� Abstract— In this paper we present a novel approach for density estimation. The proposed approach is based on using the logistic regression model to get initial density estimation for the given empirical density. The empirical data does not exactly follow the logistic regression model, so, there will be a deviation between the empirical density and the density estimated using logistic regression model. This deviation may be positive and/or negative. In this paper we use a linear combination of Gaussian (LCG) with positive and negative components as a model for this deviation. Also, we will use the expectation maximization (EM) algorithm to estimate the parameters of LCG. Experiments on real images demonstrate the accuracy of our approach.

• A submitted manuscript is the version of the article upon submission and before peer-review. There can be important differences between the submitted version and the official published version of record. People interested in the research are advised to contact the author for the final version of the publication, or visit the DOI to the publisher's website. • The final author version and the galley proof are versions of the publication after peer review. • The final published version features the final layout of the paper including the volume, issue and page numbers.

xiv

paper presents the use of linear scaling and tree search in a content-based music retrieval system that can take a user's acoustic input (8-second clip of singing or humming) via a microphone and then retrieve the intended song from over 3000 candidate songs in the database. The system, known as Super MBox, demonstrates the feasibility of real-time content-based music retrieval with a high recognition rate. Super MBox first takes the user's acoustic input from a microphone and converts it into a pitch vector. Then a fast comparison engine using linear scaling and tree search is employed to compute the similarity scores. We have tested Super MBox and found the top-20 recognition rate is about 73% with about 1000 clips of test inputs from people with mediocre singing skills.

important to empower the management and staff with data warehousing-based critical thinking and knowledge management tools for strategic decision-making. Many healthcare organizations struggle with the utilization of data collected through an organization’s on-line transaction processing (OLTP) system, which is not integrated for decision-making and pattern analysis. Do you know what your top 10 paid procedures are? Do you know the DRGs that have the lowest reimbursement from Medicare/ Medicaid? Are you receiving timely and accurate reports for your analysis demands? Is your legacy system effective in identifying patterns to maximize your operational output? Are you drowning in data instead of sailing on the power of knowledge? This article focuses on data mining technology and its ability to uncover hidden and unexpected patterns in data for strategic decision-making in healthcare.

\kdtree \citefriedman1976algorithm has long been deemed unsuitable for exact nearest-neighbor search in high dimensional data. The theoretical guarantees and the empirical performance of \kdtree do not show significant improvements over brute-force nearest-neighbor search in moderate to high dimensions. \kdtree has been used relatively more successfully for approximate search \citemuja2009flann but lack theoretical guarantees. In the article, we build upon randomized-partition trees \citedasgupta2013randomized to propose \kdtree based approximate search schemes with $O(d łog d + łog n)$ query time for data sets with n points in d dimensions and rigorous theoretical guarantees on the search accuracy. We empirically validate the search accuracy and the query time guarantees of our proposed schemes, demonstrating the significantly improved scaling for same level of accuracy.

Zero-shot recognition (ZSR) refers to predict the categories of the unseen new images that never appear in the training set. Seen classes belong to the training set and unseen classes belong to the test set. Typically, most of the previous works focus on solving how to embed the space between visual and semantic representation. In this paper, we argue that the salience of the image can provide more discriminative information for ZSR. We propose a framework that first detects the salient objects in image, and then locates the most interested salient region and extracts the features in it. We use the convolutional neural network (CNN) to extract the visual features of the salient region and maps them to both user-defined and latent attributes. Ablation study on two challenging ZSL datasets is conducted. The experimental results demonstrate the salient latent features extracting from the salient origins can improve the recognition accuracy in a great deal.

With the rise of mobile phones and the popularity of the Internet, 5G, and e-commerce, there should be more various tasks processed on smartphone, which requires a good scheduling strategy to handle running processes. Compared with a single core, one of the biggest advantages of multi-core is the efficiency of running processes. In the multi-core, each core is responsible for serving a process at each moment, hence the performance of smartphone executing programs can become much better. This paper focuses on designing a branch-and-bound strategy for process scheduling in multi-core smartphone. Since the shifts of cores must be allocated by the operating system (like Android or iOS), the scheduler in the operating system would adopt the proposed scheme to improve efficiency of process scheduling.

