Search

Contact Logics provide a natural framework for representing and reasoning about regions in several areas of computer science. In this paper, we focus our attention on reasoning methods for Contact Logics and address the satisfiability problem and the unifiability problem. Firstly, we give sound and complete tableaux-based decision procedures in Contact Logics and we obtain new results about the decidability/complexity of the satisfiability problem in these logics. Secondly, we address the computability of the unifiability problem in Contact Logics and we obtain new results about the unification type of the unifiability problem in these logics.

A common task of Web users is querying structured information from Web pages. In this paper we propose a novel query processor for systematically discovering any-k relations from Web search results with conjunctive queries. The 'any-k' phrase denotes that retrieved tuples are not ranked by the system.For realizing this interesting scenario the query processor transfers a structured query into keyword queries that are submitted to a search engine, forwards search results to relation extractors, and then combines relations into result tuples.Unfortunately, relation extractors may fail to return a relation for a result tuple. We propose a solid information theory-based approach for retrieving missing attribute values of partially retrieved relations. Moreover, user-defined data sources may not return at least k complete result tuples. To solve this problem, we extend the Eddy query processing mechanism [14] for our 'querying the Web' scenario with a continuous, adaptive routing model. The model determines the most promising next incomplete row for returning any-k complete result tuples at any point during the query execution process.We report a thorough experimental evaluation over multiple relation extractors. Our experiments demonstrate that our query processor returns complete result tuples while processing only very few Web pages.

The classification of color image that contain skin pixel is a challenging process. It is challenging due to certain factor such as brightness, background similar to human skin represent the obstacles thereof. The basic approach of skin detection is color based classification. The human skin image is the composition of RGB color. The data values of skin is calculated by randomly sampling RGB color values of different face images which have different age, race, genders sets. However, its performance has not really been high because of the high overlapped degree between “skin” and “non-skin” pixels. This paper describes the linear norm fuzzy based twin support vector machine (LN-FTSVM)) approach for discriminate skin and non-skin data values of skin dataset and to enhance the skin recognition performance. The concept of fuzzy is resolved the unclassified and overlapped data region problems. If no decision function is positive for a data set, this data set is classified into a class with the large membership value. By computational experiments shows that Experiments result shows that the accuracy is improved by linear norm fuzzy based TSVM (LN-FTSVM)) over the conventional methods.

To improve the intelligent image recognition abilities of edge devices, a parallel-optimization-based framework called POWER is introduced in this paper. With FPGA (Field-Programmable Gate Array) as its hardware module, POWER provides well extensibility and flexible customization capability for developing intelligent firmware suitable for different types of edge devices in various scenarios. Through an actual case study, we design and implement a firmware prototype following the specification of POWER and explore its performance improvement using parallel optimization. Our experimental results show that the firmware prototype we implement exhibits good performance and is applicable to substation inspection robots, which also validate the effectiveness of our POWER framework in designing edge intelligent firmware modules indirectly.

Searching for information about a specific person is a frequent online activity. In most cases, users are aided in the search process by queries containing a name in Web search engines. Typically, Web search engines provide just a few accurate results associated with a name-containing query. Most existing solutions for suggesting synonyms in online search are based on pattern matching and phonetic encoding, however very often, the performance of such solutions is less than optimal. In this paper, we propose SpokenName2Vec, a novel and generic algorithm which addresses the synonym suggestion problem by utilizing automated speech generation, and deep learning to produce novel spoken name embeddings. These embeddings capture the way people pronounce names in a particular language and accent. Utilizing a name’s pronunciation can help detect names that sound alike, but are written differently. We demonstrated the proposed approach on a large-scale dataset with more than 250,000 forenames and surnames and evaluated it on two ground truth datasets containing 7400 forenames and 25,000 surnames (including their verified synonyms). The performance of SpokenName2Vec was found superior to the 10 other algorithms evaluated, including phonetic encoding, string similarity, and machine learning algorithms. The results obtained emphasize the potential of spoken name embeddings for improved synonym suggestion.

