Search

Influence maximization is a crucial problem in the analysis of large and complex networks. The problem asks for a set of nodes of a network from which message dissemination is maximized. This paper focuses on influence maximization from samples (IMFS) where the network is hidden. Instead, the input is a set of input-output samples of the influence function over a network. The recently proposed algorithm by Balkanski, Rubinstein, and Singer (BRS) for monotone submodular optimization achieves theoretically-tight optimization ratio. Empirically, however, no work is done to evaluate the actual performance of the BRS algorithm against parameters such as sample size, sample set size, and network topology. This paper provides an empirical analysis of algorithms for IMFS. In particular, we extends BRS by (1) factoring in effects caused by the complex network topology when approximating the marginal contribution of nodes and (2) removing constraints on the sampling distribution. We conduct experiments using two information diffusion models, and samples generated on both synthetic random networks and real-world networks. In general, our proposed algorithm significantly improves the output quality from the BRS algorithm.

Path planning problem is one of the key contents in the research field of Geographic Information System (GIS), among which the finding of the shortest path is always a hot topic. When there is large quantity of data, the efficiency of traditional algorithms of Kth shortest paths is relatively low, and some path problems concerning big difference between the planned Kth paths under actual requirements cannot be solved. On the basis of Dijkstra algorithm, the concepts of favorability and repeatability are introduced to find Kth paths including the shortest in the current graph cyclically by determination of repeatability of path results and the change of graph caused by the change of favorability, thereby realizing the planning of multiple different paths. Compared with similar algorithms, this algorithm is faster, and the multiple path results obtained meet the requirement for certain repeatability while the lengths of paths are also quite reasonable.

Publish/subscribe paradigm is seen as a good candidate in constructing distributed applications due to its unique and powerful features such as anonymity, asynchrony and many-to-many communication. As applications based on publish/subscribe increases greatly it is required to support different quality of service properties. However, most current solutions adopt simple matching strategies without distinguishing crucial tasks from non-crucial ones, which lead to poor performance of key tasks when publish/subscribe systems are in heavy load. In this paper, we introduce the notion of subscription priority, and propose a new marginal revenue based scheduling (MRBS) algorithm which considers subscription priority. MRBS uses priorities of attribute constraints to sort the matching order for an event, and in such a way MRBS can distinguish important subscriptions from less important ones. MRBS is proven to be able to produce the maximum event matching revenue. A priority-sensitive algorithm for subscription information maintenance is also given. Finally, we conduct detailed experimental evaluation to verify the effectiveness of our approach.

Feature extraction plays a very important role in object recognition and categorization. In this paper, we present a method for extracting corner-based feature from images. This feature is invariant to image scale and rotation, and is shown robust to addition of noise and changes in 3D viewpoint. This paper also describes an approach to using the feature for object recognition. As baselines for comparison, we implemented three additional recognition systems using signature, moment invariant and Fourier descriptor as features. They provide a good basis for judging the importance of representation in learning. The performance analysis on the obtained experimental results demonstrates that the proposed method is effective and efficient.

This paper presents a novel method for recognizing objects using corner features. As one of the most important local features, corner contains lots of information with the shape of the objects. Our approach detects corner points from images and describes the objects based on them. The features are highly distinctive, in the sense that a single feature can be correctly matched with high probability against a large database of features from many images. The recognition proceeds by matching individual features to a database of features from known objects using a fast nearest-neighbor algorithm. The performance on the obtained experimental results demonstrates that the proposed method is efficient.

Diverse pose estimation of three-dimensional (3D) object in the whole view-space remains a challenge in the field of pattern recognition. In this paper, a pose estimation algorithm of 3D object named isomap-eigenanalysis-regression (Isomap-E-R), which estimates arbitrary pose of 3D object in the whole view space, is proposed. For the training set, the low-dimensional embedding of input pattern set is computed by isomap, and the eigen-images of the embedding are deduced on the basis of an eigenspace. A different projection direction in low dimensional embedding is utilized to improve the accuracy of pose estimation. The metrics on each direction, derived by linear regression, is then used to further deduce the projection of the training set. For a given new sample, its projection onto the eigen-images is first computed, and the training images nearest to those deduced for the new sample by the algorithm give the estimation poses. The performance analysis on the obtained experimental results demonstrated that the proposed method could estimate the diverse pose of 3D object with significant efficiency and precision. Finally, the algorithm can be also extended to real-time pose estimate of 3D object and other potential applications.

There are many algorithms to simplify meshes. But, most of them are to realize real-time rendering circumstance or models in virtual reality. This paper introduces a novel scheme to simplify meshes from a different view point--visual resolution. This paper uses visual recognizing object principle to simplify meshes. The model or scheme gives a reasonable standard which decides what extent the 3D model is simplified. This scheme can be used in the 3D object-recognition and the robot vision. In this paper, 3D model is the convex model.

A method is presented to obtain maps between the high-dimensional data and the low-dimensional space deduced by locally linear embedding (LLE). Since LLE does not provide a parametric function that build maps between the image space and the low-dimensional manifold. In this paper, multivariate linear regression is applied to deduce the maps. It can successfully project a new data point onto the embedded space. Also it can be extended to supervised LLE. The performance analysis on the obtained experimental results demonstrated that the proposed method is effective and efficient.

Diverse pose estimation of 3D object in the whole view-space is a problem perplexed many researchers. In this paper we propose an algorithm extended from LLE which can estimate the arbitrary pose of 3D object in the whole view space. First, we compute the eigen-images of training set by introducing the idea of PCA using the low-dimensional embedding coordinate deduced from LLE. For a new sample we can compute its projection to the eigen-images, and the nearest training images from the new sample are the estimation poses. Next, we set different weight for different projection direction depends on its eigen-value when computing the distance between the new sample and the training images. Experimental results obtained demonstrated that the performance of the proposed method could estimate the diverse pose of 3D object efficiently and precisely, also our algorithm can be extended to real-time pose estimate, is of a potential future.