jmlr jmlr2008 jmlr2008-87 knowledge-graph by maker-knowledge-mining

87 jmlr-2008-Stationary Features and Cat Detection

Source: pdf

Author: FranĂ§ois Fleuret, Donald Geman

Abstract: Most discriminative techniques for detecting instances from object categories in still images consist of looping over a partition of a pose space with dedicated binary classiÄ?Ĺš ers. The efÄ?Ĺš ciency of this strategy for a complex pose, that is, for Ä?Ĺš ne-grained descriptions, can be assessed by measuring the effect of sample size and pose resolution on accuracy and computation. Two conclusions emerge: (1) fragmenting the training data, which is inevitable in dealing with high in-class variation, severely reduces accuracy; (2) the computational cost at high resolution is prohibitive due to visiting a massive pose partition. To overcome data-fragmentation we propose a novel framework centered on pose-indexed features which assign a response to a pair consisting of an image and a pose, and are designed to be stationary: the probability distribution of the response is always the same if an object is actually present. Such features allow for efÄ?Ĺš cient, one-shot learning of pose-speciÄ?Ĺš c classiÄ?Ĺš ers. To avoid expensive scene processing, we arrange these classiÄ?Ĺš ers in a hierarchy based on nested partitions of the pose; as in previous work on coarse-to-Ä?Ĺš ne search, this allows for efÄ?Ĺš cient processing. The hierarchy is then Ă˘€?foldedĂ˘€? for training: all the classiÄ?Ĺš ers at each level are derived from one base predictor learned from all the data. The hierarchy is Ă˘€?unfoldedĂ˘€? for testing: parsing a scene amounts to examining increasingly Ä?Ĺš ner object descriptions only when there is sufÄ?Ĺš cient evidence for coarser ones. In this way, the detection results are equivalent to an exhaustive search at high resolution. We illustrate these ideas by detecting and localizing cats in highly cluttered greyscale scenes. Keywords: supervised learning, computer vision, image interpretation, cats, stationary features, hierarchical search

Reference: text

Summary: the most important sentenses genereted by tfidf model

sentIndex sentText sentNum sentScore

1 Charles Street Baltimore, MD 21218, USA Editor: Pietro Perona Abstract Most discriminative techniques for detecting instances from object categories in still images consist of looping over a partition of a pose space with dedicated binary classiÄ? [sent-4, score-0.889]

2 Ĺš ne-grained descriptions, can be assessed by measuring the effect of sample size and pose resolution on accuracy and computation. [sent-8, score-0.632]

3 Two conclusions emerge: (1) fragmenting the training data, which is inevitable in dealing with high in-class variation, severely reduces accuracy; (2) the computational cost at high resolution is prohibitive due to visiting a massive pose partition. [sent-9, score-0.671]

4 We want to do more than say whether or not there are objects in the scene; we want to provide a description of the pose of each detected instance, for example the locations of certain landmarks. [sent-33, score-0.662]

5 More generally, pose could refer to any properties of object instantiations which c 2008 Francois Fleuret and Donald Geman. [sent-34, score-0.664]

6 In particular, translating and scaling the training images to a reference pose allows for learning a base classiÄ? [sent-45, score-0.782]

7 Summarizing: aggregating the data avoids sparse training but at the expense of costly image transforms and restrictions on the pose; fragmenting the data can, in principle, accommodate a complex pose but at the expense of crippling performance due to impoverished training. [sent-59, score-0.721]

8 A related trade-off is the one between computation and pose resolution. [sent-60, score-0.57]

9 , higher pose resolution) allows for training a more discriminating classiÄ? [sent-64, score-0.635]

10 Ĺš ned the pose partitioning, the more online computation because regardless of how the classiÄ? [sent-67, score-0.57]

11 Ĺš ers corresponding to a recursive partitioning of the pose space, that is, parameterizations of increasing complexity. [sent-83, score-0.645]

12 The pose of the scissors could be the locations of the screw and the two tips, in which case one might measure the relative frequency a particular edge orientation inside in a disc whose radius and location, as well as the chosen orientation, depends on the pose. [sent-86, score-0.757]

13 If properly designed, the response statistics have a distribution which is invariant to the pose when in fact a pair of scissors is present (see Ă‚Â§ 3. [sent-87, score-0.612]

14 in the sense of assigning a numerical value to each combination of an image and a pose (or subset of poses). [sent-105, score-0.656]

15 The desired form of stationarity is that, for any given pose, the distribution of the responses of the features over images containing an object at that pose does not depend on the pose. [sent-106, score-0.855]

16 Said another way, if an image and an object instance at a given pose are selected, and only the responses of the stationary features are provided, one cannot guess the pose. [sent-107, score-0.961]

17 This is illustrated in Figure 1: knowing only the proportion of edges at a pose-dependent orientation in the indicated disk provides no information about the pose of the scissors. [sent-108, score-0.627]

18 Given that objects are present, a stationary feature evaluated at one pose is then the Ă˘€œsameĂ˘€? [sent-109, score-0.708]

19 Ĺš ers to train grows linearly, not exponentially, with the depth of the pose hierarchy. [sent-131, score-0.647]

20 detector is a coordinated search based on stationary features and a two-level hierarchy; the search for the belly location in the second-level is conditional on a pending head location and data fragmentation is avoided with pose-indexed features in a head-belly frame. [sent-146, score-1.276]

21 Ĺš c base features, pose hierarchy and pose-indexed features, to detecting cats in still images. [sent-155, score-0.924]

22 Related Work We characterize other work in relation to the two basic components of our detection strategy: explicit modeling of a hidden pose parameter, as in many generative and discriminative methods, and formulating detection as a controlled Ă˘€œprocess of discoveryĂ˘€? [sent-158, score-0.826]

23 The criterion for a true detection is that the estimates of the head location, head size and belly location all be close to the true pose (see Ă‚Â§ 6. [sent-177, score-1.765]

24 The H+B detector is built from separate head and body detectors while the HB detector is built upon pose indexed features (see Ă‚Â§ 6. [sent-179, score-1.332]

25 The extreme application of this conditioning paradigm is classical template matching (Grenander, 1993): if the pose is rich enough to account for all non-trivial statistical variation, then even a relatively simple metric can capture the remaining uncertainty, which is basically noise. [sent-184, score-0.57]

26 Ĺš ned with respect to the pose of the object to detect, in that case a series of control points. [sent-209, score-0.664]

27 2 A Process of Discovery We do not regard the hidden pose as a Ă˘€œnuisanceĂ˘€? [sent-212, score-0.599]

28 Hierarchical techniques, which can accomplish focusing, are based on a recursive partitioning of the pose space (or object/pose space), which can be either ad-hoc (Geman et al. [sent-220, score-0.595]

29 Ĺš ers, each one trained on a dedicated set of examplesĂ˘€”those carrying a pose in the corresponding cell of the hierarchy. [sent-224, score-0.696]

30 Our work is also related to early work on hierarchical template-matching (Gavrila, 1998) and hierarchical search of pose space using branch and bound algorithms (Huttenlocher and Rucklidge, 1993), and to the cascade of classiÄ? [sent-227, score-0.681]

31 Ĺš ce of selectivity if the pose space is coarsely explored and the sacriÄ? [sent-235, score-0.57]

32 K, let Yk be a Boolean random variable indicating whether or not there is a target in I with pose in Yk . [sent-254, score-0.57]

33 Ĺš xed size in a gray-scale image of size W Ä‚— H, natural choices would be I = [0, 1]W H and Y = [0,W ] Ä‚— [0, H], the image plane itself; that is, the pose reduces to one location. [sent-260, score-0.742]

34 If the desired detection accuracy were 5 pixels, then the pose cells might be disjoint 5 Ä‚— 5 blocks and K would be approximately W H . [sent-261, score-0.732]

35 On the other hand, if 25 the pose accommodated scale and multiple points of interest, then obviously the same accuracy in the prediction would lead to a far larger K, and any detection algorithm based on looping over pose cells would be highly costly. [sent-262, score-1.302]

36 We denote by T a training set of images labeled with the presences of targets T = I (t) , Y(t) 1Ă˘‰Â¤tĂ˘‰Â¤T , where each I (t) is a full image, and Y(t) is the Boolean vector indicating the pose cells occupied by targets in I (t) . [sent-263, score-0.739]

37 Ĺš‚ine and online and opens the way for dealing with complex pose spaces. [sent-280, score-0.57]

38 This corresponds to generating a single sample from each training scene, labeled according to whether or not there is a target with pose in Yk . [sent-287, score-0.609]

39 This is data-fragmentation: training 2555 F LEURET AND G EMAN Y , the pose space Y1 , . [sent-288, score-0.609]

40 , YK , a partition of the pose space Y Z , a W Ä‚— H pixel lattice I = {0, . [sent-291, score-0.597]

41 , 255}Z , a set of gray-scale images of size W Ä‚— H I, a random variable taking values in I Yk , a Boolean random variable indicating if there is a target in I with pose in Y k Y = (Y1 , . [sent-294, score-0.639]

42 , K} Ä‚— I Ă˘†’ I , an image transformation intended to normalize a given pose X : {1, . [sent-300, score-0.683]

43 corresponding to X(k, I) g : RQ Ă˘†’ {0, 1}, a predictor trained from all the data Table 1: Notation fk involves only those data which exactly satisfy the pose constraint; no synthesis or transformations are exploited to augment the number of samples available for training. [sent-305, score-0.656]

44 Ĺš ner the partitioning of the pose space Y , the fewer positive data points are available for training each f k . [sent-307, score-0.634]

45 The relation between the signal and the pose is obvious and normalizing the positive samples to a common reference pose by translating and scaling them is the natural procedure; only one classiÄ? [sent-309, score-1.274]

46 Unless 3D models are available, from which various views can be synthesized, the only course of action is data-fragmentation: partition the pose space into several cells corresponding to different orientation ranges and train a dedicated, range-speciÄ? [sent-313, score-0.742]

47 Consider for instance a family of edge detectors and consider again a pose consisting of a single location z. [sent-325, score-0.784]

48 In such a case, the transformation ÄŽˆ applies a translation to the image to move the center of pose cell Y k to a reference location. [sent-326, score-0.725]