With the rapid proliferation of data platforms collecting and curating data related to various domains such as governments data, education data, environment data or product ratings, more and more data are available online. This offers an unparalleled opportunity to study the behavior of individuals and the interactions between them. In the political sphere, being able to query datasets of voting records provides interesting insights for data journalists and political analysts. In particular, such data can be leveraged for the investigation of exceptionally consensual/controversial topics. Consider data describing the voting behavior in the European Parliament (EP). Such a dataset records the votes of each member (MEP) in voting sessions held in the parliament, as well as information on the parliamentarians (e.g., gender, national party, European party alliance) and the sessions (e.g., topic, date). This dataset offers opportunities to study the agreement or disagreement of coherent subgroups, especially to highlight unexpected behavior. It is to be expected that on the majority of voting sessions, MEPs will vote along the lines of their European party alliance. However, when matters are of interest to a specific nation within Europe, alignments may change and agreements can be formed or dissolved. For instance, when a legislative procedure on fishing rights is put before the MEPs, the island nation of the UK can be expected to agree on a specific course of action regardless of their party alliance, fostering an exceptional agreement where strong polarization exists otherwise. In this thesis, we aim to discover such exceptional (dis)agreement patterns not only in voting data but also in more generic data, called behavioral data, which involves individuals performing observable actions on entities. We devise two novel methods which offer complementary angles of exceptional (dis)agreement in behavioral data: within and between groups. These two approaches called Debunk and Deviant, ideally, enables the implementation of a sufficiently comprehensive tool to highlight, summarize and analyze exceptional comportments in behavioral data. We thoroughly investigate the qualitative and quantitative performances of the devised methods. Furthermore, we motivate their usage in the context of computational journalism.

With the rapid development of information technology and the arrival of the era of big data, people’s access to information is increasingly relying on information such as images. Today, image data are showing an increasing trend in the form of an index. How to use deep learning models to extract valuable information from massive data is very important. In the face of such a situation, people cannot accurately and timely find out the information they need. Therefore, the research on image retrieval technology is very important. Image retrieval is an important technology in the field of computer vision image processing. It realizes fast and accurate query of similar images in image database. The excellent feature representation not only can represent the category information of the image but also capture the relevant semantic information of the image. If the neural network feature learning expression is combined with the image retrieval field, it will definitely improve the application of image retrieval technology. To solve the above problems, this article studies the problems encountered in deep learning neural network feature learning based on image self-encoding and discusses its feature expression in the field of image retrieval. By adding the spatial relationship information obtained by image self-encoding in the neural network training process, the feature expression ability of the selected neural network is improved, and the neural network feature learning based on image coding is successfully applied to the popular field of image retrieval.

With the introduction of Web 2.0, there has been an extreme increase in the popularity of social bookmarking systems and folksonomies. In this paper, our motive is to develop a recommender system that is based on user assigned tags and content present on web pages. Although the tag recommendations in social tagging systems can be very accurate and personalized, there exists an issue of risk to the privacy of user's profile, since the social tags are given by a user expose his preferences to other users in contact. To overcome this problem, we have incorporated obfuscation privacy strategies with the well-known Delicious dataset in social tagging based recommender system. We have applied the popular supervised machine-learning algorithm, K-Nearest Neighbours classifier to the dataset that recommends relevant tags to the user. Privacy has been introduced in our tag-based recommender system by hiding some of the necessary tags, bookmarks of a user and replacing them with some random tags and bookmarks. Our experiment results indicate that the recommender system being implemented is highly efficient in terms recall and privacy measure for different values of k. The results and comparisons indicate that we have successfully employed an effective tag recommender system, which also protects the user's privacy without any significant fall in the quality of recommendation.

With the increased usage of computer networks, network intrusions have greatly threatened the Internet infrastructures. Traditional signature-based intrusion detection often suffers from an ineffectivity to those previously “unseen” attacks. In this paper, we analyze the network intrusions from a new viewpoint based on data field and propose branch and bound tree to lessen computation complexity. Finally, we evaluated our approach over KDD Cup 1999 data set.

