, thus depending on the initial X there might or might not exist such a rotation. [sent-166, score-0.228]

49 For each possible group number C we recover the rotation which best aligns X’s columns with the canonical coordinate system. [sent-169, score-0.536]

50 Let Z ∈ Rn×C be the matrix obtained after rotating the eigenvector matrix X, i. [sent-170, score-0.234]

51 We wish to recover the rotation R for which in every row in Z there will be at most one non-zero entry. [sent-173, score-0.329]

52 We thus deﬁne a cost function: n C J= i=1 j=1 2 Zij Mi2 (3) Minimizing this cost function over all possible rotations will provide the best alignment with the canonical coordinate system. [sent-174, score-0.526]

53 This is done using the gradient descent scheme described in Appendix A. [sent-175, score-0.186]

54 The number of groups is taken as the one providing the minimal cost (if several group numbers yield practically the same minimal cost, the largest of those is selected). [sent-176, score-0.52]

55 We start by aligning the top two eigenvectors (as well as possible). [sent-178, score-0.362]

56 Then, at each step of the search (up to the maximal group number), we add a single eigenvector to the already rotated ones. [sent-179, score-0.283]

57 This can be viewed as taking the alignment result of the previous group number as an initialization to the current one. [sent-180, score-0.322]

58 The alignment of this new set of eigenvectors is extremely fast (typically a few iterations) since the initialization is good. [sent-181, score-0.406]

59 The overall run time of this incremental procedure is just slightly longer than aligning all the eigenvectors in a non-incremental way. [sent-182, score-0.321]

60 Using this scheme to estimate the number of groups on the data set of Figure 3 provided a correct result for all but one (for the right-most dataset at the bottom row we predicted 2 clusters instead of 3). [sent-183, score-0.477]

61 Corresponding plots of the alignment quality for different group numbers are shown in Figure 5. [sent-184, score-0.325]

62 Yu and Shi [11] suggested rotating normalized eigenvectors to obtain an optimal segmentation. [sent-185, score-0.377]

63 , setting Mi = 1 and Zij = 0 otherwise) and using SVD to recover the rotation which best aligns the columns of X with those of Z. [sent-188, score-0.247]

64 In our experiments we noticed that this iterative method can easily get stuck in local minima and thus does not reliably ﬁnd the optimal alignment and the group number. [sent-189, score-0.364]

65 [3] who assigned points to clusters according to the maximal entry in the corresponding row of the eigenvector matrix. [sent-191, score-0.396]

66 This works well when there are no repeated eigenvalues as then the eigenvectors 0. [sent-192, score-0.422]

67 (3)) for varying group numbers corresponding to the top row data sets of Figure 3. [sent-210, score-0.293]

68 The selected group number marked by a red circle, corresponds to the largest group number providing minimal cost (costs up to 0. [sent-211, score-0.455]

69 used a non-normalized afﬁnity matrix thus were not certain to obtain a repeated eigenvalue, however, this could easily happen and then the clustering would fail. [sent-215, score-0.346]

70 (ii) Since the ﬁnal clustering can be conducted by non-maximum suppression, we obtain clustering results for all the inspected group numbers at a tiny additional cost. [sent-217, score-0.638]

71 When the data is highly noisy, one can still employ k-means, or better, EM, to cluster the rows of Z. [sent-218, score-0.187]

72 However, since the data is now aligned with the canonical coordinate scheme we can obtain by non-maximum suppression an excellent initialization so very few iterations sufﬁce. [sent-219, score-0.325]

73 Compute the local scale σi for each point si ∈ S using Eq. [sent-224, score-0.3]

74 Deﬁne D to be a diagonal matrix with Dii = j=1 Aij and construct the nor−1/2 ˆ −1/2 malized afﬁnity matrix L = D . [sent-230, score-0.181]

75 , xC the C largest eigenvectors of L and form the matrix X = [x1 , . [sent-235, score-0.33]

76 , xC ] ∈ Rn×C , where C is the largest possible group number. [sent-238, score-0.186]

77 Recover the rotation R which best aligns X’s columns with the canonical coordinate system using the incremental gradient descent scheme (see also Appendix A). [sent-240, score-0.518]

78 Grade the cost of the alignment for each group number, up to C, according to Eq. [sent-242, score-0.369]

79 Set the ﬁnal group number Cb est to be the largest group number with minimal alignment cost. [sent-245, score-0.56]

80 Take the alignment result Z of the top Cb est eigenvectors and assign the original 2 2 point si to cluster c if and only if maxj (Zij ) = Zic . [sent-247, score-0.75]

81 If highly noisy data, use the previous step result to initialize k-means, or EM, clustering on the rows of Z. [sent-249, score-0.346]

82 The number of groups and the corresponding segmentation were obtained automatically. [sent-252, score-0.232]

83 html 5 Discussion & Conclusions Spectral clustering practitioners know that selecting good parameters to tune the clustering process is an art requiring skill and patience. [sent-260, score-0.608]

84 Automating spectral clustering was the main motivation for this study. [sent-261, score-0.41]

85 The key ideas we introduced are three: (a) using a local scale, rather than a global one, (b) estimating the scale from the data, and (c) rotating the eigenvectors to create the maximally sparse representation. [sent-262, score-0.473]

86 We proposed an automated spectral clustering algorithm based on these ideas: it computes automatically the scale and the number of groups and it can handle multi-scale data which are problematic for previous approaches. [sent-263, score-0.729]

87 For instance, the local scale σ might be better estimated by a method which relies on more informative local statistics. [sent-265, score-0.215]

88 Acknowledgments: Finally, we wish to thank Yair Weiss for providing us his code for spectral clustering. [sent-270, score-0.201]

89 Longuet-Higgins “Feature grouping by ‘relocalisation’ of eigenvectors of the proximity matrix” In Proc. [sent-301, score-0.27]

90 A Recovering the Aligning Rotation To ﬁnd the best alignment for a set of eigenvectors we adopt a gradient descent scheme similar to that suggested in [2]. [sent-314, score-0.648]

91 There, Givens rotations where used to recover a rotation which diagonalizes a symmetric matrix by minimizing a cost function which measures the diagonality of the matrix. [sent-315, score-0.399]

92 Similarly, here, we deﬁne a cost function which measures the alignment quality of a set of vectors and prove that the gradient descent, using Givens rotations, converges. [sent-316, score-0.316]

93 Note, that the indices mi of the maximal entries of the rows of X might be different than those of the optimal Z. [sent-320, score-0.315]

94 Using the gradient descent scheme allows to increase the cost corresponding to part of the rows as long as the overall cost is reduced, thus enabling changing the indices mi . [sent-322, score-0.624]