49 This can be summarized algorithmically as follows: In order to predict if there is a target in image I with pose in Yk , Ä? [sent-334, score-0.656]

50 Ĺš rst normalize the image with ÄŽˆ so that a target with pose in Yk would be moved to a reference pose cell, then extract features in that transformed image using ĂŽĹž, and Ä? [sent-335, score-1.441]

51 Rotating the image does not allow for normalizing the body orientation without changing the head orientation, and designing a non-afÄ? [sent-342, score-0.558]

52 Instead, we propose a different mechanism for data-aggregation based on pose-indexed features which directly assign a response to a pair consisting of an image and a pose cell and which satisfy a stationarity requirement. [sent-345, score-0.86]

53 Ĺš cient training by normalizing the pose at the image level is difÄ? [sent-357, score-0.735]

54 The ordering matters because we want (1) to hold and hence there is a correspondence among individual components of X(k) from one pose cell to another. [sent-371, score-0.61]

55 Ĺš rst applying a normalizing mapping ÄŽˆ to transform I in accordance with a pose cell k, and then evaluating the base features, we directly compute the feature responses as functions of both the image and the pose cell. [sent-383, score-1.351]

56 The pose space is Y = (ĂŽÂ¸1 , ĂŽÂ¸2 ) Ă˘ˆˆ {1, . [sent-415, score-0.57]

57 For simplicity, assume there is at most one object instance, so we can just write Y = (ĂŽÂ¸ 1 , ĂŽÂ¸2 ) Ă˘ˆˆ Y to denote an instance with pose (ĂŽÂ¸1 , ĂŽÂ¸2 ). [sent-420, score-0.664]

58 2565 F LEURET AND G EMAN Figure 9: From the centroid of any pose cell, we deÄ? [sent-432, score-0.57]

59 3, a pose-indexed feature is a real-valued function of both a pose cell and an image. [sent-444, score-0.61]

60 Since the response of any of them depends on counting certain edge types over rectangular windows in the image, we construct our family of pose-indexed features indirectly by indexing both the edge types and the window locations with the pose cell. [sent-446, score-0.94]

61 For any pose cell index k, we compute the average head location h = h k , the average belly location b = bk , and the average head radius r = rk of the pose cell Yk . [sent-452, score-2.388]

62 The three types of frames are oriented according to the relative horizontal locations of the head and belly of the cat, so a reÄ? [sent-462, score-0.811]

63 Windows relative to the head or the belly frame are simply translated and scaled accordingly. [sent-467, score-0.791]

64 Hence the full pose space is Y Ă˘Š‚ Z Ä‚— R + Ä‚— Z . [sent-482, score-0.57]

65 For any such cell, the admissible domain for the belly locations is the convex envelope of the belly locations seen in the training examples, normalized in location and scale with respect to the head location and radius. [sent-492, score-1.361]

66 There are 500 such belly squares, hence the total number of pose cells in the second level is K2 2. [sent-497, score-0.951]

67 Ĺš er g (1) is evaluated over a sublattice of possible head locations and all alarms above a very low threshold are retained. [sent-506, score-0.662]

68 Ĺš er g(2) is evaluated for each pair of head-belly locations on a sublattice consistent with the retained head alarms and with observed statistics about joint head-belly locations. [sent-508, score-0.662]

69 For clarity, the depicted discretization of the pose space is idealized, and far coarser than in the actual experiments; for an image of size 640 Ä‚— 480 pixels, we consider 50, 000 head pose cells and 2. [sent-509, score-1.637]

70 5 Detectors Whereas our aim is to detect both the head and the body, detecting the head alone is similar to the well-studied problem of detecting frontal views of human faces. [sent-524, score-0.794]

71 As stated earlier, if the pose reduces to a single position, data-aggregation is straightforward by translating either whole images or features. [sent-525, score-0.669]

72 Ĺš nd cats since the head is clearly the most stable landmark and the part of the cat with the least variation, assuming of course that the head is visible, which is the case with our data (for the same reason that family photographs display the faces of people). [sent-528, score-0.971]

73 Since that second-level detector is designed not to exploit the information in the joint locations of the head and belly, the frames here have Ä? [sent-547, score-0.597]

74 Ĺš‚ecting the cat pose horizontally would move but not invert the frames. [sent-549, score-0.731]

75 Ĺš ned relatively to any of the three reference frames (head, belly or head-belly) in order to take into account the position of the head in searching for the belly. [sent-554, score-0.748]

76 25; and (2) Either the head or belly locations are close: min( hĂ˘ˆ’h , bĂ˘ˆ’b ) Ă˘‰Â¤ 0. [sent-561, score-0.762]

77 5; (2) Ă˘ˆš the head locations are nearby each other: h Ă˘ˆ’ h Ă˘‰Â¤ 2 rr ; and (3) the belly locations are nearby Ă˘ˆš each other: b Ă˘ˆ’ b Ă˘‰Â¤ 4 rr . [sent-565, score-0.854]

78 Ĺš gure, the distance between the two head locations is less than the average head radius, and if the distance between the belly locations is less than twice the average head radius. [sent-568, score-1.554]

79 Based on that criterion, if the true pose is the one shown in thin lines and the thick poses are detections, only the leftmost one would be counted as a true hit. [sent-571, score-0.656]

80 An alarm is counted as true positive if there exists a cat in the image with a similar pose according to the criterion described above. [sent-581, score-0.817]

81 Ĺš er exploit the full pose can be seen in Figure 11. [sent-593, score-0.61]

82 For instance, a false head detection lying around or on the belly will be supported by the second-level classiÄ? [sent-599, score-0.816]

83 Moreover, combining simultaneous training with a sequential exploration of the pose space overcomes the main drawback of previous coarse-to-Ä? [sent-635, score-0.609]

84 Ĺš rst visiting potential head locations alone and then examining additional aspects of the pose consistent with and informed by candidate head locations. [sent-649, score-1.362]

85 The circle shows the estimated head size and location, and the dot the estimated belly location. [sent-658, score-0.67]

86 The circle shows the estimated head size and location, and the dot the estimated belly location. [sent-660, score-0.67]

87 The crosses depict the head and belly centers provided on the training data. [sent-684, score-0.709]

88 1 Settings Since training with fragmentation is not feasible for any complete partition of a complex pose space at a realistic resolution, the images we consider in these experiments have been cropped from the original data set so that the pose space Y is already strongly constrained. [sent-691, score-1.343]

89 Ĺš rst series of experiments the target pose is the center of the cat head, constrained to Y = [Ă˘ˆ’20, 20] Ä‚— [Ă˘ˆ’20, 20] in a 100 Ä‚— 100 image. [sent-693, score-0.731]

90 It is this pose space that will be investigated at different resolutions. [sent-694, score-0.57]

91 In the second series of experiments the pose is the pair of locations (h, b) for the head and belly, constrained to Y = ([0, 5] Ä‚— [0, 5]) Ä‚— ([Ă˘ˆ’80, 80] Ä‚— [Ă˘ˆ’20, 80]) in a 200 Ä‚— 140 image centered at the square. [sent-696, score-1.128]

92 More precisely, we consider three partitions of Y into K = 1, 4 and 16 pose cells. [sent-700, score-0.57]

93 Three of them are trained under data-fragmentation at the three considered resolutions, namely K = 1, K = 4 or K = 16 pose cells. [sent-702, score-0.599]

94 The stationary features are based on the head frame alone for the head experiments, and on both the head frame and the head-belly frame for the head-belly experiments. [sent-706, score-1.624]

95 This is not true for the head-belly experiments, where sixteen pose cells do worse than four, with or without cost equalization, which can be explained to some extent by the lower variation in the appearance of cat heads than full cat bodies, and hence fewer samples may be sufÄ? [sent-720, score-1.06]

96 Obtaining an accuracy in the locations of the head and the belly of Ä? [sent-732, score-0.762]

97 Ĺš ve pixels requires more than 7 Ä‚— 106 pose cells. [sent-733, score-0.57]

98 The need for data-aggregation: Dealing with a rich pose by training specialized predictors from constrained sub-populations is not feasible, both in terms of ofÄ? [sent-736, score-0.609]

99 Ĺš ers, the computation necessary to cover a partition of a pose space of reasonable accuracy is not realistic if the effort is uniformly distributed over cells. [sent-742, score-0.597]

100 Learning methods for generic object recognition with invariance to pose and lighting. [sent-830, score-0.692]

similar papers computed by tfidf model

tfidf for this paper:

wordName wordTfidf (topN-words)

[('pose', 0.57), ('head', 0.35), ('belly', 0.32), ('cat', 0.161), ('hierarchy', 0.152), ('alarms', 0.14), ('stationary', 0.138), ('frame', 0.121), ('eman', 0.12), ('leuret', 0.12), ('tationary', 0.12), ('cats', 0.11), ('detector', 0.106), ('etection', 0.102), ('detectors', 0.102), ('detection', 0.101), ('yk', 0.094), ('object', 0.094), ('locations', 0.092), ('geman', 0.091), ('image', 0.086), ('poses', 0.086), ('scene', 0.085), ('hb', 0.077), ('location', 0.074), ('features', 0.073), ('eatures', 0.073), ('folded', 0.07), ('images', 0.069), ('fleuret', 0.068), ('fragmentation', 0.068), ('resolution', 0.062), ('cells', 0.061), ('dedicated', 0.057), ('orientation', 0.057), ('huttenlocher', 0.051), ('ers', 0.05), ('stationarity', 0.049), ('frames', 0.049), ('detecting', 0.047), ('false', 0.045), ('base', 0.045), ('hierarchical', 0.044), ('heads', 0.042), ('viola', 0.042), ('response', 0.042), ('vision', 0.041), ('er', 0.04), ('normalizing', 0.04), ('sublattice', 0.04), ('xh', 0.04), ('cell', 0.04), ('training', 0.039), ('variation', 0.039), ('classi', 0.038), ('edge', 0.038), ('windows', 0.036), ('signal', 0.035), ('parsing', 0.035), ('scenes', 0.032), ('orientations', 0.031), ('yh', 0.03), ('translating', 0.03), ('cascades', 0.03), ('crandall', 0.03), ('detections', 0.03), ('ratemykitten', 0.03), ('stenger', 0.03), ('centered', 0.03), ('fk', 0.03), ('reference', 0.029), ('trained', 0.029), ('rq', 0.029), ('hidden', 0.029), ('invariance', 0.028), ('predictor', 0.027), ('normalize', 0.027), ('partition', 0.027), ('geometric', 0.027), ('face', 0.027), ('train', 0.027), ('window', 0.026), ('discriminating', 0.026), ('appearance', 0.026), ('aggregating', 0.026), ('aggregated', 0.026), ('fragmented', 0.025), ('dilemma', 0.025), ('indexing', 0.025), ('resolutions', 0.025), ('discriminative', 0.025), ('body', 0.025), ('partitioning', 0.025), ('strategy', 0.024), ('roc', 0.024), ('jones', 0.023), ('un', 0.023), ('cascade', 0.023), ('radii', 0.023), ('parts', 0.022)]

