Distance Metric Between 3D Models and 3D Images for Recognition and Classification

Similarity measurements between 3D objects and 2D images are useful for the tasks of object recognition and classification. We distinguish between two types of similarity metrics: metrics computed in image-space (image metrics) and metrics computed in transformation-space (transformation metrics). Existing methods typically use image and the nearest view of the object. Example for such a measure is the Euclidean distance between feature points in the image and corresponding points in the nearest view. (Computing this measure is equivalent to solving the exterior orientation calibration problem.) In this paper we introduce a different type of metrics: transformation metrics. These metrics penalize for the deformatoins applied to the object to produce the observed image. We present a transformation metric that optimally penalizes for "affine deformations" under weak-perspective. A closed-form solution, together with the nearest view according to this metric, are derived. The metric is shown to be equivalent to the Euclidean image metric, in the sense that they bound each other from both above and below. For Euclidean image metric we offier a sub-optimal closed-form solution and an iterative scheme to compute the exact solution.

A link density clustering algorithm based on automatically selecting density peaks for overlapping community detection

Peer reviewed

Modified Large Margin Nearest Neighbor Metric Learning for Regression

The main objective of this letter is to formulate a new approach of learning a Mahalanobis distance metric for nearest neighbor regression from a training sample set. We propose a modified version of the large margin nearest neighbor metric learning method to deal with regression problems. As an application, the prediction of post-operative trunk 3-D shapes in scoliosis surgery using nearest neighbor regression is described. Accuracy of the proposed method is quantitatively evaluated through experiments on real medical data.

Bayes Factor based speaker segmentation for speaker diarization

This paper proposes the use of the Bayes Factor as a distance metric for speaker segmentation within a speaker diarization system. The proposed approach uses a pair of constant sized, sliding windows to compute the value of the Bayes Factor between the adjacent windows over the entire audio. Results obtained on the 2002 Rich Transcription Evaluation dataset show an improved segmentation performance compared to previous approaches reported in literature using the Generalized Likelihood Ratio. When applied in a speaker diarization system, this approach results in a 5.1% relative improvement in the overall Diarization Error Rate compared to the baseline.

Automatic population HARDI white matter tract clustering by label fusion of multiple tract atlases

Automatic labeling of white matter fibres in diffusion-weighted brain MRI is vital for comparing brain integrity and connectivity across populations, but is challenging. Whole brain tractography generates a vast set of fibres throughout the brain, but it is hard to cluster them into anatomically meaningful tracts, due to wide individual variations in the trajectory and shape of white matter pathways. We propose a novel automatic tract labeling algorithm that fuses information from tractography and multiple hand-labeled fibre tract atlases. As streamline tractography can generate a large number of false positive fibres, we developed a top-down approach to extract tracts consistent with known anatomy, based on a distance metric to multiple hand-labeled atlases. Clustering results from different atlases were fused, using a multi-stage fusion scheme. Our "label fusion" method reliably extracted the major tracts from 105-gradient HARDI scans of 100 young normal adults. © 2012 Springer-Verlag.

Parts based representation for pedestrian using NMF with robustness to partial occlusion

Computer Vision has seen a resurgence in the parts-based representation for objects over the past few years. The parts are usually annotated beforehand for training. We present an annotation free parts-based representation for the pedestrian using Non-Negative Matrix Factorization (NMF). We show that NMF is able to capture the wide range of pose and clothing of the pedestrians. We use a modified form of NMF i.e. NMF with sparsity constraints on the factored matrices. We also make use of Riemannian distance metric for similarity measurements in NMF space as the basis vectors generated by NMF aren't orthogonal. We show that for 1% drop in accuracy as compared to the Histogram of Oriented Gradients (HOG) representation we can achieve robustness to partial occlusion.