95 Similar to [2] we wish to represent the rotation matrix R in terms of the smallest possible ˜ number of parameters. [sent-323, score-0.221]

96 Let Gi,j,θ denote a Givens rotation [1] of θ radians (counterclockwise) in the (i, j) coordinate plane. [sent-324, score-0.2]

97 Hence, ﬁnding the aligning rotation amounts to minimizing the cost function J over Θ ∈ [−π/2, π/2)K . [sent-329, score-0.3]

98 Note, that at Θ = 0 we have Zij = 0 for j = mi , Zimi = Mi , and ∂Mi ∂θk = Θ=0 ∂Zimi = ∂θk (k) Aimi (i. [sent-338, score-0.171]

99 , near Θ = 0 the maximal ∂2J ∂θl ∂θk entry for each row does not change its index). [sent-340, score-0.203]

100 mi = ik or mi = jk 0 if k = l otherwise where (ik , jk ) is the pair (i, j) corresponding to the index k in the index re-mapping discussed above. [sent-344, score-0.51]

similar papers computed by tfidf model

tfidf for this paper:

wordName wordTfidf (topN-words)

[('zij', 0.265), ('clustering', 0.247), ('eigenvectors', 0.228), ('af', 0.221), ('nities', 0.182), ('aij', 0.179), ('mi', 0.171), ('groups', 0.168), ('si', 0.164), ('spectral', 0.163), ('eigenvalues', 0.155), ('group', 0.144), ('sj', 0.142), ('alignment', 0.141), ('clusters', 0.135), ('nity', 0.133), ('rotation', 0.123), ('fkl', 0.122), ('row', 0.108), ('givens', 0.106), ('xc', 0.106), ('perona', 0.101), ('rows', 0.099), ('automatically', 0.094), ('suggested', 0.093), ('aligning', 0.093), ('zimi', 0.091), ('cluster', 0.088), ('cost', 0.084), ('scaling', 0.083), ('xr', 0.079), ('local', 0.079), ('suppression', 0.077), ('selecting', 0.077), ('eigenvalue', 0.077), ('coordinate', 0.077), ('kannan', 0.073), ('multiplicity', 0.073), ('rotations', 0.072), ('scales', 0.07), ('descent', 0.069), ('rn', 0.069), ('canonical', 0.068), ('shi', 0.067), ('scheme', 0.066), ('segmentation', 0.064), ('aligns', 0.064), ('malized', 0.061), ('njw', 0.061), ('padded', 0.061), ('polito', 0.061), ('recover', 0.06), ('matrix', 0.06), ('eigenvector', 0.058), ('scale', 0.057), ('rotating', 0.056), ('blocks', 0.055), ('lihi', 0.053), ('eigensolver', 0.053), ('wise', 0.053), ('estimating', 0.053), ('gradient', 0.051), ('entry', 0.05), ('weiss', 0.049), ('cb', 0.048), ('est', 0.048), ('resides', 0.045), ('award', 0.045), ('jk', 0.045), ('maximal', 0.045), ('locally', 0.045), ('union', 0.043), ('dii', 0.043), ('pasadena', 0.043), ('zeros', 0.043), ('largest', 0.042), ('uk', 0.042), ('grouping', 0.042), ('scaled', 0.042), ('minimal', 0.041), ('circle', 0.041), ('top', 0.041), ('highlights', 0.04), ('maxj', 0.04), ('rl', 0.04), ('irregular', 0.04), ('quality', 0.04), ('magnitude', 0.039), ('background', 0.039), ('index', 0.039), ('repeated', 0.039), ('wish', 0.038), ('suggest', 0.037), ('iv', 0.037), ('ga', 0.037), ('tune', 0.037), ('initialization', 0.037), ('rotated', 0.036), ('yu', 0.036), ('clutter', 0.036)]

similar papers list:

simIndex simValue paperId paperTitle

same-paper 1 0.99999958 161 nips-2004-Self-Tuning Spectral Clustering

Author: Lihi Zelnik-manor, Pietro Perona

2 0.25222942 79 nips-2004-Hierarchical Eigensolver for Transition Matrices in Spectral Methods