similar papers list:

simIndex simValue paperId paperTitle

same-paper 1 0.99999917 87 jmlr-2008-Stationary Features and Cat Detection

Author: FranĂ§ois Fleuret, Donald Geman

2 0.072883487 54 jmlr-2008-Learning to Select Features using their Properties

Author: Eyal Krupka, Amir Navot, Naftali Tishby

Abstract: Feature selection is the task of choosing a small subset of features that is sufﬁcient to predict the target labels well. Here, instead of trying to directly determine which features are better, we attempt to learn the properties of good features. For this purpose we assume that each feature is represented by a set of properties, referred to as meta-features. This approach enables prediction of the quality of features without measuring their value on the training instances. We use this ability to devise new selection algorithms that can efﬁciently search for new good features in the presence of a huge number of features, and to dramatically reduce the number of feature measurements needed. We demonstrate our algorithms on a handwritten digit recognition problem and a visual object category recognition problem. In addition, we show how this novel viewpoint enables derivation of better generalization bounds for the joint learning problem of selection and classiﬁcation, and how it contributes to a better understanding of the problem. Speciﬁcally, in the context of object recognition, previous works showed that it is possible to ﬁnd one set of features which ﬁts most object categories (aka a universal dictionary). Here we use our framework to analyze one such universal dictionary and ﬁnd that the quality of features in this dictionary can be predicted accurately by its meta-features. Keywords: feature selection, unobserved features, meta-features

3 0.068297789 31 jmlr-2008-Dynamic Hierarchical Markov Random Fields for Integrated Web Data Extraction

Author: Jun Zhu, Zaiqing Nie, Bo Zhang, Ji-Rong Wen

Abstract: Existing template-independent web data extraction approaches adopt highly ineffective decoupled strategies—attempting to do data record detection and attribute labeling in two separate phases. In this paper, we propose an integrated web data extraction paradigm with hierarchical models. The proposed model is called Dynamic Hierarchical Markov Random Fields (DHMRFs). DHMRFs take structural uncertainty into consideration and deﬁne a joint distribution of both model structure and class labels. The joint distribution is an exponential family distribution. As a conditional model, DHMRFs relax the independence assumption as made in directed models. Since exact inference is intractable, a variational method is developed to learn the model’s parameters and to ﬁnd the MAP model structure and label assignments. We apply DHMRFs to a real-world web data extraction task. Experimental results show that: (1) integrated web data extraction models can achieve signiﬁcant improvements on both record detection and attribute labeling compared to decoupled models; (2) in diverse web data extraction DHMRFs can potentially address the blocky artifact issue which is suffered by ﬁxed-structured hierarchical models. Keywords: conditional random ﬁelds, dynamic hierarchical Markov random ﬁelds, integrated web data extraction, statistical hierarchical modeling, blocky artifact issue

4 0.067274235 58 jmlr-2008-Max-margin Classification of Data with Absent Features

Author: Gal Chechik, Geremy Heitz, Gal Elidan, Pieter Abbeel, Daphne Koller

Abstract: We consider the problem of learning classiﬁers in structured domains, where some objects have a subset of features that are inherently absent due to complex relationships between the features. Unlike the case where a feature exists but its value is not observed, here we focus on the case where a feature may not even exist (structurally absent) for some of the samples. The common approach for handling missing features in discriminative models is to ﬁrst complete their unknown values, and then use a standard classiﬁcation procedure over the completed data. This paper focuses on features that are known to be non-existing, rather than have an unknown value. We show how incomplete data can be classiﬁed directly without any completion of the missing features using a max-margin learning framework. We formulate an objective function, based on the geometric interpretation of the margin, that aims to maximize the margin of each sample in its own relevant subspace. In this formulation, the linearly separable case can be transformed into a binary search over a series of second order cone programs (SOCP), a convex problem that can be solved efﬁciently. We also describe two approaches for optimizing the general case: an approximation that can be solved as a standard quadratic program (QP) and an iterative approach for solving the exact problem. By avoiding the pre-processing phase in which the data is completed, both of these approaches could offer considerable computational savings. More importantly, we show that the elegant handling of missing values by our approach allows it to both outperform other methods when the missing values have non-trivial structure, and be competitive with other methods when the values are missing at random. We demonstrate our results on several standard benchmarks and two real-world problems: edge prediction in metabolic pathways, and automobile detection in natural images. Keywords: max margin, missing features, network reconstruction, metabolic pa

5 0.055690553 51 jmlr-2008-Learning Similarity with Operator-valued Large-margin Classifiers

Author: Andreas Maurer

Abstract: A method is introduced to learn and represent similarity with linear operators in kernel induced Hilbert spaces. Transferring error bounds for vector valued large-margin classiﬁers to the setting of Hilbert-Schmidt operators leads to dimension free bounds on a risk functional for linear representations and motivates a regularized objective functional. Minimization of this objective is effected by a simple technique of stochastic gradient descent. The resulting representations are tested on transfer problems in image processing, involving plane and spatial geometric invariants, handwritten characters and face recognition. Keywords: learning similarity, similarity, transfer learning

6 0.03833333 29 jmlr-2008-Cross-Validation Optimization for Large Scale Structured Classification Kernel Methods

7 0.037703164 13 jmlr-2008-An Error Bound Based on a Worst Likely Assignment

8 0.035942297 73 jmlr-2008-On the Suitable Domain for SVM Training in Image Coding

9 0.03435589 81 jmlr-2008-Regularization on Graphs with Function-adapted Diffusion Processes

10 0.034306746 94 jmlr-2008-Value Function Approximation using Multiple Aggregation for Multiattribute Resource Management

11 0.033924915 67 jmlr-2008-Near-Optimal Sensor Placements in Gaussian Processes: Theory, Efficient Algorithms and Empirical Studies

12 0.033302028 59 jmlr-2008-Maximal Causes for Non-linear Component Extraction

13 0.032908674 25 jmlr-2008-Consistency of Random Forests and Other Averaging Classifiers

14 0.031812269 41 jmlr-2008-Graphical Models for Structured Classification, with an Application to Interpreting Images of Protein Subcellular Location Patterns

15 0.031567927 96 jmlr-2008-Visualizing Data using t-SNE

16 0.030464483 17 jmlr-2008-Automatic PCA Dimension Selection for High Dimensional Data and Small Sample Sizes

17 0.030249869 11 jmlr-2008-Aggregation of SVM Classifiers Using Sobolev Spaces

18 0.029758839 55 jmlr-2008-Linear-Time Computation of Similarity Measures for Sequential Data

19 0.029550286 56 jmlr-2008-Magic Moments for Structured Output Prediction

20 0.02936212 3 jmlr-2008-A Moment Bound for Multi-hinge Classifiers

similar papers computed by lsi model

lsi for this paper:

topicId topicWeight

[(0, 0.155), (1, -0.019), (2, 0.037), (3, -0.042), (4, -0.139), (5, 0.058), (6, 0.096), (7, 0.03), (8, 0.061), (9, -0.03), (10, -0.063), (11, 0.118), (12, -0.035), (13, -0.023), (14, -0.028), (15, -0.004), (16, -0.04), (17, -0.016), (18, -0.056), (19, 0.069), (20, 0.15), (21, 0.04), (22, 0.069), (23, -0.124), (24, 0.078), (25, -0.148), (26, -0.019), (27, -0.208), (28, 0.119), (29, 0.019), (30, 0.103), (31, 0.224), (32, -0.027), (33, -0.21), (34, 0.004), (35, -0.099), (36, -0.183), (37, 0.361), (38, 0.037), (39, 0.005), (40, 0.166), (41, 0.006), (42, 0.014), (43, 0.058), (44, 0.341), (45, 0.083), (46, -0.116), (47, 0.038), (48, -0.293), (49, 0.085)]

similar papers list:

simIndex simValue paperId paperTitle

same-paper 1 0.9599607 87 jmlr-2008-Stationary Features and Cat Detection

Author: FranĂ§ois Fleuret, Donald Geman

2 0.37186375 31 jmlr-2008-Dynamic Hierarchical Markov Random Fields for Integrated Web Data Extraction

Author: Jun Zhu, Zaiqing Nie, Bo Zhang, Ji-Rong Wen

3 0.32339519 54 jmlr-2008-Learning to Select Features using their Properties

Author: Eyal Krupka, Amir Navot, Naftali Tishby

4 0.25978944 73 jmlr-2008-On the Suitable Domain for SVM Training in Image Coding