With the flowering of graph mining and computation technology, the field of graph clustering has become popular. Particularly, to cluster increasingly massive data represented as graph become common today. There have been many research efforts on graph clustering algorithms as well as reported applications. However, graph clustering technology is highly application specific and computationally expensive especially when applied to huge data. In this paper, we propose an efficient graph contraction algorithm for speeding up graph clustering. We target at parallelizing edge contraction for reducing computing time for graph mining. We adopt massive multithreading for edge contraction on huge non planar graph. In experiment, our algorithm achieves good parallel efficiency for computation time of both contracting edges and clustering graph as the number of threads increases. Our edge contraction algorithm achieves the speedup ranging from 3 to 6 in contracting 10000 to 30000 edges. Finally, we observe that with edge contraction, parallel clustering achieves speed up is 8 to 80 times faster compared with the result of no edge contraction.

With the advancement of Unmanned Aerial Vehicles (UAVs) and their application in the most diverse areas, the demand for increasingly better precision in their positioning and navigation task has arisen. Based on this context, this article proposes a new approach for localization and navigation UAVs using topological maps and Convolutional Neural Networks (CNN). CNN is used as features extractor, according to the concept of Transfer Learning. The use of topological maps helps to guide the vehicle through the exploration environment. To evaluate the performance of the approach, parameters such as Accuracy, F1-Score, and processing time are considered. For the classification were used Bayesian Classifier, k-Nearest Neighbor (kNN), Multi-layer Perceptron (MLP), Optimum-Path Forest (OPF) and Support Vector Machine (SVM). The results show that CNN achieved 99.97% Accuracy and also F1-Score in combination with most of the considered classifiers, proving the effectiveness of our approach.

With heightened security concerns across the globe and the increasing need to monitor, preserve and protect infrastructure and public spaces to ensure proper operation, quality assurance and safety, numerous video cameras have been deployed. Accordingly, they also need to be monitored effectively and efficiently. However, relying on human operators to constantly monitor all the video streams is not scalable or cost effective. Humans can become subjective, fatigued, even exhibit bias and it is difficult to maintain high levels of vigilance when capturing, searching and recognizing events that occur infrequently or in isolation. These limitations are addressed in the Live Video Database Management System (LVDBMS), a framework for managing and processing live motion imagery data. It enables rapid development of video surveillance software much like traditional database applications are developed today. Such developed video stream processing applications and ad hoc queries are able to “reuse” advanced image processing techniques that have been developed. This results in lower software development and maintenance costs. Furthermore, the LVDBMS can be intensively tested to ensure consistent quality across all associated video database applications. Its intrinsic privacy framework facilitates a formalized approach to the specification and enforcement of verifiable privacy policies. This is an important step towards enabling a general privacy certification for video surveillance systems by leveraging a standardized privacy specification language. With the potential to impact many important fields ranging from security and assembly line monitoring to wildlife studies and the environment, the broader impact of this work is clear. The privacy framework protects the general public from abusive use of surveillance technology;

When using k-nearest neighbors, an unknown object is classified by comparing it to all the prototypes stored in the training database. When the size of the database is large, and especially if prototypes are represented by graphs, the search of k-nearest neighbors can be very time consuming. On this basis, some researchers have tried to propose optimization techniques to speed up or to approximate the search of the nearest neighbors of a query. However, these studies pay attention only to the case of vector space. In this paper, we propose an optimization technique dedicated to structural pattern recognition. We take advantage of a recent branch-and-bound graph edit distance approach in order to speed up the classification stage. Instead of considering each graph edit distance problem as an independent search tree, the search trees whose purpose is to classify an unknown graph are considered as a one search tree. Results showed that this approach drastically outperformed the classical one under limited time constraints. Moreover, this approach beat fast graph matching algorithms in terms of average execution time.