Author: Chakra Chennubhotla, Allan D. Jepson

Abstract: We show how to build hierarchical, reduced-rank representation for large stochastic matrices and use this representation to design an efﬁcient algorithm for computing the largest eigenvalues, and the corresponding eigenvectors. In particular, the eigen problem is ﬁrst solved at the coarsest level of the representation. The approximate eigen solution is then interpolated over successive levels of the hierarchy. A small number of power iterations are employed at each stage to correct the eigen solution. The typical speedups obtained by a Matlab implementation of our fast eigensolver over a standard sparse matrix eigensolver [13] are at least a factor of ten for large image sizes. The hierarchical representation has proven to be effective in a min-cut based segmentation algorithm that we proposed recently [8]. 1 Spectral Methods Graph-theoretic spectral methods have gained popularity in a variety of application domains: segmenting images [22]; embedding in low-dimensional spaces [4, 5, 8]; and clustering parallel scientiﬁc computation tasks [19]. Spectral methods enable the study of properties global to a dataset, using only local (pairwise) similarity or afﬁnity measurements between the data points. The global properties that emerge are best understood in terms of a random walk formulation on the graph. For example, the graph can be partitioned into clusters by analyzing the perturbations to the stationary distribution of a Markovian relaxation process deﬁned in terms of the afﬁnity weights [17, 18, 24, 7]. The Markovian relaxation process need never be explicitly carried out; instead, it can be analytically expressed using the leading order eigenvectors, and eigenvalues, of the Markov transition matrix. In this paper we consider the practical application of spectral methods to large datasets. In particular, the eigen decomposition can be very expensive, on the order of O(n 3 ), where n is the number of nodes in the graph. While it is possible to compute analytically the ﬁrst eigenvector (see §3 below), the remaining subspace of vectors (necessary for say clustering) has to be explicitly computed. A typical approach to dealing with this difﬁculty is to ﬁrst sparsify the links in the graph [22] and then apply an efﬁcient eigensolver [13, 23, 3]. In comparison, we propose in this paper a specialized eigensolver suitable for large stochastic matrices with known stationary distributions. In particular, we exploit the spectral properties of the Markov transition matrix to generate hierarchical, successively lower-ranked approximations to the full transition matrix. The eigen problem is solved directly at the coarsest level of representation. The approximate eigen solution is then interpolated over successive levels of the hierarchy, using a small number of power iterations to correct the solution at each stage. 2 Previous Work One approach to speeding up the eigen decomposition is to use the fact that the columns of the afﬁnity matrix are typically correlated. The idea then is to pick a small number of representative columns to perform eigen decomposition via SVD. For example, in the Nystrom approximation procedure, originally proposed for integral eigenvalue problems, the idea is to randomly pick a small set of m columns; generate the corresponding afﬁnity matrix; solve the eigenproblem and ﬁnally extend the solution to the complete graph [9, 10]. The Nystrom method has also been recently applied in the kernel learning methods for fast Gaussian process classiﬁcation and regression [25]. Other sampling-based approaches include the work reported in [1, 2, 11]. Our starting point is the transition matrix generated from afﬁnity weights and we show how building a representational hierarchy follows naturally from considering the stochastic matrix. A closely related work is the paper by Lin on reduced rank approximations of transition matrices [14]. We differ in how we approximate the transition matrices, in particular our objective function is computationally less expensive to solve. In particular, one of our goals in reducing transition matrices is to develop a fast, specialized eigen solver for spectral clustering. Fast eigensolving is also the goal in ACE [12], where successive levels in the hierarchy can potentially have negative afﬁnities. A graph coarsening process for clustering was also pursued in [21, 3]. 3 Markov Chain Terminology We ﬁrst provide a brief overview of the Markov chain terminology here (for more details see [17, 15, 6]). We consider an undirected graph G = (V, E) with vertices vi , for i = {1, . . . , n}, and edges ei,j with non-negative weights ai,j . Here the weight ai,j represents the afﬁnity between vertices vi and vj . The afﬁnities are represented by a non-negative, symmetric n × n matrix A having weights ai,j as elements. The degree of a node j is n n deﬁned to be: dj = i=1 ai,j = j=1 aj,i , where we deﬁne D = diag(d1 , . . . , dn ). A Markov chain is deﬁned using these afﬁnities by setting a transition probability matrix M = AD −1 , where the columns of M each sum to 1. The transition probability matrix deﬁnes the random walk of a particle on the graph G. The random walk need never be explicitly carried out; instead, it can be analytically expressed using the leading order eigenvectors, and eigenvalues, of the Markov transition matrix. Because the stochastic matrices need not be symmetric in general, a direct eigen decomposition step is not preferred for reasons of instability. This problem is easily circumvented by considering a normalized afﬁnity matrix: L = D −1/2 AD−1/2 , which is related to the stochastic matrix by a similarity transformation: L = D −1/2 M D1/2 . Because L is symmetric, it can be diagonalized: L = U ΛU T , where U = [u1 , u2 , · · · , un ] is an orthogonal set of eigenvectors and Λ is a diagonal matrix of eigenvalues [λ1 , λ2 , · · · , λn ] sorted in decreasing order. The eigenvectors have unit length uk = 1 and from the form of A and D it can be shown that the eigenvalues λi ∈ (−1, 1], with at least one eigenvalue equal to one. Without loss of generality, we take λ1 = 1. Because L and M are similar we can perform an eigen decomposition of the Markov transition matrix as: M = D1/2 LD−1/2 = D1/2 U Λ U T D−1/2 . Thus an eigenvector u of L corresponds to an eigenvector D 1/2 u of M with the same eigenvalue λ. The Markovian relaxation process after β iterations, namely M β , can be represented as: M β = D1/2 U Λβ U T D−1/2 . Therefore, a particle undertaking a random walk with an initial distribution p 0 acquires after β steps a distribution p β given by: p β = M β p 0 . Assuming the graph is connected, as β → ∞, the Markov chain approaches a unique n stationary distribution given by π = diag(D)/ i=1 di , and thus, M ∞ = π1T , where 1 is a n-dim column vector of all ones. Observe that π is an eigenvector of M as it is easy to show that M π = π and the corresponding eigenvalue is 1. Next, we show how to generate hierarchical, successively low-ranked approximations for the transition matrix M . 