Author: Gustavo Camps-Valls, Juan Gutiérrez, Gabriel Gómez-Pérez, Jesús Malo

Abstract: Conventional SVM-based image coding methods are founded on independently restricting the distortion in every image coefﬁcient at some particular image representation. Geometrically, this implies allowing arbitrary signal distortions in an n-dimensional rectangle deﬁned by the ε-insensitivity zone in each dimension of the selected image representation domain. Unfortunately, not every image representation domain is well-suited for such a simple, scalar-wise, approach because statistical and/or perceptual interactions between the coefﬁcients may exist. These interactions imply that scalar approaches may induce distortions that do not follow the image statistics and/or are perceptually annoying. Taking into account these relations would imply using non-rectangular εinsensitivity regions (allowing coupled distortions in different coefﬁcients), which is beyond the conventional SVM formulation. In this paper, we report a condition on the suitable domain for developing efﬁcient SVM image coding schemes. We analytically demonstrate that no linear domain fulﬁlls this condition because of the statistical and perceptual inter-coefﬁcient relations that exist in these domains. This theoretical result is experimentally conﬁrmed by comparing SVM learning in previously reported linear domains and in a recently proposed non-linear perceptual domain that simultaneously reduces the statistical and perceptual relations (so it is closer to fulﬁlling the proposed condition). These results highlight the relevance of an appropriate choice of the image representation before SVM learning. Keywords: image coding, non-linear perception models, statistical independence, support vector machines, insensitivity zone c 2008 Gustavo Camps-Valls, Juan Guti´ rrez, Gabriel G´ mez-P´ rez and Jes´ s Malo. e o e u ´ ´ ´ C AMPS -VALLS , G UTI E RREZ , G OMEZ -P E REZ AND M ALO 1. Problem Statement: The Diagonal Jacobian Condition Image coding schemes based on support vector machines (SVM) have been successfully introduced in the literature. SVMs have been used in the spatial domain (Robinson and Kecman, 2000), in the block-DCT domain (Robinson and Kecman, 2003), and in the wavelet domain (Ahmed, 2005; Jiao et al., 2005). These coding methods take advantage of the ability of the support vector regression (SVR) algorithm for function approximation using a small number of parameters (signal samples, or ¨ support vectors) (Smola and Scholkopf, 2004). In all current SVM-based image coding techniques, a representation of the image is described by the entropy-coded weights associated to the support vectors necessary to approximate the signal with a given accuracy. Relaxing the accuracy bounds reduces the number of needed support vectors. In a given representation domain, reducing the number of support vectors increases the compression ratio at the expense of bigger distortion (lower image quality). By applying the standard SVR formulation, a certain amount of distortion in each sample of the image representation is allowed. In the original formulation, scalar restrictions on the errors are introduced using a constant ε-insensitivity value for every sample. ´ Recently, this procedure has been reﬁned by Gomez-P´ rez et al. (2005) using a proﬁle-dependent e SVR (Camps-Valls et al., 2001) that considers a different ε for each sample or frequency. This frequency-dependent insensitivity, ε f , accounts for the fact that, according to simple (linear) perception models, not every sample in linear frequency domains (such as DCT or wavelets) contributes to the perceived distortion in the same way. Despite different domains have been proposed for SVM training (spatial domain, block-DCT and wavelets) and different ε insensitivities per sample have been proposed, in conventional SVR formulation, the particular distortions introduced by regression in the different samples are not coupled. In all the reported SVM-based image coding schemes, the RBF kernel is used and the penalization parameter is ﬁxed to an arbitrarily large value. In this setting, considering n-sample signals as n-dimensional vectors, the SVR guarantees that the approximated vectors are conﬁned in n-dimensional rectangles around the original vectors. These rectangles are just n-dimensional cubes in the standard formulation or they have certain elongation if different ε f are considered in each axis, f . Therefore, in all the reported SVM-based coding methods, these rectangles are always oriented along the axes of the (linear) image representation. According to this, a common feature of these (scalar-wise) approaches is that they give rise to decoupled distortions in each dimension. P´ rez-Cruz et al. (2002) proposed a hyperspherical insensitivity zone to correct the penalization e factor in each dimension of multi-output regression problems, but again, restrictions to each sample were still uncoupled. This scalar-wise strategy is not the best option in domains where the different dimensions of the image representation are not independent. For instance, consider the situation where actually independent components, r f , are obtained from a given image representation, y, applying some eventually non-linear transform, R: R y −→ r. In this case, SVM regression with scalar-wise error restriction makes sense in the r domain. However, the original y domain will not be suitable for the standard SVM regression unless the matrix ∇R is diagonal (up to any permutation of the dimensions, that is, only one non-zero element per row). Therefore, if transforms that achieve independence have non-diagonal Jacobian, scalar-wise restrictions in the original (coupled coefﬁcients) domain y are not allowed. 50 O N THE S UITABLE D OMAIN FOR SVM T RAINING IN I MAGE C ODING y2 1 −0,5 −0,5 0,4 R= r2 y1 r1 R−1 = 0,83 −0,83 0,67 −1,67 Figure 1: Insensitivity regions in different representation domains, y (left) and r (right), related by a non-diagonal transform ∇R and its inverse ∇R−1 . Figure 1 illustrates this situation. The shaded region in the right plot (r domain) represents the n-dimensional box determined by the ε f insensitivities in each dimension ( f =1,2), in which a scalar-wise approach is appropriate due to independence among signal coefﬁcients. Given that the particular ∇R transform is not diagonal, the corresponding shaded region in the left plot (the original y domain) is not aligned along the axes of the representation. This has negative implications: note that for the highlighted points, smaller distortions in both dimensions in the y domain (as implied by SVM with tighter but scalar ε f insensitivities) do not necessarily imply lying inside the insensitivity region in the ﬁnal truly independent (and meaningful) r domain. Therefore, the original y domain is not suitable for the direct application of conventional SVM, and consequently, non-trivial coupled insensitivity regions are required. Summarizing, in the image coding context, the condition for an image representation y to be strictly suitable for conventional SVM learning is that the transform that maps the original representation y to an independent coefﬁcient representation r must be locally diagonal. As will be reviewed below, independence among coefﬁcients (and the transforms to obtain them) may be deﬁned in both statistical and perceptual terms (Hyvarinen et al., 2001; Malo et al., 2001; Epifanio et al., 2003; Malo et al., 2006). On the one hand, a locally diagonal relation to a statistically independent representation is desirable because independently induced distortions (as the conventional SVM approach does) will preserve the statistics of the distorted signal, that is, it will not introduce artiﬁcial-looking artifacts. On the other hand, a locally diagonal relation to a perceptually 51 ´ ´ ´ C AMPS -VALLS , G UTI E RREZ , G OMEZ -P E REZ AND M ALO independent representation is desirable because independently induced distortions do not give rise to increased subjective distortions due to non-trivial masking or facilitation interactions between the distortions in each dimension (Watson and Solomon, 1997). In this work, we show that conventional linear domains do not fulﬁll the diagonal Jacobian condition in either the statistical case or in the perceptual case. This theoretical result is experimentally conﬁrmed by comparing SVM learning in previously reported linear domains (Robinson and Kecman, 2003; G´ mez-P´ rez et al., 2005) and in a recently proposed non-linear perceptual domain o e that simultaneously reduces the statistical and the perceptual relations (Malo et al., 2006), thus, this non-linear perceptual domain is closer to fulﬁlling the proposed condition. The rest of the paper is structured as follows. Section 2 reviews the fact that linear coefﬁcients of the image representations commonly used for SVM training are neither statistically independent nor perceptually independent. Section 3 shows that transforms for obtaining statistical and/or perceptual independence from linear domains have non-diagonal Jacobian. This suggests that there is room to improve the performance of conventional SVM learning reported in linear domains. In Section 4, we propose the use of a perceptual representation for SVM training because it strictly fulﬁlls the diagonal Jacobian condition in the perceptual sense and increases the statistical independence among coefﬁcients, bringing it closer to fulﬁlling the condition in the statistical sense. The experimental image coding results conﬁrm the superiority of this domain for SVM training in Section 5. Section 6 presents the conclusions and ﬁnal remarks. 2. Statistical and Perceptual Relations Among Image Coefﬁcients Statistical independence among the coefﬁcients of a signal representation refers to the fact that the joint PDF of the class of signals to be considered can be expressed as a product of the marginal PDFs in each dimension (Hyvarinen et al., 2001). Simple (second-order) descriptions of statistical dependence use the non-diagonal nature of the covariance matrix (Clarke, 1985; Gersho and Gray, 1992). More recent and accurate descriptions use higher-order moments, mutual information, or the non-Gaussian nature (sparsity) of marginal PDFs (Hyvarinen et al., 2001; Simoncelli, 1997). Perceptual independence refers to the fact that the visibility of errors in coefﬁcients of an image may depend on the energy of neighboring coefﬁcients, a phenomenon known in the perceptual literature as masking or facilitation (Watson and Solomon, 1997). Perceptual dependence has been formalized just up to second order, and this may be described by the non-Euclidean nature of the perceptual metric matrix (Malo et al., 2001; Epifanio et al., 2003; Malo et al., 2006). 