Even though clustering trajectory data attracted considerable attention in the last few years, most of prior work assumed that moving objects can move freely in an euclidean space and did not consider the eventual presence of an underlying road network and its influence on evaluating the similarity between trajectories. In this paper, we present two approaches to clustering network-constrained trajectory data. The first approach discovers clusters of trajectories that traveled along the same parts of the road network. The second approach is segment-oriented and aims to group together road segments based on trajectories that they have in common. Both approaches use a graph model to depict the interactions between observations w.r.t. their similarity and cluster this similarity graph using a community detection algorithm. We also present experimental results obtained on synthetic data to showcase our propositions.

Modelling signals as sparse linear combinations of atoms selected from a dictionary has become a popular paradigm in many fields, including signal processing, statistics, and machine learning. This line of research has witnessed the development of several well-founded theoretical frameworks (see, e.g., Wainwright [2009], Zhang [2009]) and efficient algorithmic tools (see, e.g., Bach et al. [2011] and references therein). However, the performance of such approaches hinges on the representation of the signals, which makes the question of designing "good" dictionaries prominent. A great deal of effort has been dedicated to come up with efficient predefined dictionaries, e.g., the various types of wavelets [Mallat, 2008]. These representations have notably contributed to many successful image processing applications such as compression, denoising and deblurring. More recently, the idea of simultaneously learning the dictionary and the sparse decompositions of the signals--also known as sparse dictionary learning, or simply, sparse coding--has emerged as a powerful framework, with state-of-the-art performance in many tasks, including inpainting and image classification (see, e.g., Mairal et al. [2010] and references therein). Although sparse dictionary learning can sometimes be formulated as convex[Bach et al., 2008, Bradley and Bagnell, 2009], nonparametric Bayesian [Zhou et al., 2009] and submodular [Krause and Cevher, 2010] problems, the most popular and widely used definition of sparse coding brings into play a non-convex optimization problem. Despite its empirical and practical success, there has only been little theoretical analysis of the properties of sparse dictionary learning.

We propose a version of least-mean-square (LMS) algorithm for sparse system identification. Our algorithm called online linearized Bregman iteration (OLBI) is derived from minimizing the cumulative prediction error squared along with an l1-l2 norm regularizer. By systematically treating the non-differentiable regularizer we arrive at a simple two-step iteration. We demonstrate that OLBI is bias free and compare its operation with existing sparse LMS algorithms by rederiving them in the online convex optimization framework. We perform convergence analysis of OLBI for white input signals and derive theoretical expressions for both the steady state and instantaneous mean square deviations (MSD). We demonstrate numerically that OLBI improves the performance of LMS type algorithms for signals generated from sparse tap weights.

We investigate a robust penalized logistic regression algorithm based on a minimum distance criterion. Influential outliers are often associated with the explosion of parameter vector estimates, but in the context of standard logistic regression, the bias due to outliers always causes the parameter vector to implode, that is shrink towards the zero vector. Thus, using LASSO-like penalties to perform variable selection in the presence of outliers can result in missed detections of relevant covariates. We show that by choosing a minimum distance criterion together with an Elastic Net penalty, we can simultaneously find a parsimonious model and avoid estimation implosion even in the presence of many outliers in the important small $n$ large $p$ situation. Implementation using an MM algorithm is described and performance evaluated.

Hyperspectral images (HIS) classification is a high technical remote sensing tool. The goal is to reproduce a thematic map that will be compared with a reference ground truth map (GT), constructed by expecting the region. The HIS contains more than a hundred bidirectional measures, called bands (or simply images), of the same region. They are taken at juxtaposed frequencies. Unfortunately, some bands contain redundant information, others are affected by the noise, and the high dimensionality of features made the accuracy of classification lower. The problematic is how to find the good bands to classify the pixels of regions. Some methods use Mutual Information (MI) and threshold, to select relevant bands, without treatment of redundancy. Others control and eliminate redundancy by selecting the band top ranking the MI, and if its neighbors have sensibly the same MI with the GT, they will be considered redundant and so discarded. This is the most inconvenient of this method, because this avoids the advantage of hyperspectral images: some precious information can be discarded. In this paper we'll accept the useful redundancy. A band contains useful redundancy if it contributes to produce an estimated reference map that has higher MI with the GT.nTo control redundancy, we introduce a complementary threshold added to last value of MI. This process is a Filter strategy; it gets a better performance of classification accuracy and not expensive, but less preferment than Wrapper strategy.