4 Building a Hierarchy of Transition Matrices The goal is to generate a very fast approximation, while simultaneously achieving sufﬁcient accuracy. For notational ease, we think of M as a ﬁne-scale representation and M as some coarse-scale approximation to be derived here. By coarsening M further, we can generate successive levels of the representation hierarchy. We use the stationary distribution π to construct a corresponding coarse-scale stationary distribution δ. As we just discussed a critical property of the ﬁne scale Markov matrix M is that it is similar to the symmetric matrix L and we wish to preserve this property at every level of the representation hierarchy. 4.1 Deriving Coarse-Scale Stationary Distribution We begin by expressing the stationary distribution π as a probabilistic mixture of latent distributions. In matrix notation, we have (1) π = K δ, where δ is an unknown mixture coefﬁcient vector of length m, K is an n × m non-negative n kernel matrix whose columns are latent distributions that each sum to 1: i=1 Ki,j = 1 and m n. It is easy to derive a maximum likelihood approximation of δ using an EM type algorithm [16]. The main step is to ﬁnd a stationary point δ, λ for the Lagrangian: m n i=1 m Ki,j δj + λ πi ln E≡− j=1 δj − 1 . (2) j=1 An implicit step in this EM procedure is to compute the the ownership probability r i,j of the j th kernel (or node) at the coarse scale for the ith node on the ﬁne scale and is given by ri,j = δj Ki,j . m k=1 δk Ki,k (3) The EM procedure allows for an update of both δ and the latent distributions in the kernel matrix K (see §8.3.1 in [6]). For initialization, δ is taken to be uniform over the coarse-scale states. But in choosing kernels K, we provide a good initialization for the EM procedure. Speciﬁcally, the Markov matrix M is diffused using a small number of iterations to get M β . The diffusion causes random walks from neighboring nodes to be less distinguishable. This in turn helps us select a small number of columns of M β in a fast and greedy way to be the kernel matrix K. We defer the exact details on kernel selection to a later section (§4.3). 4.2 Deriving the Coarse-Scale Transition Matrix In order to deﬁne M , the coarse-scale transition matrix, we break it down into three steps. First, the Markov chain propagation at the coarse scale can be deﬁned as: q k+1 = M q k , (4) where q is the coarse scale probability distribution after k steps of the random walk. Second, we expand q k into the ﬁne scale using the kernels K resulting in a ﬁne scale probability distribution p k : p k = Kq k . (5) k Finally, we lift p k back into the coarse scale by using the ownership probability of the j th kernel for the ith node on the ﬁne grid: n qjk+1 = ri,j pik i=1 (6) Substituting for Eqs.(3) and (5) in Eq. 6 gives n m qjk+1 = i=1 n Ki,t qtk = ri,j t=1 i=1 δj Ki,j m k=1 δk Ki,k m Ki,t qtk . (7) t=1 We can write the preceding equation in a matrix form: q k+1 = diag( δ ) K T diag K δ −1 Kq k . (8) Comparing this with Eq. 4, we can derive the transition matrix M as: M = diag( δ ) K T diag K δ −1 (9) K. It is easy to see that δ = M δ, so δ is the stationary distribution for M . Following the deﬁnition of M , and its stationary distribution δ, we can generate a symmetric coarse scale afﬁnity matrix A given by A = M diag(δ) = diag( δ ) K T diag K δ −1 Kdiag(δ) , (10) where we substitute for the expression M from Eq. 9. The coarse-scale afﬁnity matrix A is then normalized to get: L = D−1/2 AD−1/2 ; D = diag(d1 , d2 , · · · , dm ), (11) where dj is the degree of node j in the coarse-scale graph represented by the matrix A (see §3 for degree deﬁnition). Thus, the coarse scale Markov matrix M is precisely similar to a symmetric matrix L. 4.3 Selecting Kernels For demonstration purpose, we present the kernel selection details on the image of an eye shown below. To begin with, a random walk is deﬁned where each pixel in the test image is associated with a vertex of the graph G. The edges in G are deﬁned by the standard 8-neighbourhood of each pixel. For the demonstrations in this paper, the edge weight ai,j between neighbouring pixels xi and xj is given by a function of the difference in the 2 corresponding intensities I(xi ) and I(xj ): ai,j = exp(−(I(xi ) − I(xj ))2 /2σa ), where σa is set according to the median absolute difference |I(xi ) − I(xj )| between neighbours measured over the entire image. The afﬁnity matrix A with the edge weights is then used to generate a Markov transition matrix M . The kernel selection process we use is fast and greedy. First, the ﬁne scale Markov matrix M is diffused to M β using β = 4. The Markov matrix M is sparse as we make the afﬁnity matrix A sparse. Every column in the diffused matrix M β is a potential kernel. To facilitate the selection process, the second step is to rank order the columns of M β based on a probability value in the stationary distribution π. Third, the kernels (i.e. columns of M β ) are picked in such a way that for a kernel Ki all of the neighbours of pixel i which are within the half-height of the the maximum value in the kernel Ki are suppressed from the selection process. Finally, the kernel selection is continued until every pixel in the image is within a half-height of the peak value of at least one kernel. If M is a full matrix, to avoid the expense of computing M β explicitly, random kernel centers can be selected, and only the corresponding columns of M β need be computed. We show results from a three-scale hierarchy on the eye image (below). The image has 25 × 20 pixels but is shown here enlarged for clarity. At the ﬁrst coarse scale 83 kernels are picked. The kernels each correspond to a different column in the ﬁne scale transition matrix and the pixels giving rise to these kernels are shown numbered on the image. Using these kernels as an initialization, the EM procedure derives a coarse-scale stationary distribution δ 21 14 26 4 (Eq. 2), while simultaneously updating the kernel ma12 27 2 19 trix. Using the newly updated kernel matrix K and the 5 8 13 23 30 18 6 9 derived stationary distribution δ a transition matrix M 28 20 15 32 10 22 is generated (Eq. 9). The coarse scale Markov matrix 24 17 7 is then diffused to M β , again using β = 4. The kernel Coarse Scale 1 Coarse Scale 2 selection algorithm is reapplied, this time picking 32 kernels for the second coarse scale. Larger values of β cause the coarser level to have fewer elements. But the exact number of elements depends on the form of the kernels themselves. For the random experiments that we describe later in §6 we found β = 2 in the ﬁrst iteration and 4 thereafter causes the number of kernels to be reduced by a factor of roughly 1/3 to 1/4 at each level. 