2.1 Statistical Relations In recent years, a variety of approaches, known collectively as “independent component analysis” (ICA), have been developed to exploit higher-order statistics for the purpose of achieving a unique linear solution for coefﬁcient independence (Hyvarinen et al., 2001). The basis functions obtained when these methods are applied to images are spatially localized and selective for both orientation and spatial frequency (Olshausen and Field, 1996; Bell and Sejnowski, 1997). Thus, they are similar to basis functions of multi-scale wavelet representations. Despite its name, linear ICA does not actually produce statistically independent coefﬁcients when applied to photographic images. Intuitively, independence would seem unlikely, since images are not formed from linear superpositions of independent patterns: the typical combination rule for the elements of an image is occlusion. Empirically, the coefﬁcients of natural image decom52 O N THE S UITABLE D OMAIN FOR SVM T RAINING IN I MAGE C ODING | f | = 10.8 cpd | f | = 24.4 cpd −20 −10 −10 fy (cpd) −30 −20 fy (cpd) −30 0 0 10 10 20 20 30 30 −30 −20 −10 0 f (cpd) 10 20 30 −30 x −20 −10 0 f (cpd) 10 20 30 x Figure 2: Statistical interaction of two particular coefﬁcients of the local Fourier Transform with their neighbors in a natural image database. The absolute value of the frequency of these coefﬁcients is | f | = 10.8 and | f | = 24.4 cycles/degree (cpd). positions in spatially localized oscillating basis functions are found to be fairly well decorrelated (i.e., their covariance is almost zero). However, the amplitudes of coefﬁcients at nearby spatial positions, orientations, and scales are highly correlated (even with orthonormal transforms) (Simoncelli, ´ 1997; Buccigrossi and Simoncelli, 1999; Wainwright et al., 2001; Hyvarinen et al., 2003; Guti errez et al., 2006; Malo et al., 2006; Malo and Guti´ rrez, 2006). This suggests that achieving statistical e independence requires the introduction of non-linearities beyond linear ICA transforms. Figure 2 reproduces one of many results that highlight the presence of statistical relations of natural image coefﬁcients in block PCA or linear ICA-like domains: the energy of spatially localized oscillating ﬁlters is correlated with the energy of neighboring ﬁlters in scale and orientation (see Guti´ rrez et al., 2006). A remarkable feature is that the interaction width increases with frequency, e as has been reported in other domains, for example, wavelets (Buccigrossi and Simoncelli, 1999; Wainwright et al., 2001; Hyvarinen et al., 2003), and block-DCT (Malo et al., 2006). In order to remove the remaining statistical relations in the linear domains y, non-linear ICA methods are necessary (Hyvarinen et al., 2001; Lin, 1999; Karhunen et al., 2000; Jutten and Karhunen, 2003). Without lack of generality, non-linear ICA transforms can be schematically understood as a two-stage process (Malo and Guti´ rrez, 2006): e T R (( y hh x hh (( r, (1) R−1 T−1 where x is the image representation in the spatial domain, and T is a global unitary linear transform that removes second-order and eventually higher-order relations among coefﬁcients in the spatial domain. Particular examples of T include block PCA, linear ICAs, DCT or wavelets. In the ICA literature notation, T is the separating matrix and T−1 is the mixing matrix. The second transform 53 ´ ´ ´ C AMPS -VALLS , G UTI E RREZ , G OMEZ -P E REZ AND M ALO R is an additional non-linearity that is introduced in order to remove the statistical relations that still remain in the y domain. 2.2 Perceptual Relations Perceptual dependence among coefﬁcients in different image representations can be understood by using the current model of V1 cortex. This model can also be summarized by the two-stage (linear and non-linear) process described in Equation (1). In this perceptual case, T is also a linear ﬁlter bank applied to the original input image in the spatial domain. This ﬁlter bank represents the linear behavior of V1 neurons whose receptive ﬁelds happen to be similar to wavelets or linear ICA basis functions (Olshausen and Field, 1996; Bell and Sejnowski, 1997). The second transform, R, is a non-linear function that accounts for the masking and facilitation phenomena that have been reported in the linear y domain (Foley, 1994; Watson and Solomon, 1997). Section 3.2 gives a parametric expression for the second non-linear stage, R: the divisive normalization model (Heeger, 1992; Foley, 1994; Watson and Solomon, 1997). This class of models is based on psychophysical experiments assuming that the last domain, r, is perceptually Euclidean (i.e., perfect perceptual independence). An additional conﬁrmation of this assumption is the success of (Euclidean) subjective image distortion measures deﬁned in that domain (Teo and Heeger, 1994). Straightforward application of Riemannian geometry to obtain the perceptual metric matrix in other domains shows that the coefﬁcients of linear domains x and y, or any other linear transform of them, are not perceptually independent (Epifanio et al., 2003). Figure 3 illustrates the presence of perceptual relations between coefﬁcients when using linear block frequency or wavelet-like domains, y: the cross-masking behavior. In this example, the visibility of the distortions added on top of the background image made of periodic patterns has to be assessed. This is a measure of the sensitivity of a particular perceptual mechanism to distortions in that dimension, ∆y f , when mechanisms tuned to other dimensions are simultaneously active, that is, y f = 0, with f = f . As can be observed, low frequency noise is more visible in high frequency backgrounds than in low frequency backgrounds (e.g., left image). Similarly, high frequency noise is more visible in low frequency backgrounds than in high frequency ones (e.g., right image). That is to say, a signal of a speciﬁc frequency strongly masks the corresponding frequency analyzer, but it induces a smaller sensitivity reduction in the analyzers that are tuned to different frequencies. In other words, the reduction in sensitivity of a speciﬁc analyzer gets larger as the distance between the background frequency and the frequency of the analyzer gets smaller. The response of each frequency analyzer not only depends on the energy of the signal for that frequency band, but also on the energy of the signal in other frequency bands (cross-masking). This implies that a different amount of noise in each frequency band may be acceptable depending on the energy of that frequency band and on the energy of neighboring bands. This is what we have called perceptual dependence among different coefﬁcients in the y domain. At this point, it is important to stress the similarity between the set of computations to obtain statistically decoupled image coefﬁcients and the known stages of biological vision. In fact, it has been hypothesized that biological visual systems have organized their sensors to exploit the particular statistics of the signals they have to process. See Barlow (2001), Simoncelli and Olshausen (2001), and Simoncelli (2003) for reviews on this hypothesis. In particular, both the linear and the non-linear stages of the cortical processing have been successfully derived using redundancy reduction arguments: nowadays, the same class of linear 54 O N THE S UITABLE D OMAIN FOR SVM T RAINING IN I MAGE C ODING 3 cpd 6 cpd 12 cpd 24 cpd Figure 3: Illustrative example of perceptual dependence (cross-masking phenomenon). Equal energy noise of different frequency content, 3 cycl/deg (cpd), 6 cpd, 12 cpd and 24 cpd, shown on top of a background image. Sampling frequency assumes that these images subtend an angle of 3 deg. stage T is used in transform coding algorithms and in vision models (Olshausen and Field, 1996; Bell and Sejnowski, 1997; Taubman and Marcellin, 2001), and new evidence supports the same idea for the second non-linear stage (Schwartz and Simoncelli, 2001; Malo and Guti´ rrez, 2006). e According to this, the statistical and perceptual transforms, R, that remove the above relations from the linear domains, y, would be very similar if not the same. 3. Statistical and Perceptual Independence Imply Non-diagonal Jacobian In this section, we show that both statistical redundancy reduction transforms (e.g., non-linear ICA) and perceptual independence transforms (e.g., divisive normalization), have non-diagonal Jacobian for any linear image representation, so they are not strictly suitable for conventional SVM training. 3.1 Non-diagonal Jacobian in Non-linear ICA Transforms One possible approach for dealing with global non-linear ICA is to act differentially by breaking the problem into local linear pieces that can then be integrated to obtain the global independent coefﬁcient domain (Malo and Guti´ rrez, 2006). Each differential sub-problem around a particular e point (image) can be locally solved using the standard linear ICA methods restricted to the neighbors of that point (Lin, 1999). Using the differential approach in the context of a two-stage process such as the one in Equation (1), it can be shown that (Malo and Guti´ rrez, 2006): e r = r0 + Z x x0 T (x ) dx = r0 + Z x x0 ∇R(Tx ) T dx , (2) where T (x ) is the local separating matrix for a neighborhood of the image x , and T is the global separating matrix for the whole PDF. Therefore, the Jacobian of the second non-linear stage is: ∇R(y) = ∇R(Tx) = T (x) T−1 . 55 (3) ´ ´ ´ C AMPS -VALLS , G UTI E RREZ , G OMEZ -P E REZ AND M ALO As local linear independent features around a particular image, x, differ in general from global linear independent features, that is, T (x) = T, the above product is not the identity nor diagonal in general. 3.2 Non-diagonal Jacobian in Non-linear Perceptual Transforms The current response model for the cortical frequency analyzers is non-linear (Heeger, 1992; Watson and Solomon, 1997). The outputs of the ﬁlters of the ﬁrst linear stage, y, undergo a non-linear sigmoid transform in which the energy of each linear coefﬁcient is weighted by a linear Contrast Sensitivity Function (CSF) (Campbell and Robson, 1968; Malo et al., 1997) and is further normalized by a combination of the energies of neighbor coefﬁcients in frequency, r f = R(y) f = sgn(y f ) |α f y f |γ , β f + ∑n =1 h f f |α f y f |γ f (4) where α f (Figure 4[top left]) are CSF-like weights, β f (Figure 4[top right]) control the sharpness of the response saturation for each coefﬁcient, γ is the so called excitation exponent, and the matrix h f f determines the interaction neighborhood in the non-linear normalization of the energy. This interaction matrix models the cross-masking behavior (cf. Section 2.2). The interaction in this matrix is assumed to be Gaussian (Watson and Solomon, 1997), and its width increases with the frequency. Figure 4[bottom] shows two examples of this Gaussian interaction for two particular coefﬁcients in a local Fourier domain. Note that the width of the perceptual interaction neighborhood increases with the frequency in the same way as the width of the statistical interaction neighborhood shown in Figure 2. We used a value of γ = 2 in the experiments. Taking derivatives in the general divisive normalization model, Equation (4), we obtain ∇R(y) f f = sgn(y f )γ α f |α f y f |γ |α f y f |γ−1 α f |α f y f |γ−1 δf f − hf f β f + ∑n =1 h f f |α f y f |γ (β f + ∑n =1 h f f |α f y f |γ )2 f f , (5) which is not diagonal because of the interaction matrix, h, which describes the cross-masking between each frequency f and the remaining f = f . Note that the intrinsic non-linear nature of both the statistical and perceptual transforms, Equations (3) and (5), makes the above results true for any linear domain under consideration. Speciﬁcally, if any other possible linear domain for image representation is considered, y = T y, then the Jacobian of the corresponding