Turning visual search time on its head

Our everyday visual experience frequently involves searching for objects in clutter. Why are some searches easy and others hard? It is generally believed that the time taken to find a target increases as it becomes similar to its surrounding distractors. Here, I show that while this is qualitatively true, the exact relationship is in fact not linear. In a simple search experiment, when subjects searched for a bar differing in orientation from its distractors, search time was inversely proportional to the angular difference in orientation. Thus, rather than taking search reaction time (RT) to be a measure of target-distractor similarity, we can literally turn search time on its head (i.e. take its reciprocal 1/RT) to obtain a measure of search dissimilarity that varies linearly over a large range of target-distractor differences. I show that this dissimilarity measure has the properties of a distance metric, and report two interesting insights come from this measure: First, for a large number of searches, search asymmetries are relatively rare and when they do occur, differ by a fixed distance. Second, search distances can be used to elucidate object representations that underlie search - for example, these representations are roughly invariant to three-dimensional view. Finally, search distance has a straightforward interpretation in the context of accumulator models of search, where it is proportional to the discriminative signal that is integrated to produce a response. This is consistent with recent studies that have linked this distance to neuronal discriminability in visual cortex. Thus, while search time remains the more direct measure of visual search, its reciprocal also has the potential for interesting and novel insights. (C) 2012 Elsevier Ltd. All rights reserved.

Linear dimensionality reduction: Survey, insights, and generalizations

© 2015 John P. Cunningham and Zoubin Ghahramani. Linear dimensionality reduction methods are a cornerstone of analyzing high dimensional data, due to their simple geometric interpretations and typically attractive computational properties. These methods capture many data features of interest, such as covariance, dynamical structure, correlation between data sets, input-output relationships, and margin between data classes. Methods have been developed with a variety of names and motivations in many fields, and perhaps as a result the connections between all these methods have not been highlighted. Here we survey methods from this disparate literature as optimization programs over matrix manifolds. We discuss principal component analysis, factor analysis, linear multidimensional scaling, Fisher's linear discriminant analysis, canonical correlations analysis, maximum autocorrelation factors, slow feature analysis, sufficient dimensionality reduction, undercomplete independent component analysis, linear regression, distance metric learning, and more. This optimization framework gives insight to some rarely discussed shortcomings of well-known methods, such as the suboptimality of certain eigenvector solutions. Modern techniques for optimization over matrix manifolds enable a generic linear dimensionality reduction solver, which accepts as input data and an objective to be optimized, and returns, as output, an optimal low-dimensional projection of the data. This simple optimization framework further allows straightforward generalizations and novel variants of classical methods, which we demonstrate here by creating an orthogonal-projection canonical correlations analysis. More broadly, this survey and generic solver suggest that linear dimensionality reduction can move toward becoming a blackbox, objective-agnostic numerical technology.

Hierarchical Multi-classification with Predictive Clustering Trees in Functional Genomics

Struyf, J., Dzeroski, S. Blockeel, H. and Clare, A. (2005) Hierarchical Multi-classification with Predictive Clustering Trees in Functional Genomics. In proceedings of the EPIA 2005 CMB Workshop

Dual-orthogonal radial basis function networks for nonlinear time series prediction

A new structure of Radial Basis Function (RBF) neural network called the Dual-orthogonal RBF Network (DRBF) is introduced for nonlinear time series prediction. The hidden nodes of a conventional RBF network compare the Euclidean distance between the network input vector and the centres, and the node responses are radially symmetrical. But in time series prediction where the system input vectors are lagged system outputs, which are usually highly correlated, the Euclidean distance measure may not be appropriate. The DRBF network modifies the distance metric by introducing a classification function which is based on the estimation data set. Training the DRBF networks consists of two stages. Learning the classification related basis functions and the important input nodes, followed by selecting the regressors and learning the weights of the hidden nodes. In both cases, a forward Orthogonal Least Squares (OLS) selection procedure is applied, initially to select the important input nodes and then to select the important centres. Simulation results of single-step and multi-step ahead predictions over a test data set are included to demonstrate the effectiveness of the new approach.