72 28 35 44 51 64 82 4 12 31 56 19 77 36 45 52 65 13 57 23 37 5 40 53 63 73 14 29 6 66 38 74 47 24 7 30 41 54 71 78 58 15 8 20 39 48 59 67 25 68 79 21 16 2 11 26 42 49 55 60 75 32 83 43 9 76 50 17 27 61 33 69 80 3 46 18 70 81 34 10 62 22 1 25 11 1 3 16 31 29 At coarser levels of the hierarchy, we expect the kernels to get less sparse and so will the afﬁnity and the transition matrices. In order to promote sparsity at successive levels of the hierarchy we sparsify A by zeroing out elements associated with “small” transition probabilities in M . However, in the experiments described later in §6, we observe this sparsiﬁcation step to be not critical. To summarize, we use the stationary distribution π at the ﬁne-scale to derive a transition matrix M , and its stationary distribution δ, at the coarse-scale. The coarse scale transition in turn helps to derive an afﬁnity matrix A and its normalized version L. It is obvious that this procedure can be repeated recursively. We describe next how to use this representation hierarchy for building a fast eigensolver. 5 Fast EigenSolver Our goal in generating a hierarchical representation of a transition matrix is to develop a fast, specialized eigen solver for spectral clustering. To this end, we perform a full eigen decomposition of the normalized afﬁnity matrix only at the coarsest level. As discussed in the previous section, the afﬁnity matrix at the coarsest level is not likely to be sparse, hence it will need a full (as opposed to a sparse) version of an eigen solver. However it is typically the case that e ≤ m n (even in the case of the three-scale hierarchy that we just considered) and hence we expect this step to be the least expensive computationally. The resulting eigenvectors are interpolated to the next lower level of the hierarchy by a process which will be described next. Because the eigen interpolation process between every adjacent pair of scales in the hierarchy is similar, we will assume we have access to the leading eigenvectors U (size: m × e) for the normalized afﬁnity matrix L (size: m × m) and describe how to generate the leading eigenvectors U (size: n × e), and the leading eigenvalues S (size: e × 1), for the ﬁne-scale normalized afﬁnity matrix L (size: n × n). There are several steps to the eigen interpolation process and in the discussion that follows we refer to the lines in the pseudo-code presented below. First, the coarse-scale eigenvectors U can be interpolated using the kernel matrix K to generate U = K U , an approximation for the ﬁne-scale eigenvectors (line 9). Second, interpolation alone is unlikely to set the directions of U exactly aligned with U L , the vectors one would obtain by a direct eigen decomposition of the ﬁne-scale normalized afﬁnity matrix L. We therefore update the directions in U by applying a small number of power iterations with L, as given in lines 13-15. e e function (U, S) = CoarseToFine(L, K, U , S) 1: INPUT 2: L, K ⇐ {L is n × n and K is n × m where m n} e e e e 3: U /S ⇐ {leading coarse-scale eigenvectors/eigenvalues of L. U is of size m × e, e ≤ m} 4: OUTPUT 5: U, S ⇐ {leading ﬁne-scale eigenvectors/eigenvalues of L. U is n × e and S is e × 1.} x 10 0.4 3 0.96 0.94 0.92 0.9 0.35 2.5 Relative Error Absolute Relative Error 0.98 Eigen Value |δλ|λ−1 −3 Eigen Spectrum 1 2 1.5 1 5 10 15 20 Eigen Index (a) 25 30 0.2 0.15 0.1 0.5 0.88 0.3 0.25 0.05 5 10 15 20 Eigen Index (b) 25 30 5 10 15 20 Eigen Index 25 30 (c) Figure 1: Hierarchical eigensolver results. (a) comparing ground truth eigenvalues S L (red circles) with multi-scale eigensolver spectrum S (blue line) (b) Relative absolute error between eigenvalues: |S−SL | (c) Eigenvector mismatch: 1 − diag |U T UL | , between SL eigenvectors U derived by the multi-scale eigensolver and the ground truth U L . Observe the slight mismatch in the last few eigenvectors, but excellent agreement in the leading eigenvectors (see text). 6: CONSTANTS: TOL = 1e-4; POWER ITERS = 50 7: “ ” e 8: TPI = min POWER ITERS, log(e × eps/TOL)/ log(min(S)) {eps: machine accuracy} e 9: U = K U {interpolation from coarse to ﬁne} 10: while not converged do 11: Uold = U {n × e matrix, e n} 12: for i = 1 to TPI do 13: U ⇐ LU 14: end for 15: U ⇐ Gram-Schmidt(U ) {orthogonalize U } 16: Le = U T LU {L may be sparse, but Le need not be.} 17: Ue Se UeT = svd(Le ) {eigenanalysis of Le , which is of size e × e.} 18: U ⇐ U Ue {update the leading eigenvectors of L} 19: S = diag(Se ) {grab the leading eigenvalues of L} T 20: innerProd = 1 − diag( Uold U ) {1 is a e × 1 vector of all ones} 21: converged = max[abs(innerProd)] < TOL 22: end while The number of power iterations TPI can be bounded as discussed next. Suppose v = U c where U is a matrix of true eigenvectors and c is a coefﬁcient vector for an arbitrary vector v. After TPI power iterations v becomes v = U diag(S TPI )c, where S has the exact eigenvalues. In order for the component of a vector v in the direction Ue (the eth column of U ) not to be swamped by other components, we can limit it’s decay after TPI iterations as TPI follows: (S(e)/S(1)) >= e×eps/TOL, where S(e) is the exact eth eigenvalue, S(1) = 1, eps is the machine precision, TOL is requested accuracy. Because we do not have access to the exact value S(e) at the beginning of the interpolation procedure, we estimate it from the coarse eigenvalues S. This leads to a bound on the power iterations TPI, as derived on the line 9 above. Third, the interpolation process and the power iterations need not preserve orthogonality in the eigenvectors in U . We ﬁx this by Gram-Schmidt orthogonalization procedure (line 16). Finally, there is a still a problem with power iterations that needs to be resolved, in that it is very hard to separate nearby eigenvalues. In particular, for the convergence of the power iterations the ratio that matters is between the (e + 1) st and eth eigenvalues. So the idea we pursue is to use the power iterations only to separate the reduced space of eigenvectors (of dimension e) from the orthogonal subspace (of dimension n − e). We then use a full SVD on the reduced space to update the leading eigenvectors U , and eigenvalues S, for the ﬁne-scale (lines 17-20). This idea is similar to computing the Ritz values and Ritz vectors in a Rayleigh-Ritz method. 6 Interpolation Results Our multi-scale decomposition code is in Matlab. For the direct eigen decomposition, we have used the Matlab program svds.m which invokes the compiled ARPACKC routine [13], with a default convergence tolerance of 1e-10. In Fig. 1a we compare the spectrum S obtained from a three-scale decomposition on the eye image (blue line) with the ground truth, which is the spectrum SL resulting from direct eigen decomposition of the ﬁne-scale normalized afﬁnity matrices L (red circles). There is an excellent agreement in the leading eigenvalues. To illustrate this, we show absolute relative error between the spectra: |S−SL | in Fig. 1b. The spectra agree mostly, except for SL the last few eigenvalues. For a quantitative comparison between the eigenvectors, we plot in Fig. 1c the following measure: 1 − diag(|U T UL |), where U is the matrix of eigenvectors obtained by the multi-scale approximation, UL is the ground-truth resulting from a direct eigen decomposition of the ﬁne-scale afﬁnity matrix L and 1 is a vector of all ones. The relative