independence transform, R , is ∇R (y ) = ∇R(y) T −1 , which, in general, will also be non-diagonal because of the non-diagonal and point-dependent nature of ∇R(y). To summarize, since no linear domain fulﬁlls the diagonal Jacobian condition in either statistical or perceptual terms, the negative situation illustrated in Figure 1 may occur when using SVM in these domains. Therefore, improved results could be obtained if SVM learning were applied after some transform achieving independent coefﬁcients, R. 4. SVM Learning in a Perceptually Independent Representation In order to conﬁrm the above theoretical results (i.e., the unsuitability of linear representation domains for SVM learning) and to assess the eventual gain that can be obtained from training SVR 56 O N THE S UITABLE D OMAIN FOR SVM T RAINING IN I MAGE C ODING 0.04 αf 0.015 βf 0.03 0.012 0.02 0.009 0.006 0.01 0.003 0 0 4 8 12 16 20 24 28 32 0 f (frequency cycles/degree) 4 8 | f | = 10.8 cpd 16 20 24 28 32 | f | = 24.4 cpd −30 −20 −20 −10 −10 f (cpd) −30 0 y 0 y f (cpd) 12 f (frequency cycles/degree) 10 10 20 20 30 −30 −20 −10 0 fx (cpd) 10 20 30 −30 30 −20 −10 0 fx (cpd) 10 20 30 Figure 4: Parameters of the perceptual model: α f (top left), β f (top right). Bottom ﬁgures represent perceptual interaction neighborhoods h f f of two particular coefﬁcients of the local Fourier domain. in a more appropriate domain, we should compare the performance of SVRs in previously reported linear domains (e.g., block-DCT or wavelets) and in one of the proposed non-linear domains (either the statistically independent domain or the perceptually independent domain). Exploration of the statistical independence transform may have academic interest but, in its present formulation, it is not practical for coding purposes: direct application of non-linear ICA as in Equation (2) is very time-consuming for high dimensional vectors since lots of local ICA computations are needed to transform each block, and a very large image database is needed for a robust and signiﬁcant computation of R. Besides, an equally expensive differential approach is also needed to compute the inverse R−1 for image decoding. In contrast, the perceptual non-linearity (and its inverse) are analytical. These analytical expressions are feasible for reasonable block sizes, and there are efﬁcient iterative methods that can be used for larger vectors (Malo et al., 2006). In this paper, we explore the use of a psychophysically-based divisive normalized domain: ﬁrst compute a block-DCT transform and then apply the divisive normalization model described above for each block. The results will be compared to the ﬁrst competitive SVM coding results (Robinson 57 ´ ´ ´ C AMPS -VALLS , G UTI E RREZ , G OMEZ -P E REZ AND M ALO ´ and Kecman, 2003) and the posterior improvements reported by G omez-P´ rez et al. (2005), both e formulated in the linear block-DCT domain. As stated in Section 2, by construction, the proposed domain is perceptually Euclidean with perceptually independent components. The Euclidean nature of this domain has an additional beneﬁt: the ε-insensitivity design is very simple because a constant value is appropriate due to the constant perceptual relevance of all coefﬁcients. Thus, direct application of the standard SVR method is theoretically appropriate in this domain. Moreover, beyond its built-in perceptual beneﬁts, this psychophysically-based divisive normalization has attractive statistical properties: it strongly reduces the mutual information between the ﬁnal coefﬁcients r (Malo et al., 2006). This is not surprising according to the hypothesis that try to explain the early stages of biological vision systems using information theory arguments (Barlow, 1961; Simoncelli and Olshausen, 2001). Speciﬁcally, dividing the energy of each linear coefﬁcient by the energy of the neighbors, which are statistically related with it, cf. Figure 2, gives coefﬁcients with reduced statistical dependence. Moreover, as the empirical non-linearities of perception have been reproduced using non-linear ICA in Equation (2) (Malo and Guti´ rrez, 2006), the empirical die visive normalization can be seen as a convenient parametric way to obtain statistical independence. 5. Performance of SVM Learning in Different Domains In this section, we analyze the performance of SVM-based coding algorithms in linear and nonlinear domains through rate-distortion curves and explicit examples for visual comparison. In addition, we discuss how SVM selects support vectors in these domains to represent the image features. 5.1 Model Development and Experimental Setup In the (linear) block-DCT domain, y, we use the method introduced by Robinson and Kecman (2003) (RKi-1), in which the SVR is trained to learn a ﬁxed (low-pass) number of DCT coefﬁcients ´ (those with frequency bigger than 20 cycl/deg are discarded); and the method proposed by G omezP´ rez et al. (2005) (CSF-SVR), in which the relevance of all DCT coefﬁcients is weighted according e to the CSF criterion using an appropriately modulated ε f . In the non-linear domain, r, we use the SVR with constant insensitivity parameter ε (NL-SVR). In all cases, the block-size is 16×16, that is, y, r ∈ R256 . The behavior of JPEG standard is also included in the experiments for comparison purposes. As stated in Section 1, we used the RBF kernel and arbitrarily large penalization parameter in every SVR case. In all experiments, we trained the SVR models without the bias term, and modelled the absolute value of the DCT, y, or response coefﬁcients, r. All the remaining free parameters (ε-insensitivity and Gaussian width of the RBF kernel σ) were optimized for all the considered models and different compression ratios. In the NL-SVM case, the parameters of the divisive normalization used in the experiments are shown in Figure 4. After training, the signal is described by the uniformly quantized Lagrange multipliers of the support vectors needed to keep the regression error below the thresholds ε f . The last step is entropy coding of the quantized weights. The compression ratio is controlled by a factor applied to the thresholds, ε f . 58 O N THE S UITABLE D OMAIN FOR SVM T RAINING IN I MAGE C ODING 5.2 Model Comparison In order to assess the quality of the coded images, three different measures were used: the standard ´ (Euclidean) RMSE, the Maximum Perceptual Error (MPE) (Malo et al., 2000; G omez-P´ rez et al., e 2005; Malo et al., 2006) and the also perceptually meaningful Structural SIMilarity (SSIM) index (Wang et al., 2004). Eight standard 256×256 monochrome 8 bits/pix images were used in the experiments. Average rate-distortion curves are plotted in Figure 5 in the range [0.05, 0.6] bits/pix (bpp). According to these entropy-per-sample data, original ﬁle size was 64 KBytes in every case, while the compressed image sizes were in the range [0.4, 4.8] KBytes. This implies that the compression ratios were in the range [160:1, 13:1]. In general, a clear gain over standard JPEG is obtained by all SVM-based methods. According to the standard Euclidean MSE point of view, the performance of RKi-1 and CSF-SVR algorithms is basically the same (note the overlapped curves in Figure 5(a)). However, it is widely known that the MSE results are not useful to represent the subjective quality of images, as extensively reported elsewhere (Girod, 1993; Teo and Heeger, 1994; Watson and Malo, 2002). When using more appropriate (perceptually meaningful) quality measures (Figures 5(b)-(c)), the CSF-SVR obtains a certain advantage over the RKi-1 algorithm for all compression rates, which was already reported by G´ mez-P´ rez et al. (2005). In all measures, and for the whole considered entropy range, the o e proposed NL-SVR clearly outperforms all previously reported methods, obtaining a noticeable gain at medium-to-high compression ratios (between 0.1 bpp (80:1) and 0.3 bpp (27:1)). Taking into account that the recommended bit rate for JPEG is about 0.5 bpp, from Figure 5 we can also conclude that the proposed technique achieves the similar quality levels at a lower bit rate in the range [0.15, 0.3] bpp. Figure 6 shows representative visual results of the considered SVM strategies on standard images (Lena and Barbara) at the same bit rate (0.3 bpp, 27:1 compression ratio or 2.4 KBytes in 256×256 images). The visual inspection conﬁrms that the numerical gain in MPE and SSIM shown in Figure 5 is also perceptually signiﬁcant. Some conclusions can be extracted from this ﬁgure. ´ First, as previously reported by Gomez-P´ rez et al. (2005), RKi-1 leads to poorer (blocky) results e because of the crude approximation of the CSF (as an ideal low-pass ﬁlter) and the equal relevance applied to the low-frequency DCT-coefﬁcients. Second, despite the good performance yielded by the CSF-SVR approach to avoid blocking effects, it is worth noting that high frequency details are smoothed (e.g., see Barbara’s scarf). These effects are highly alleviated by introducing SVR in the non-linear domain. See, for instance, Lena’s eyes, her hat’s feathers or the better reproduction of the high frequency pattern in Barbara’s clothes. Figure 7 shows the results obtained by all considered methods at a very high compression ratio for the Barbara image (0.05 bpp, 160:1 compression ratio or 0.4 KBytes in 256×256 images). This experiment is just intended to show the limits of methods performance since it is out of the recommended rate ranges. Even though this scenario is unrealistic, differences among methods are still noticeable: the proposed NL-SVR method reduces the blocky effects (note for instance that the face is better reproduced). This is due to a better distribution of support vectors in the perceptually independent domain. 5.3 Support Vector Distribution The observed different perceptual image quality obtained with each approach is a direct consequence of support vector distribution in different domains. Figure 8 shows a representative example 59 ´ ´ ´ C AMPS -VALLS , G UTI E RREZ , G OMEZ -P E REZ AND M ALO 20 JPEG RKi−1 CSF−SVR NL−SVR 18 RMSE 16 14 12 10 8 0.1 0.2 0.3 0.4 Entropy (bits/pix) 0.5 0.6 22 JPEG RKi−1 CSF−SVR NL−SVR 20 18 16 MPE 14 12 10 8 6 4 0.1 0.2 0.3 0.4 Entropy (bits/pix) 0.5 0.6 0.85 0.8 SSIM 0.75 0.7 0.65 JPEG 0.6 RKi−1 CSF−SVR 0.55 NL−SVR 0.5 0.1 0.2 0.3 0.4 Entropy (bits/pix) 0.5 0.6 Figure 5: Average rate distortion curves over eight standard images (Lena, Barbara, Boats, Einstein, Peppers, Mandrill, Goldhill, Camera man) using objective and subjective measures for the considered JPEG (dotted) and the SVM approaches (RKi-1 dash-dotted, CSF-SVR dashed and NL-SVR solid). RMSE distortion (top), Maximum Perceptual Error, MPE (middle) (Malo et al., 2000; G´ mez-P´ rez et al., 2005; Malo et al., 2006), and Structural o e SIMilarity index, SSIM (bottom) (Wang et al., 2004). 