In this paper we study general $l_p$ regularized unconstrained minimization problems. In particular, we derive lower bounds for nonzero entries of first- and second-order stationary points, and hence also of local minimizers of the $l_p$ minimization problems. We extend some existing iterative reweighted $l_1$ (IRL1) and $l_2$ (IRL2) minimization methods to solve these problems and proposed new variants for them in which each subproblem has a closed form solution. Also, we provide a unified convergence analysis for these methods. In addition, we propose a novel Lipschitz continuous $\epsilon$-approximation to $\|x\|^p_p$. Using this result, we develop new IRL1 methods for the $l_p$ minimization problems and showed that any accumulation point of the sequence generated by these methods is a first-order stationary point, provided that the approximation parameter $\epsilon$ is below a computable threshold value. This is a remarkable result since all existing iterative reweighted minimization methods require that $\epsilon$ be dynamically updated and approach zero. Our computational results demonstrate that the new IRL1 method is generally more stable than the existing IRL1 methods [21,18] in terms of objective function value and CPU time.

For such Gaussian graphical models (GGMs), it is usually assumed that a given variable can bepredicted by a small numberof other variables. This assumption implies that the precision matrix is sparse. Therefore estimating Gaussian graphical model can be reduced to the problem of estimating a sparse precision matrix. One approach to sparse precision matrix estimation is covariance selection or neighborhood selection (Dempster, 1972; Meinshausen & Bühlmann, 2006), which tries to estimate each row (or column) of the precision matrix by predicting the corresponding variable using a sparse linear combination of other variables. An alternative formulation is maximum-likelihood estimation method that directly estimate the full precision matrix.

Typical cohorts in brain imaging studies are not large enough for systematic testing of all the information contained in the images. To build testable working hypotheses, investigators thus rely on analysis of previous work, sometimes formalized in a so-called meta-analysis. In brain imaging, this approach underlies the specification of regions of interest (ROIs) that are usually selected on the basis of the coordinates of previously detected effects. In this paper, we propose to use a database of images, rather than coordinates, and frame the problem as transfer learning: learning a discriminant model on a reference task to apply it to a different but related new task. To facilitate statistical analysis of small cohorts, we use a sparse discriminant model that selects predictive voxels on the reference task and thus provides a principled procedure to define ROIs. The benefits of our approach are twofold. First it uses the reference database for prediction, i.e. to provide potential biomarkers in a clinical setting. Second it increases statistical power on the new task. We demonstrate on a set of 18 pairs of functional MRI experimental conditions that our approach gives good prediction. In addition, on a specific transfer situation involving different scanners at different locations, we show that voxel selection based on transfer learning leads to higher detection power on small cohorts.

We consider the problem of estimating a sparse multi-response regression function, with an application to expression quantitative trait locus (eQTL) mapping, where the goal is to discover genetic variations that influence gene-expression levels. In particular, we investigate a shrinkage technique capable of capturing a given hierarchical structure over the responses, such as a hierarchical clustering tree with leaf nodes for responses and internal nodes for clusters of related responses at multiple granularity, and we seek to leverage this structure to recover covariates relevant to each hierarchically-defined cluster of responses. We propose a tree-guided group lasso, or tree lasso, for estimating such structured sparsity under multi-response regression by employing a novel penalty function constructed from the tree. We describe a systematic weighting scheme for the overlapping groups in the tree-penalty such that each regression coefficient is penalized in a balanced manner despite the inhomogeneous multiplicity of group memberships of the regression coefficients due to overlaps among groups. For efficient optimization, we employ a smoothing proximal gradient method that was originally developed for a general class of structured-sparsity-inducing penalties. Using simulated and yeast data sets, we demonstrate that our method shows a superior performance in terms of both prediction errors and recovery of true sparsity patterns, compared to other methods for learning a multivariate-response regression.

Covering-based rough set theory is a useful tool to deal with inexact, uncertain or vague knowledge in information systems. Geometric lattice has widely used in diverse fields, especially search algorithm design which plays important role in covering reductions. In this paper, we construct four geometric lattice structures of covering-based rough sets through matroids, and compare their relationships. First, a geometric lattice structure of covering-based rough sets is established through the transversal matroid induced by the covering, and its characteristics including atoms, modular elements and modular pairs are studied. We also construct a one-to-one correspondence between this type of geometric lattices and transversal matroids in the context of covering-based rough sets. Second, sufficient and necessary conditions for three types of covering upper approximation operators to be closure operators of matroids are presented. We exhibit three types of matroids through closure axioms, and then obtain three geometric lattice structures of covering-based rough sets. Third, these four geometric lattice structures are compared. Some core concepts such as reducible elements in covering-based rough sets are investigated with geometric lattices. In a word, this work points out an interesting view, namely geometric lattice, to study covering-based rough sets.