error plot demonstrates a close match, within the tolerance threshold of 1e-4 that we chose for the multi-scale method, in the leading eigenvector directions between the two methods. The relative error is high with the last few eigen vectors, which suggests that the power iterations have not clearly separated them from other directions. So, the strategy we suggest is to pad the required number of leading eigen basis by about 20% before invoking the multi-scale procedure. Obviously, the number of hierarchical stages for the multi-scale procedure must be chosen such that the transition matrix at the coarsest scale can accommodate the slight increase in the subspace dimensions. For lack of space we are omitting extra results (see Ch.8 in [6]). Next we measure the time the hierarchical eigensolver takes to compute the leading eigenbasis for various input sizes, in comparison with the svds.m procedure [13]. We form images of different input sizes by Gaussian smoothing of i.i.d noise. The Gaussian function has a standard deviation of 3 pixels. The edges in graph G are deﬁned by the standard 8-neighbourhood of each pixel. The edge weights between neighbouring pixels are simply given by a function of the difference in the corresponding intensities (see §4.3). The afﬁnity matrix A with the edge weights is then used to generate a Markov transition matrix M . The fast eigensolver is run on ten different instances of the input image of a given size and the average of these times is reported here. For a fair comparison between the two procedures, we set the convergence tolerance value for the svds.m procedure to be 1e-4, the same as the one used for the fast eigensolver. We found the hierarchical representation derived from this tolerance threshold to be sufﬁciently accurate for a novel min-cut based segmentation results that we reported in [8]. Also, the subspace dimensionality is ﬁxed to be 51 where we expect (and indeed observe) the leading 40 eigenpairs derived from the multi-scale procedure to be accurate. Hence, while invoking svds.m we compute only the leading 41 eigenpairs. In the table shown below, the ﬁrst column corresponds to the number of nodes in the graph, while the second and third columns report the time taken in seconds by the svds.m procedure and the Matlab implementation of the multi-scale eigensolver respectively. The fourth column reports the speedups of the multi-scale eigensolver over svds.m procedure on a standard desktop (Intel P4, 2.5GHz, 1GB RAM). Lowering the tolerance threshold for svds.m made it faster by about 20 − 30%. Despite this, the multi-scale algorithm clearly outperforms the svds.m procedure. The most expensive step in the multi-scale algorithm is the power iteration required in the last stage, that is interpolating eigenvectors from the ﬁrst coarse scale to the required ﬁne scale. The complexity is of the order of n × e where e is the subspace dimensionality and n is the size of the graph. Indeed, from the table we can see that the multi-scale procedure is taking time roughly proportional to n. Deviations from the linear trend are observed at speciﬁc values of n, which we believe are due to the n 322 632 642 652 1002 1272 1282 1292 1602 2552 2562 2572 5112 5122 5132 6002 7002 8002 svds.m 1.6 10.8 20.5 12.6 44.2 91.1 230.9 96.9 179.3 819.2 2170.8 871.7 7977.2 20269 7887.2 10841.4 15048.8 Multi-Scale 1.5 4.9 5.5 5.1 13.1 20.4 35.2 20.9 34.4 90.3 188.7 93.3 458.8 739.3 461.9 644.2 1162.4 1936.6 Speedup 1.1 2.2 3.7 2.5 3.4 4.5 6.6 4.6 5.2 9.1 11.5 9.3 17.4 27.4 17.1 16.8 12.9 variations in the difﬁculty of the speciﬁc eigenvalue problem (eg. nearly multiple eigenvalues). The hierarchical representation has proven to be effective in a min-cut based segmentation algorithm that we proposed recently [8]. Here we explored the use of random walks and associated spectral embedding techniques for the automatic generation of suitable proposal (source and sink) regions for a min-cut based algorithm. The multiscale algorithm was used to generate the 40 leading eigenvectors of large transition matrices (eg. size 20K × 20K). In terms of future work, it will be useful to compare our work with other approximate methods for SVD such as [23]. Ack: We thank S. Roweis, F. Estrada and M. Sakr for valuable comments. References [1] D. Achlioptas and F. McSherry. Fast Computation of Low-Rank Approximations. STOC, 2001. [2] D. Achlioptas et al Sampling Techniques for Kernel Methods. NIPS, 2001. [3] S. Barnard and H. Simon Fast Multilevel Implementation of Recursive Spectral Bisection for Partitioning Unstructured Problems. PPSC, 627-632. [4] M. Belkin et al Laplacian Eigenmaps and Spectral Techniques for Embedding. NIPS, 2001. [5] M. Brand et al A unifying theorem for spectral embedding and clustering. AI & STATS, 2002. [6] C. Chennubhotla. Spectral Methods for Multi-scale Feature Extraction and Spectral Clustering. http://www.cs.toronto.edu/˜chakra/thesis.pdf Ph.D Thesis, Department of Computer Science, University of Toronto, Canada, 2004. [7] C. Chennubhotla and A. Jepson. Half-Lives of EigenFlows for Spectral Clustering. NIPS, 2002. [8] F. Estrada, A. Jepson and C. Chennubhotla. Spectral Embedding and Min-Cut for Image Segmentation. Manuscript Under Review, 2004. [9] C. Fowlkes et al Efﬁcient spatiotemporal grouping using the Nystrom method. CVPR, 2001. [10] S. Belongie et al Spectral Partitioning with Indeﬁnite Kernels using Nystrom app. ECCV, 2002. [11] A. Frieze et al Fast Monte-Carlo Algorithms for ﬁnding low-rank approximations. FOCS, 1998. [12] Y. Koren et al ACE: A Fast Multiscale Eigenvectors Computation for Drawing Huge Graphs IEEE Symp. on InfoVis 2002, pp. 137-144 [13] R. B. Lehoucq, D. C. Sorensen and C. Yang. ARPACK User Guide: Solution of Large Scale Eigenvalue Problems by Implicitly Restarted Arnoldi Methods. SIAM 1998. [14] J. J. Lin. Reduced Rank Approximations of Transition Matrices. AI & STATS, 2002. [15] L. Lova’sz. Random Walks on Graphs: A Survey Combinatorics, 1996, 353–398. [16] G. J. McLachlan et al Mixture Models: Inference and Applications to Clustering. 1988 [17] M. Meila and J. Shi. A random walks view of spectral segmentation. AI & STATS, 2001. [18] A. Ng, M. Jordan and Y. Weiss. On Spectral Clustering: analysis and an algorithm NIPS, 2001. [19] A. Pothen Graph partitioning algorithms with applications to scientiﬁc computing. Parallel Numerical Algorithms, D. E. Keyes et al (eds.), Kluwer Academic Press, 1996. [20] G. L. Scott et al Feature grouping by relocalization of eigenvectors of the proximity matrix. BMVC, pg. 103-108, 1990. [21] E. Sharon et al Fast Multiscale Image Segmentation CVPR, I:70-77, 2000. [22] J. Shi and J. Malik. Normalized cuts and image segmentation. PAMI, August, 2000. [23] H. Simon et al Low-Rank Matrix Approximation Using the Lanczos Bidiagonalization Process with Applications SIAM J. of Sci. Comp. 21(6):2257-2274, 2000. [24] N. Tishby et al Data clustering by Markovian Relaxation NIPS, 2001. [25] C. Williams et al Using the Nystrom method to speed up the kernel machines. NIPS, 2001.