60 O N THE S UITABLE D OMAIN FOR SVM T RAINING IN I MAGE C ODING Rki Rki SVR+CSF SVR+CSF NL+SVR NL+SVR Figure 6: Examples of decoded Lena (left) and Barbara (right) images at 0.3 bits/pix. From top to bottom: JPEG, RKi-1, CSF-SVR, and NL-SVR. 61 ´ ´ ´ C AMPS -VALLS , G UTI E RREZ , G OMEZ -P E REZ AND M ALO (a) (b) (c) (d) Figure 7: Examples of decoded Barbara images at a high compression ratio of 0.05 bits/pix (160:1) for (a) JPEG, (b) RKi-1, (c) CSF-SVR, and (d) NL-SVR. of the distribution of the selected support vectors by the RKi-1 and the CSF-SVR models working in the linear DCT domain, and the NL-SVM working in the perceptually independent non-linear domain r. Speciﬁcally, a block of Barbara’s scarf at different compression ratios is used for illustration purposes. The RKi-1 approach (Robinson and Kecman, 2003) uses a constant ε but, in order to consider the low subjective relevance of the high-frequency region, the corresponding coefﬁcients are neglected. As a result, this approach only allocates support vectors in the low/medium frequency regions. The CSF-SVR approach uses a variable ε according to the CSF and gives rise to a more natural concentration of support vectors in the low/medium frequency region, which captures medium to high frequency details at lower compression rates (0.5 bits/pix). Note that the number of support vectors is bigger than in the RKi-1 approach, but it selects some necessary high-frequency coefﬁcients to keep the error below the selected threshold. However, for bigger compression ratios (0.3 bits/pix), it misrepresents some high frequency, yet relevant, features (e.g., the peak from the stripes). The NL-SVM approach works in the non-linear transform domain, in which a more uniform coverage 62 O N THE S UITABLE D OMAIN FOR SVM T RAINING IN I MAGE C ODING (a) (b) (c) 0 200 10 100 0 200 5 5 15 0 0 20 30 25 20 10 20 30 fy (cpd) 10 0 fy (cpd) 30 10 fx (cpd) 0 200 5 10 100 0 200 5 10 100 15 25 10 30 20 30 f (cpd) y 10 0 30 10 0.5 20 30 25 20 5 1 0 0 20 30 0 1.5 15 15 0 fy (cpd) 30 0 fx (cpd) 20 25 20 0 fx (cpd) fx (cpd) 20 30 25 20 30 10 0.5 15 0 5 1 10 100 15 0 1.5 25 20 f (cpd) y 10 30 fy (cpd) 0 0 fx (cpd) fx (cpd) Figure 8: Signal in different domains and the selected support vectors by the SVM models in a block of the Barbara image at 0.3 bits/pix (top row) and 0.5 bits/pix (bottom row). Different domains are analyzed: (a) linear DCT using RKi-1, (b) linear DCT with CSF-SVM, and (c) non-linear perceptual domain with standard ε-SVM (NL-SVR). of the domain is done, accounting for richer (and perceptually independent) coefﬁcients to perform efﬁcient sparse signal reconstruction. It is important to remark that, for a given method (or domain), tightening ε f implies (1) considering more support vectors, and (2) an increase in entropy (top and bottom rows in Figure 8, 0.3 bpp to 0.5 bpp). However, note that the relevant measure is the entropy and not the number of support vectors: even though the number of selected support vectors in the r domain is higher, their variance is lower, thus giving rise to the same entropy after entropy coding. 6. Conclusions In this paper, we have reported a condition on the suitable domain for developing efﬁcient SVM image coding schemes. The so-called diagonal Jacobian condition states that SVM regression with scalar-wise error restriction in a particular domain makes sense only if the transform that maps this domain to an independent coefﬁcient representation is locally diagonal. We have demonstrated that, 63 ´ ´ ´ C AMPS -VALLS , G UTI E RREZ , G OMEZ -P E REZ AND M ALO in general, linear domains do not fulﬁll this condition because non-trivial statistical and perceptual inter-coefﬁcient relations do exist in these domains. This theoretical ﬁnding has been experimentally conﬁrmed by observing that improved compression results are obtained when SVM is applied in a non-linear perceptual domain that starts from the same linear domain used by previously reported SVM-based image coding schemes. These results highlight the relevance of an appropriate image representation choice before SVM learning. Further work is tied to the use of SVM-based coding schemes in statistically, rather than perceptually, independent non-linear ICA domains. In order to do so, local PCA instead of local ICA may be used in the local-to-global differential approach (Malo and Guti´ rrez, 2006) to speed up the e non-linear computation. Acknowledgments This work has been partly supported by the Spanish Ministry of Education and Science under grant CICYT TEC2006-13845/TCM and by the Generalitat Valenciana under grant GV-06/215. References R. Ahmed. Wavelet-based image compression using SVM learning and encoding techniques. Proc. 8th IASTED International Conference Computer Graphics and Imaging, 1:162–166, 2005. H. B. Barlow. Redundancy reduction revisited. Network: Comp. Neur. Sys., 12:241–253, 2001. H. B. Barlow. Possible principles underlying the transformation of sensory messages. In WA Rosenblith, editor, Sensory Communication, pages 217–234. MIT Press, Cambridge, MA, 1961. A. J. Bell and T. J. Sejnowski. The ‘independent components’ of natural scenes are edge ﬁlters. Vision Research, 37(23):3327–3338, 1997. R. W. Buccigrossi and E. P. Simoncelli. Image compression via joint statistical characterization in the wavelet domain. IEEE Transactions on Image Processing, 8(12):1688–1701, 1999. F. W. Campbell and J. G. Robson. Application of Fourier analysis to the visibility of gratings. Journal of Physiology, 197(3):551–566, August 1968. G. Camps-Valls, E. Soria-Olivas, J. P´ rez-Ruixo, A. Art´ s-Rodr´guez, F. P´ rez-Cruz, and e e ı e A. Figueiras-Vidal. A proﬁle-dependent kernel-based regression for cyclosporine concentration prediction. In Neural Information Processing Systems (NIPS) – Workshop on New Directions in Kernel-Based Learning Methods, Vancouver, Canada, December 2001. R. J. Clarke. Transform Coding of Images. Academic Press, New York, 1985. I. Epifanio, J. Guti´ rrez, and J. Malo. Linear transform for simultaneous diagonalization of coe variance and perceptual metric matrix in image coding. Pattern Recognition, 36(8):1799–1811, August 2003. J. M. Foley. Human luminance pattern mechanisms: Masking experiments require a new model. Journal of the Optical Society of America A, 11(6):1710–1719, 1994. 64 O N THE S UITABLE D OMAIN FOR SVM T RAINING IN I MAGE C ODING A. Gersho and R. M. Gray. Vector Quantization and Signal Compression. Kluwer Academic Press, Boston, 1992. B. Girod. What’s wrong with mean-squared error. In A. B. Watson, editor, Digital Images and Human Vision, pages 207–220. The MIT press, 1993. G. G´ mez-P´ rez, G. Camps-Valls, J. Guti´ rrez, and J. Malo. Perceptually adaptive insensitivity for o e e support vector machine image coding. IEEE Transactions on Neural Networks, 16(6):1574–1581, 2005. J. Guti´ rrez, F. J. Ferri, and J. Malo. Regularization operators for natural images based on nonlinear e perception models. IEEE Transactions on Image Processing, 15(1):189–2000, January 2006. D. J. Heeger. Normalization of cell responses in cat striate cortex. Vis. Neurosci., 9:181–198, 1992. A. Hyvarinen, J. Karhunen, and E. Oja. Independent Component Analysis. John Wiley & Sons, New York, 2001. A. Hyvarinen, J. Hurri, and J. Vayrynenm. Bubbles: a unifying framework for low-level statistical properties of natural image sequences. Journal of the Optical Society of America A, 20(7), 2003. R. Jiao, Y. Li, Q. Wang, and B. Li. SVM regression and its application to image compression. In International Conference on Intelligent Computing, volume 3645, pages 747–756. Lecture Notes on Computer Science, 2005. C. Jutten and J. Karhunen. Advances in nonlinear blind source separation. Proc. of the 4th Int. Symp. on Independent Component Analysis and Blind Signal Separation (ICA2003), pages 245– 256, 2003. J. Karhunen, S. Malaroiu, and M. Ilmoniemi. Local linear independent component analysis based on clustering. Intl. J. Neur. Syst., 10(6), December 2000. J. K. Lin. Factorizing probability density functions: Generalizing ICA. In Proc. of the First Intl. Workshop on ICA and Signal Separation, pages 313–318, Aussois, France, 1999. J. Malo and J. Guti´ rrez. V1 non-linear properties emerge from local-to-global non-linear ICA. e Network: Computation in Neural Systems, 17:85–102, 2006. J. Malo, A. M. Pons, A. Felipe, and J. M. Artigas. Characterization of human visual system threshold performance by a weighting function in the Gabor domain. Journal of Modern Optics, 44(1): 127–148, 1997. J. Malo, F. Ferri, J. Albert, J. Soret, and J. M. Artigas. The role of perceptual contrast non-linearities in image transform coding. Image & Vision Computing, 18(3):233–246, February 2000. J. Malo, J. Guti´ rrez, I. Epifanio, F. Ferri, and J. M. Artigas. Perceptual feed-back in multigrid moe tion estimation using an improved DCT quantization. IEEE Transactions on Image Processing, 10(10):1411–1427, October 2001. J. Malo, I. Epifanio, R. Navarro, and E. Simoncelli. Non-linear image representation for efﬁcient perceptual coding. IEEE Transactions on Image Processing, 15(1):68–80, 2006. 65 ´ ´ ´ C AMPS -VALLS , G UTI E RREZ , G OMEZ -P E REZ AND M ALO B. A. Olshausen and D. J. Field. Emergence of simple-cell receptive ﬁeld properties by learning a sparse code for natural images. Nature, 381:607–609, 1996. F. P´ rez-Cruz, G. Camps-Valls, E. Soria-Olivas, J. J. P´ rez-Ruixo, A. R. Figueiras-Vidal, and e e A. Art´ s-Rodr´guez. Multi-dimensional function approximation and regression estimation. In e ı International Conference on Artiﬁcial Neural Networks, ICANN2002, Madrid, Spain., Aug 2002. Lecture Notes in Computer Science. Springer–Verlag. J. Robinson and V. Kecman. The use of Support Vector Machines in image compression. In Proceedings of the International Conference on Engineering Intelligence Systems, EIS’2000, volume 36, pages 93–96, University of Paisley, Scotland, U.K., June 2000. J. Robinson and V. Kecman. Combining Support Vector Machine learning with the discrete cosine transform in image compression. IEEE Transactions Neural Networks, 14(4):950–958, July 2003. O. Schwartz and E. P. Simoncelli. Natural signal statistics and sensory gain control. Nature Neuroscience, 4(8):819–825, 2001. E. P. Simoncelli. Vision and the statistics of the visual environment. Current Opinion in Neurobiology, 13(2):144–149, 2003. E. P. Simoncelli. Statistical models for images: Compression, restoration and synthesis. In 31st Asilomar Conference on Signals, Systems and Computers, Paciﬁc Grove, CA, 1997. E. P. Simoncelli and B. O. Olshausen. Natural image statistics and neural representation. Annual Review of Neuroscience, 24:1193–1216, 2001. A. J. Smola and B. Sch¨ lkopf. A tutorial on support vector regression. Statistics and Computing, o 14:199–222, 2004. D. S. Taubman and M. W. Marcellin. JPEG2000: Image Compression Fundamentals, Standards and Practice. Kluwer Academic Publishers, Boston, 2001. P. C. Teo and D. J. Heeger. Perceptual image distortion. Proceedings of the First IEEE International Conference on Image Processing, 2:982–986, 1994. M. J. Wainwright, E. P. Simoncelli, and A. S. Willsky. Random cascades on wavelet trees and their use in analyzing and modeling natural images. Applied and Computational Harmonic Analysis, 11:89–123, 2001. Z. Wang, A. C. Bovik, H. R. Sheikh, and E. P. Simoncelli. Image quality assessment: From error visibility to structural similarity. IEEE Transactions on Image Processing, 13(4):600–612, 2004. A. B. Watson and J. Malo. Video quality measures based on the standard spatial observer. In Proceedings of the IEEE International Conference on Image Proceedings, volume 3, pages 41– 44, 2002. A. B. Watson and J. A. Solomon. A model of visual contrast gain control and pattern masking. Journal of the Optical Society of America A, 14(9):2379–2391, September 1997. 66