Cost733cat: a database of weather and circulation type classifications

A new database of weather and circulation type catalogs is presented comprising 17 automated classification methods and five subjective classifications. It was compiled within COST Action 733 "Harmonisation and Applications of Weather Type Classifications for European regions" in order to evaluate different methods for weather and circulation type classification. This paper gives a technical description of the included methods using a new conceptual categorization for classification methods reflecting the strategy for the definition of types. Methods using predefined types include manual and threshold based classifications while methods producing types derived from the input data include those based on eigenvector techniques, leader algorithms and optimization algorithms. In order to allow direct comparisons between the methods, the circulation input data and the methods' configuration were harmonized for producing a subset of standard catalogs of the automated methods. The harmonization includes the data source, the climatic parameters used, the classification period as well as the spatial domain and the number of types. Frequency based characteristics of the resulting catalogs are presented, including variation of class sizes, persistence, seasonal and inter-annual variability as well as trends of the annual frequency time series. The methodological concept of the classifications is partly reflected by these properties of the resulting catalogs. It is shown that the types of subjective classifications compared to automated methods show higher persistence, inter-annual variation and long-term trends. Among the automated classifications optimization methods show a tendency for longer persistence and higher seasonal variation. However, it is also concluded that the distance metric used and the data preprocessing play at least an equally important role for the properties of the resulting classification compared to the algorithm used for type definition and assignment.

An integer linear programming approach for approximate string comparison

We introduce a problem called maximum common characters in blocks (MCCB), which arises in applications of approximate string comparison, particularly in the unification of possibly erroneous textual data coming from different sources. We show that this problem is NP-complete, but can nevertheless be solved satisfactorily using integer linear programming for instances of practical interest. Two integer linear formulations are proposed and compared in terms of their linear relaxations. We also compare the results of the approximate matching with other known measures such as the Levenshtein (edit) distance. (C) 2008 Elsevier B.V. All rights reserved.

Statistics of the received power for free space optical channels

Free space optical (FSO) communication links can experience extreme signal degradation due to atmospheric turbulence induced spatial and temporal irradiance fuctuations (scintillation) in the laser wavefront. In addition, turbulence can cause the laser beam centroid to wander resulting in power fading, and sometimes complete loss of the signal. Spreading of the laser beam and jitter are also artifacts of atmospheric turbulence. To accurately predict the signal fading that occurs in a laser communication system and to get a true picture of how this affects crucial performance parameters like bit error rate (BER) it is important to analyze the probability density function (PDF) of the integrated irradiance fuctuations at the receiver. In addition, it is desirable to find a theoretical distribution that accurately models these ?uctuations under all propagation conditions. The PDF of integrated irradiance fuctuations is calculated from numerical wave-optic simulations of a laser after propagating through atmospheric turbulence to investigate the evolution of the distribution as the aperture diameter is increased. The simulation data distribution is compared to theoretical gamma-gamma and lognormal PDF models under a variety of scintillation regimes from weak to very strong. Our results show that the gamma-gamma PDF provides a good fit to the simulated data distribution for all aperture sizes studied from weak through moderate scintillation. In strong scintillation, the gamma-gamma PDF is a better fit to the distribution for point-like apertures and the lognormal PDF is a better fit for apertures the size of the atmospheric spatial coherence radius ρ0 or larger. In addition, the PDF of received power from a Gaussian laser beam, which has been adaptively compensated at the transmitter before propagation to the receiver of a FSO link in the moderate scintillation regime is investigated. The complexity of the adaptive optics (AO) system is increased in order to investigate the changes in the distribution of the received power and how this affects the BER. For the 10 km link, due to the non-reciprocal nature of the propagation path the optimal beam to transmit is unknown. These results show that a low-order level of complexity in the AO provides a better estimate for the optimal beam to transmit than a higher order for non-reciprocal paths. For the 20 km link distance it was found that, although minimal, all AO complexity levels provided an equivalent improvement in BER and that no AO complexity provided the correction needed for the optimal beam to transmit. Finally, the temporal power spectral density of received power from a FSO communication link is investigated. Simulated and experimental results for the coherence time calculated from the temporal correlation function are presented. Results for both simulation and experimental data show that the coherence time increases as the receiving aperture diameter increases. For finite apertures the coherence time increases as the communication link distance is increased. We conjecture that this is due to the increasing speckle size within the pupil plane of the receiving aperture for an increasing link distance.