3 0.22374403 103 nips-2004-Limits of Spectral Clustering

Author: Ulrike V. Luxburg, Olivier Bousquet, Mikhail Belkin

Abstract: An important aspect of clustering algorithms is whether the partitions constructed on ﬁnite samples converge to a useful clustering of the whole data space as the sample size increases. This paper investigates this question for normalized and unnormalized versions of the popular spectral clustering algorithm. Surprisingly, the convergence of unnormalized spectral clustering is more difﬁcult to handle than the normalized case. Even though recently some ﬁrst results on the convergence of normalized spectral clustering have been obtained, for the unnormalized case we have to develop a completely new approach combining tools from numerical integration, spectral and perturbation theory, and probability. It turns out that while in the normalized case, spectral clustering usually converges to a nice partition of the data space, in the unnormalized case the same only holds under strong additional assumptions which are not always satisﬁed. We conclude that our analysis gives strong evidence for the superiority of normalized spectral clustering. It also provides a basis for future exploration of other Laplacian-based methods. 1

4 0.1720594 31 nips-2004-Blind One-microphone Speech Separation: A Spectral Learning Approach

Author: Francis R. Bach, Michael I. Jordan

Abstract: We present an algorithm to perform blind, one-microphone speech separation. Our algorithm separates mixtures of speech without modeling individual speakers. Instead, we formulate the problem of speech separation as a problem in segmenting the spectrogram of the signal into two or more disjoint sets. We build feature sets for our segmenter using classical cues from speech psychophysics. We then combine these features into parameterized afﬁnity matrices. We also take advantage of the fact that we can generate training examples for segmentation by artiﬁcially superposing separately-recorded signals. Thus the parameters of the afﬁnity matrices can be tuned using recent work on learning spectral clustering [1]. This yields an adaptive, speech-speciﬁc segmentation algorithm that can successfully separate one-microphone speech mixtures. 1

5 0.1715461 13 nips-2004-A Three Tiered Approach for Articulated Object Action Modeling and Recognition

Author: Le Lu, Gregory D. Hager, Laurent Younes

Abstract: Visual action recognition is an important problem in computer vision. In this paper, we propose a new method to probabilistically model and recognize actions of articulated objects, such as hand or body gestures, in image sequences. Our method consists of three levels of representation. At the low level, we ﬁrst extract a feature vector invariant to scale and in-plane rotation by using the Fourier transform of a circular spatial histogram. Then, spectral partitioning [20] is utilized to obtain an initial clustering; this clustering is then reﬁned using a temporal smoothness constraint. Gaussian mixture model (GMM) based clustering and density estimation in the subspace of linear discriminant analysis (LDA) are then applied to thousands of image feature vectors to obtain an intermediate level representation. Finally, at the high level we build a temporal multiresolution histogram model for each action by aggregating the clustering weights of sampled images belonging to that action. We discuss how this high level representation can be extended to achieve temporal scaling invariance and to include Bi-gram or Multi-gram transition information. Both image clustering and action recognition/segmentation results are given to show the validity of our three tiered representation.

6 0.17047203 115 nips-2004-Maximum Margin Clustering

7 0.15037368 90 nips-2004-Joint Probabilistic Curve Clustering and Alignment

8 0.14431772 133 nips-2004-Nonparametric Transforms of Graph Kernels for Semi-Supervised Learning

9 0.12728186 61 nips-2004-Efficient Out-of-Sample Extension of Dominant-Set Clusters

10 0.12194566 150 nips-2004-Proximity Graphs for Clustering and Manifold Learning

11 0.11852661 174 nips-2004-Spike Sorting: Bayesian Clustering of Non-Stationary Data

12 0.10880237 77 nips-2004-Hierarchical Clustering of a Mixture Model

13 0.10203286 167 nips-2004-Semi-supervised Learning with Penalized Probabilistic Clustering

14 0.093724191 188 nips-2004-The Laplacian PDF Distance: A Cost Function for Clustering in a Kernel Feature Space

15 0.084137216 9 nips-2004-A Method for Inferring Label Sampling Mechanisms in Semi-Supervised Learning

16 0.069258049 185 nips-2004-The Convergence of Contrastive Divergences

17 0.067414947 18 nips-2004-Algebraic Set Kernels with Application to Inference Over Local Image Representations

18 0.066397235 122 nips-2004-Modelling Uncertainty in the Game of Go

19 0.065107316 125 nips-2004-Multiple Relational Embedding

20 0.064714834 10 nips-2004-A Probabilistic Model for Online Document Clustering with Application to Novelty Detection

similar papers computed by lsi model

lsi for this paper:

topicId topicWeight

[(0, -0.257), (1, 0.091), (2, -0.092), (3, -0.156), (4, -0.137), (5, -0.132), (6, -0.254), (7, 0.084), (8, -0.232), (9, -0.022), (10, -0.214), (11, 0.045), (12, -0.101), (13, -0.021), (14, 0.031), (15, -0.054), (16, -0.115), (17, -0.016), (18, -0.168), (19, 0.004), (20, 0.061), (21, 0.035), (22, -0.008), (23, -0.071), (24, 0.004), (25, 0.109), (26, -0.015), (27, 0.016), (28, 0.022), (29, 0.063), (30, 0.044), (31, 0.059), (32, -0.053), (33, 0.05), (34, -0.003), (35, 0.094), (36, -0.064), (37, -0.041), (38, -0.066), (39, -0.042), (40, -0.012), (41, -0.022), (42, -0.056), (43, -0.026), (44, 0.0), (45, -0.137), (46, -0.053), (47, -0.027), (48, 0.083), (49, 0.108)]

similar papers list:

simIndex simValue paperId paperTitle

same-paper 1 0.98028123 161 nips-2004-Self-Tuning Spectral Clustering

Author: Lihi Zelnik-manor, Pietro Perona

2 0.80918354 103 nips-2004-Limits of Spectral Clustering

Author: Ulrike V. Luxburg, Olivier Bousquet, Mikhail Belkin

3 0.78293663 79 nips-2004-Hierarchical Eigensolver for Transition Matrices in Spectral Methods

Author: Chakra Chennubhotla, Allan D. Jepson

4 0.59537059 61 nips-2004-Efficient Out-of-Sample Extension of Dominant-Set Clusters

Author: Massimiliano Pavan, Marcello Pelillo

Abstract: Dominant sets are a new graph-theoretic concept that has proven to be relevant in pairwise data clustering problems, such as image segmentation. They generalize the notion of a maximal clique to edgeweighted graphs and have intriguing, non-trivial connections to continuous quadratic optimization and spectral-based grouping. We address the problem of grouping out-of-sample examples after the clustering process has taken place. This may serve either to drastically reduce the computational burden associated to the processing of very large data sets, or to efﬁciently deal with dynamic situations whereby data sets need to be updated continually. We show that the very notion of a dominant set offers a simple and efﬁcient way of doing this. Numerical experiments on various grouping problems show the effectiveness of the approach. 1

5 0.58607417 13 nips-2004-A Three Tiered Approach for Articulated Object Action Modeling and Recognition

Author: Le Lu, Gregory D. Hager, Laurent Younes

6 0.56056261 115 nips-2004-Maximum Margin Clustering

7 0.56030262 150 nips-2004-Proximity Graphs for Clustering and Manifold Learning

8 0.47719514 31 nips-2004-Blind One-microphone Speech Separation: A Spectral Learning Approach

9 0.45548329 167 nips-2004-Semi-supervised Learning with Penalized Probabilistic Clustering

10 0.45400563 90 nips-2004-Joint Probabilistic Curve Clustering and Alignment

11 0.43922827 174 nips-2004-Spike Sorting: Bayesian Clustering of Non-Stationary Data

