Support Vector Machines (SVM) project feature vectors into a linear or non-linear state space using kernel function(s) and attempts to maximize the margin between classes. The projection of feature vectors into a high dimensional hyperspace structure helps to provide sparse separable clusters of data. Biometric data such as fingerprints transformed into feature vectors are candidates for support vector machine sparse classification. Fingerprint position and ridge flow pattern classification provide a feature vector for a kernel function(s). As samples are projected into a hyperspace construct, fingerprint identification velocity will improve while performance will increase or remain constant. This paper discusses and shows the results of a fingerprint identification support vector machine within a novel hyperspace structure. The support vector machine is formalized with a high dimensional hyperspace structure with an internal bootstrapped c-means clustering algorithm and probabilistic neural network (PNN). The National Institute of Standards and Technology (NIST) provided data set Special Database 14 for all experiments.

This paper presents a ranking framework for content-based image retrieval using relevance feature mapping. Each relevance feature measures the relevance of an image to some profile underlying the image database. The framework is a two-stage process. In the off-line modeling stage, it constructs a collection of models which maps all images in the database to the relevance feature space. In the on-line retrieval stage, it assigns a weight to every relevance feature based on the query image, and then ranks images in the database according to their weighted average feature values. The framework also incorporates relevance feedback which modifies the ranking based on the feedbacks through reweighted features. We show that the power of the proposed framework is coming from the relevance features. Experiments on a large image database validate the efficacy and efficiency of the proposed framework.

Visual information retrieval systems have gained importance due to the increasing amount of available digital multimedia data. Local features employing a bag of words approach from text retrieval have outperformed global features and have enhanced retrieval performance in large data sets. In this paper we conduct an exploratory study revisiting the bag of visual words approach for content based image retrieval. We apply a fuzzy clustering technique for visual words creation and visual words assignment and show in a first attempt that fuzzy clustering leads to more robust results in terms of retrieval performance.

Twitter is a user-generated content system that allows its users to share short text messages, called tweets, for a variety of purposes, including daily conversations, URLs sharing and information news. Considering its world-wide distributed network of users of any age and social condition, it represents a low level news flashes portal that, in its impressive short response time, has the principal advantage.

In this paper we recognize this primary role of Twitter and we propose a novel topic detection technique that permits to retrieve in real-time the most emergent topics expressed by the community. First, we extract the contents (set of terms) of the tweets and model the term life cycle according to a novel aging theory intended to mine the emerging ones. A term can be defined as emerging if it frequently occurs in the specified time interval and it was relatively rare in the past. Moreover, considering that the importance of a content also depends on its source, we analyze the social relationships in the network with the well-known Page Rank algorithm in order to determine the authority of the users. Finally, we leverage a navigable topic graph which connects the emerging terms with other semantically related keywords, allowing the detection of the emerging topics, under user-specified time constraints. We provide different case studies which show the validity of the proposed approach.

Support Vector Machine Clustering (SVMC) is a model-based clustering method designed primarily for solving 2-class clustering problems. In this paper, we generalize the SVMC method to multi-class clustering via two different strategies, namely One-Against-All and hierarchical clustering. We applied the resulting multi-class SVMC techniques to large scale image clustering based on the visual keywords representation and Histogram Intersection Kernel. Experiments on two benchmark databases show that compared with traditional Support Vector Clustering (SVC) method, our proposed approach is particularly suited to large scale data and large number of classes clustering problems, in terms of computational efficiency and clustering quality.

In this paper, we develop a method for retrieving events of interest in a video archive. To this end, we address the following two issues. First, due to camera techniques, locations and so on, shots of an event contain significantly different features. So, they cannot be retrieved by a single retrieval model. Thus, we use "rough set theory" to extract multiple classification rules, each of which correctly identifies a subset of shots of the event. Second, although concepts like Person, Car and Cityspace are useful for event retrieval, we need to distinguish between relevant concepts to an event and irrelevant ones. Otherwise, the retrieval performance degrades. So, in order to select concepts relevant to the event, we organize concepts into "video ontology" which is a formal and explicit specification of concepts, concept properties and relations among concepts. Experimental results show both the effectiveness of rough set theory and the one of video ontology.

Owing to the rapid mounting of massive image data, image classification has attracted lots of research efforts. Several diverse research disciplines have been confluent on this important theme, looking for more powerful solutions. In this paper, we propose a novel image representation method B2S (Bag to Set) that keeps all frequency information and is more discriminative than traditional histogram based bag representation. Based on B2S, we construct two different image classification approaches. First, we apply B2S to a state-of-the-art image classification algorithm SPM in computer vision. Second, we design a framework DisIClass (Discriminative Frequent Pattern-Based Image Classification) to utilize data mining algorithms to classify images, which was hardly done before due to the intrinsic differences between the data of computer vision and data mining fields. DisIClass adapts the locality property of image data, and apply sequential covering method to induce the most discriminative feature sets from a closed frequent item set mining method. Our experiments with real image data show the high accuracy and good scalability of both approaches.

Images on the Web appear with textual contents providing meaningful information to their semantics. Methods that automatically determine and extract the Web image context from an HTML document are widely used in different applications. However, the performance of the image context extraction has rather been evaluated on its own. Keeping this imperative in mind, we present a framework to objective evaluation and comparison of the performance of image context extraction methods. This is achieved by collecting a large ground truth dataset consisting of diverse Web documents from real Web servers and by defining performance measures adapted to fit the special properties of the context extraction task. To show the capabilities of the proposed framework, common extraction methods from the literature have been evaluated and the results are summarized in this paper.

This work explores computer vision applications of the MapReduce framework that are relevant to the data mining community. An overview of MapReduce and common design patterns are provided for those with limited MapReduce background. We discuss both the high level theory and the low level implementation for several computer vision algorithms: classifier training, sliding windows, clustering, bag-of-features, background subtraction, and image registration. Experimental results for the k-means clustering and single Gaussian background subtraction algorithms are performed on a 410 node Hadoop cluster.

The problem of identifying frequently occurring patterns, or motifs, in time series data has received a lot of attention in the past few years. Most existing work on finding time series motifs require that the length of the patterns be known in advance. However, such information is not always available. In addition, motifs of different lengths may co-exist in a time series dataset. In this work, we propose a novel approach, based on grammar induction, for approximate variable-length time series motif discovery. Our algorithm offers the advantage of discovering hierarchical structure, regularity and grammar from the data. The preliminary results are promising. They show that the grammar-based approach is able to find some important motifs, and suggest that the new direction of using grammar-based algorithms for time series pattern discovery might be worth exploring.