When planning a new development (facility/service site), location decisions are always the major issue. In this paper we introduce a novel query capacity constraint MaxBRNN, which can solve the facility location selection problem efficiently.The MaxBRNN (maximizing BRNN) query is based on bichromatic reverse nearest neighbor (BRNN) query which uses the number of reverse nearest customers to model the influence of a facility location. The MaxBRNN query has been appreciated extensively in spatial database studies because of its great potential in real life applications, such as, markets decision, sensor network clustering and the design of GSM (global system for mobile communication). The existing researches mostly suppose that the service facility's capacity is unlimited. However, in real cases, facilities are inevitably constrained by designed capacities. For example, if the government wants to select a new place to set up an emergency center to share the existing centers' patients, they need to know the current emergency centers' capacity so that they can estimate the new center's scale. Thus, the capacity constrained MaxBRNN query is significantly important in planning a new development. As far as we know, the capacity constrained MaxBRNN query has not been studied yet, so, we formulate this problem, propose a basic solution and develop some efficient algorithms for the query.Our major contributions are as follows: (1) we propose a novel query capacity constraint MaxBRNN which can solve the facility location selection problem effectively and efficiently; (2) we develop a basic algorithm CCMB and two improved algorithms which can find out the optimal region in terms of building a new facility, maximize its impact and deal with the complicated reassignment when adding new facilities into the dataset; (3) we prove the algorithms' effectiveness and efficiency by extensive experiments using both real and synthetic data sets. We propose and formalize the capacity constrained MaxBRNN query.We propose a basic algorithm CCMB which can solve the problem efficiently.We develop two improved methods: Prog-CCMB and Pruning-CCMB.We prove the algorithms' effectiveness and efficiency for facility selection query.

When given a collection of graphs on over-lapping, but possibly non-identical, vertex sets, many inference tasks involve learning the across-network alignment. Often, this task is difficult due to the enormity of the graphs, as network alignment (that is, graph matching) is NP-hard in its most general form. That being said, often a partial alignment is known a priori. We leverage the knowledge of this partial alignment (seeds) to assist the graph matching task. We further present principled methods of handling non-identical vertex sets and multiple-node alignment for both the tasks of graph matching and vertex nomination. We demonstrate the applicability of our methodologies via synthetic and real-data experiments. In particular, we show a real-time example of network alignment of the leftand right-hemispheres for the Drosophila larvae connectome and demonstrate that our structure-based algorithm improves their currently used techniques. Further, inspired by the nuances inherent in this data, and in particular the presence of neural type-based correlation, we present a network model which is robust enough to handle this type of data while still being practical in the sense that statistical theory is still relevant. We compare this model ii ABSTRACT to other network models and present relevant existing theory.to other network models and present relevant existing theory. Primary Reader: Carey E. Priebe Secondary Reader: Vince Lyzinski

Wearable devices have recently gained a foothold in the market with the uptake of smartwatches. The strong tie between a smartwatch and its owner, the highly predictable position of a smartwatch on the body, and its internal sensors are enabling a wide array of applications that leverage the user context. In this paper we focus on a gesture recognition system to augment the user interaction with multimedia applications. We define a set of seven gestures that are relevant across several applications and we collect an extensive dataset with two smartwatches (the Motorola Moto360 and Apple's Watch). We use Long Short Term Memory neural networks for gesture recognition based on sensor data from both smartwatches. We provide an extensive evaluation of the classification accuracy of the system and provide a sensitivity analysis to find the Long Short Term Memory configuration that maximizes the classification accuracy. We also show the extent to which Long Short Term Memory neural networks outperform traditional machine learning approaches. We also illustrate an application we built for the Android and iOS platforms that allows developers to easily integrate the gesture recognition in their own systems. We conclude the paper with a description of use cases to underscore the potential impact of our contribution.

We survey some of the main Crisp and Fuzzy (Computational) Intelligence works, which have been developed for the solution of the real world production planning and scheduling problems. Development of better production planning and scheduling can help many industries, since timely deliveries and good quality of products are important for customers, as well as the special goals of manufacturers, e.g., the reduction of the inventory costs, improvement of the utilization of the productive capacity, synchronization of the different sections of the plants, etc. However, attaining an optimal scheduling of production in different industries is very hard, because of the conflicts among different goals. The examples of such conflicting goals are: the optimization of the production rate, production quality, intra-order production consistency, etc. Within the last thirty years or so, it has been found that fuzzy set theory incorporated with fuzzy cluster analysis is a suitable tool to analyse and diagnose these sorts of complicated problems.