12 0.4200083 133 nips-2004-Nonparametric Transforms of Graph Kernels for Semi-Supervised Learning

13 0.41586569 77 nips-2004-Hierarchical Clustering of a Mixture Model

14 0.39495951 185 nips-2004-The Convergence of Contrastive Divergences

15 0.38972786 207 nips-2004-ℓ₀-norm Minimization for Basis Selection

16 0.34274149 130 nips-2004-Newscast EM

17 0.33260915 199 nips-2004-Using Machine Learning to Break Visual Human Interaction Proofs (HIPs)

18 0.32405046 196 nips-2004-Triangle Fixing Algorithms for the Metric Nearness Problem

19 0.31874603 96 nips-2004-Learning, Regularization and Ill-Posed Inverse Problems

20 0.31606618 18 nips-2004-Algebraic Set Kernels with Application to Inference Over Local Image Representations

similar papers computed by lda model

lda for this paper:

topicId topicWeight

[(13, 0.096), (15, 0.176), (26, 0.097), (31, 0.016), (33, 0.238), (35, 0.022), (39, 0.044), (50, 0.038), (71, 0.015), (74, 0.167), (87, 0.014)]

similar papers list:

simIndex simValue paperId paperTitle

same-paper 1 0.9213919 161 nips-2004-Self-Tuning Spectral Clustering

Author: Lihi Zelnik-manor, Pietro Perona

2 0.8747533 68 nips-2004-Face Detection --- Efficient and Rank Deficient

Author: Wolf Kienzle, Matthias O. Franz, Bernhard Schölkopf, Gökhan H. Bakir

Abstract: This paper proposes a method for computing fast approximations to support vector decision functions in the ﬁeld of object detection. In the present approach we are building on an existing algorithm where the set of support vectors is replaced by a smaller, so-called reduced set of synthesized input space points. In contrast to the existing method that ﬁnds the reduced set via unconstrained optimization, we impose a structural constraint on the synthetic points such that the resulting approximations can be evaluated via separable ﬁlters. For applications that require scanning large images, this decreases the computational complexity by a signiﬁcant amount. Experimental results show that in face detection, rank deﬁcient approximations are 4 to 6 times faster than unconstrained reduced set systems. 1

3 0.87323743 174 nips-2004-Spike Sorting: Bayesian Clustering of Non-Stationary Data

Author: Aharon Bar-hillel, Adam Spiro, Eran Stark

Abstract: Spike sorting involves clustering spike trains recorded by a microelectrode according to the source neuron. It is a complicated problem, which requires a lot of human labor, partly due to the non-stationary nature of the data. We propose an automated technique for the clustering of non-stationary Gaussian sources in a Bayesian framework. At a ﬁrst search stage, data is divided into short time frames and candidate descriptions of the data as a mixture of Gaussians are computed for each frame. At a second stage transition probabilities between candidate mixtures are computed, and a globally optimal clustering is found as the MAP solution of the resulting probabilistic model. Transition probabilities are computed using local stationarity assumptions and are based on a Gaussian version of the Jensen-Shannon divergence. The method was applied to several recordings. The performance appeared almost indistinguishable from humans in a wide range of scenarios, including movement, merges, and splits of clusters. 1

4 0.87231034 189 nips-2004-The Power of Selective Memory: Self-Bounded Learning of Prediction Suffix Trees

Author: Ofer Dekel, Shai Shalev-shwartz, Yoram Singer

Abstract: Prediction sufﬁx trees (PST) provide a popular and effective tool for tasks such as compression, classiﬁcation, and language modeling. In this paper we take a decision theoretic view of PSTs for the task of sequence prediction. Generalizing the notion of margin to PSTs, we present an online PST learning algorithm and derive a loss bound for it. The depth of the PST generated by this algorithm scales linearly with the length of the input. We then describe a self-bounded enhancement of our learning algorithm which automatically grows a bounded-depth PST. We also prove an analogous mistake-bound for the self-bounded algorithm. The result is an efﬁcient algorithm that neither relies on a-priori assumptions on the shape or maximal depth of the target PST nor does it require any parameters. To our knowledge, this is the ﬁrst provably-correct PST learning algorithm which generates a bounded-depth PST while being competitive with any ﬁxed PST determined in hindsight. 1

5 0.87219959 16 nips-2004-Adaptive Discriminative Generative Model and Its Applications

Author: Ruei-sung Lin, David A. Ross, Jongwoo Lim, Ming-Hsuan Yang

Abstract: This paper presents an adaptive discriminative generative model that generalizes the conventional Fisher Linear Discriminant algorithm and renders a proper probabilistic interpretation. Within the context of object tracking, we aim to ﬁnd a discriminative generative model that best separates the target from the background. We present a computationally efﬁcient algorithm to constantly update this discriminative model as time progresses. While most tracking algorithms operate on the premise that the object appearance or ambient lighting condition does not signiﬁcantly change as time progresses, our method adapts a discriminative generative model to reﬂect appearance variation of the target and background, thereby facilitating the tracking task in ever-changing environments. Numerous experiments show that our method is able to learn a discriminative generative model for tracking target objects undergoing large pose and lighting changes.

6 0.87184381 133 nips-2004-Nonparametric Transforms of Graph Kernels for Semi-Supervised Learning

7 0.87178469 69 nips-2004-Fast Rates to Bayes for Kernel Machines

8 0.87118679 178 nips-2004-Support Vector Classification with Input Data Uncertainty

9 0.87112409 187 nips-2004-The Entire Regularization Path for the Support Vector Machine

10 0.87082565 31 nips-2004-Blind One-microphone Speech Separation: A Spectral Learning Approach

11 0.86954325 4 nips-2004-A Generalized Bradley-Terry Model: From Group Competition to Individual Skill

12 0.86855799 206 nips-2004-Worst-Case Analysis of Selective Sampling for Linear-Threshold Algorithms

13 0.86837023 131 nips-2004-Non-Local Manifold Tangent Learning

14 0.86788476 103 nips-2004-Limits of Spectral Clustering

15 0.86783689 45 nips-2004-Confidence Intervals for the Area Under the ROC Curve

16 0.86756516 110 nips-2004-Matrix Exponential Gradient Updates for On-line Learning and Bregman Projection

17 0.86723983 167 nips-2004-Semi-supervised Learning with Penalized Probabilistic Clustering

18 0.86667252 3 nips-2004-A Feature Selection Algorithm Based on the Global Minimization of a Generalization Error Bound

19 0.86527121 207 nips-2004-ℓ₀-norm Minimization for Basis Selection

20 0.86523783 2 nips-2004-A Direct Formulation for Sparse PCA Using Semidefinite Programming