5 0.22560056 58 jmlr-2008-Max-margin Classification of Data with Absent Features

Author: Gal Chechik, Geremy Heitz, Gal Elidan, Pieter Abbeel, Daphne Koller

6 0.22560038 51 jmlr-2008-Learning Similarity with Operator-valued Large-margin Classifiers

7 0.18345051 81 jmlr-2008-Regularization on Graphs with Function-adapted Diffusion Processes

8 0.17578602 30 jmlr-2008-Discriminative Learning of Max-Sum Classifiers

9 0.17477521 29 jmlr-2008-Cross-Validation Optimization for Large Scale Structured Classification Kernel Methods

10 0.15996525 41 jmlr-2008-Graphical Models for Structured Classification, with an Application to Interpreting Images of Protein Subcellular Location Patterns

11 0.15672486 12 jmlr-2008-Algorithms for Sparse Linear Classifiers in the Massive Data Setting

12 0.14483202 9 jmlr-2008-Active Learning by Spherical Subdivision

13 0.1434544 25 jmlr-2008-Consistency of Random Forests and Other Averaging Classifiers

14 0.13943835 65 jmlr-2008-Multi-Agent Reinforcement Learning in Common Interest and Fixed Sum Stochastic Games: An Experimental Study

15 0.13709842 22 jmlr-2008-Closed Sets for Labeled Data

16 0.12441877 10 jmlr-2008-Active Learning of Causal Networks with Intervention Experiments and Optimal Designs (Special Topic on Causality)

17 0.1234646 42 jmlr-2008-HPB: A Model for Handling BN Nodes with High Cardinality Parents

18 0.1226653 14 jmlr-2008-An Extension on "Statistical Comparisons of Classifiers over Multiple Data Sets" for all Pairwise Comparisons

19 0.11911935 67 jmlr-2008-Near-Optimal Sensor Placements in Gaussian Processes: Theory, Efficient Algorithms and Empirical Studies

20 0.11665204 33 jmlr-2008-Evidence Contrary to the Statistical View of Boosting

similar papers computed by lda model

lda for this paper:

topicId topicWeight

[(0, 0.027), (1, 0.016), (5, 0.51), (9, 0.011), (31, 0.02), (40, 0.052), (54, 0.027), (58, 0.023), (66, 0.03), (76, 0.025), (88, 0.073), (92, 0.03), (94, 0.029), (99, 0.024)]

similar papers list:

simIndex simValue paperId paperTitle

same-paper 1 0.89467704 87 jmlr-2008-Stationary Features and Cat Detection

Author: FranĂ§ois Fleuret, Donald Geman

2 0.78457302 76 jmlr-2008-Optimization Techniques for Semi-Supervised Support Vector Machines

Author: Olivier Chapelle, Vikas Sindhwani, Sathiya S. Keerthi

Abstract: Due to its wide applicability, the problem of semi-supervised classiﬁcation is attracting increasing attention in machine learning. Semi-Supervised Support Vector Machines (S 3 VMs) are based on applying the margin maximization principle to both labeled and unlabeled examples. Unlike SVMs, their formulation leads to a non-convex optimization problem. A suite of algorithms have recently been proposed for solving S3 VMs. This paper reviews key ideas in this literature. The performance and behavior of various S3 VM algorithms is studied together, under a common experimental setting. Keywords: semi-supervised learning, support vector machines, non-convex optimization, transductive learning

3 0.53380668 41 jmlr-2008-Graphical Models for Structured Classification, with an Application to Interpreting Images of Protein Subcellular Location Patterns

Author: Shann-Ching Chen, Geoffrey J. Gordon, Robert F. Murphy

Abstract: In structured classiﬁcation problems, there is a direct conﬂict between expressive models and efﬁcient inference: while graphical models such as Markov random ﬁelds or factor graphs can represent arbitrary dependences among instance labels, the cost of inference via belief propagation in these models grows rapidly as the graph structure becomes more complicated. One important source of complexity in belief propagation is the need to marginalize large factors to compute messages. This operation takes time exponential in the number of variables in the factor, and can limit the expressiveness of the models we can use. In this paper, we study a new class of potential functions, which we call decomposable k-way potentials, and provide efﬁcient algorithms for computing messages from these potentials during belief propagation. We believe these new potentials provide a good balance between expressive power and efﬁcient inference in practical structured classiﬁcation problems. We discuss three instances of decomposable potentials: the associative Markov network potential, the nested junction tree, and a new type of potential which we call the voting potential. We use these potentials to classify images of protein subcellular location patterns in groups of cells. Classifying subcellular location patterns can help us answer many important questions in computational biology, including questions about how various treatments affect the synthesis and behavior of proteins and networks of proteins within a cell. Our new representation and algorithm lead to substantial improvements in both inference speed and classiﬁcation accuracy. Keywords: factor graphs, approximate inference algorithms, structured classiﬁcation, protein subcellular location patterns, location proteomics

4 0.31175387 81 jmlr-2008-Regularization on Graphs with Function-adapted Diffusion Processes

Author: Arthur D. Szlam, Mauro Maggioni, Ronald R. Coifman

Abstract: Harmonic analysis and diffusion on discrete data has been shown to lead to state-of-the-art algorithms for machine learning tasks, especially in the context of semi-supervised and transductive learning. The success of these algorithms rests on the assumption that the function(s) to be studied (learned, interpolated, etc.) are smooth with respect to the geometry of the data. In this paper we present a method for modifying the given geometry so the function(s) to be studied are smoother with respect to the modiﬁed geometry, and thus more amenable to treatment using harmonic analysis methods. Among the many possible applications, we consider the problems of image denoising and transductive classiﬁcation. In both settings, our approach improves on standard diffusion based methods. Keywords: diffusion processes, diffusion geometry, spectral graph theory, image denoising, transductive learning, semi-supervised learning

5 0.31070286 18 jmlr-2008-Bayesian Inference and Optimal Design for the Sparse Linear Model

Author: Matthias W. Seeger

Abstract: The linear model with sparsity-favouring prior on the coefﬁcients has important applications in many different domains. In machine learning, most methods to date search for maximum a posteriori sparse solutions and neglect to represent posterior uncertainties. In this paper, we address problems of Bayesian optimal design (or experiment planning), for which accurate estimates of uncertainty are essential. To this end, we employ expectation propagation approximate inference for the linear model with Laplace prior, giving new insight into numerical stability properties and proposing a robust algorithm. We also show how to estimate model hyperparameters by empirical Bayesian maximisation of the marginal likelihood, and propose ideas in order to scale up the method to very large underdetermined problems. We demonstrate the versatility of our framework on the application of gene regulatory network identiﬁcation from micro-array expression data, where both the Laplace prior and the active experimental design approach are shown to result in signiﬁcant improvements. We also address the problem of sparse coding of natural images, and show how our framework can be used for compressive sensing tasks. Part of this work appeared in Seeger et al. (2007b). The gene network identiﬁcation application appears in Steinke et al. (2007). Keywords: sparse linear model, Laplace prior, expectation propagation, approximate inference, optimal design, Bayesian statistics, gene network recovery, image coding, compressive sensing

6 0.30520305 31 jmlr-2008-Dynamic Hierarchical Markov Random Fields for Integrated Web Data Extraction

7 0.30460688 83 jmlr-2008-Robust Submodular Observation Selection

8 0.30020115 51 jmlr-2008-Learning Similarity with Operator-valued Large-margin Classifiers

9 0.29487079 58 jmlr-2008-Max-margin Classification of Data with Absent Features

10 0.2872037 59 jmlr-2008-Maximal Causes for Non-linear Component Extraction

11 0.28513986 34 jmlr-2008-Exponentiated Gradient Algorithms for Conditional Random Fields and Max-Margin Markov Networks

12 0.28318453 86 jmlr-2008-SimpleMKL

13 0.28218356 39 jmlr-2008-Gradient Tree Boosting for Training Conditional Random Fields

14 0.28178334 73 jmlr-2008-On the Suitable Domain for SVM Training in Image Coding

15 0.27374291 66 jmlr-2008-Multi-class Discriminant Kernel Learning via Convex Programming (Special Topic on Model Selection)

16 0.2709699 67 jmlr-2008-Near-Optimal Sensor Placements in Gaussian Processes: Theory, Efficient Algorithms and Empirical Studies

17 0.26489881 96 jmlr-2008-Visualizing Data using t-SNE

18 0.26450583 89 jmlr-2008-Support Vector Machinery for Infinite Ensemble Learning

19 0.26365516 16 jmlr-2008-Approximations for Binary Gaussian Process Classification

20 0.2632255 43 jmlr-2008-Hit Miss Networks with Applications to Instance Selection