Ambient Assisted Living (AAL) services are emerging as context-awareness solutions to support elderly people’s autonomy. The context-aware paradigm makes applications more user-adaptive. In this way, context and user models expressed in ontologies are employed by applications to describe user and environment characteristics. The rapid advance of technology allows creating context server to relieve applications of context reasoning techniques. Specifically, the Next Generation Networks (NGN) provides by means of the presence service a framework to manage the current user’s state as well as the user’s profile information extracted from Internet and mobile context. This paper propose a user modeling ontology for AAL services which can be deployed in a NGN environment with the aim at adapting their functionalities to the elderly’s context information and state.

Location-based services are quickly becoming immensely popular. In addition to services based on users' current location, many potential services rely on users' location history, or their spatial-temporal provenance. Malicious users may lie about their spatial-temporal provenance without a carefully designed security system for users to prove their past locations. In this paper, we present the Spatial-Temporal provenance Assurance with Mutual Proofs (STAMP) scheme. STAMP is designed for ad-hoc mobile users generating location proofs for each other in a distributed setting. However, it can easily accommodate trusted mobile users and wireless access points. STAMP ensures the integrity and non-transferability of the location proofs and protects users' privacy. A semi-trusted Certification Authority is used to distribute cryptographic keys as well as guard users against collusion by a light-weight entropy-based trust evaluation approach. Our prototype implementation on the Android platform shows that STAMP is low-cost in terms of computational and storage resources. Extensive simulation experiments show that our entropy-based trust model is able to achieve high $(>0.9)$ collusion detection accuracy.

The paper presents a novel short-range wireless sensor node architecture, based on Integral Pulse Frequency Modulator (IPFM) and Ultra-Wideband (UWB) pulse generator. Due to the lack of internal clock signal source and multi user implementation without application of a microprocessor, the architecture is simple and energy efficient. Multi-user coding is performed using delay elements, where each user has unique delay time value. The output of the IPFM is fed to the delay element. Delayed and original signal from the IPFM are feeding UWB pulse generator. The paper presents the transmitter architecture and measurements of the transmitter signals in time and frequency domain.

International audience; This paper addresses the localization and super-resolutionof license plate in a unifying approach. Higher qualitylicense plate can be obtained using super-resolution onsuccessive lower resolution plate images. All existingmethods assume that plate zones are correctly extractedfrom every frame. However, the accurate localization needsa sufficient quality of the image, which is not always truein real video. Super-resolution on all pixels is a possiblebut much time consuming alternative. We propose a frameworkwhich interlaces successfully these two modules. First,coarse candidates are found by an weak but fast licenseplate detection based on edge map sub-sampling. Then, animproved fast MAP-based super-resolution, using local phaseaccurate registration and edge preserving prior, is appliedon these regions of interest. Finally, our robustICHT-based localizer rejects false-alarms and localizes thehigh resolution license plate more accurately. Experimentswhich were conducted on synthetic and real data, proved therobustness of our approach with real-time possibility.

In this study a multi-objective problem considering uncertainty and flexibility of job sequence in an automated flexible job shop (AFJS) is considered using manufacturing simulation. The AFJS production system is considered as a complex problem due to automatic elements requiring planning and optimization. Several solution approaches are proposed lately in different categories of meta-heuristics, combinatorial optimization and mathematically originated methods. This paper provides the metamodel using simulation optimization approach based on multi-objective efficiency. The proposed metamodel includes different general techniques and swarm intelligent technique to reach the optimum solution of uncertain resource assignment and job sequences in an AFJS. In order to show the efficiency and productivity of the proposed approach, various experimental scenarios are considered. Results show the optimal resources assignment and optimal job sequence which cause efficiency and productivity maximization. The makespan, number of late jobs, total flow time and total weighted flow time minimization have been resulted in an automated flexible job shop too.

The increasing transistor scale integration poses, among others, the thermal-aware floorplanning problem consisting of how to place the hardware components in order to reduce overheating by dissipation. Because of the huge amount of feasible floorplans, most of the solutions found in the literature include an evolutionary algorithm for, either partially or completely, carrying out the task of floorplanning. Evolutionary algorithms usually have a bottleneck in the fitness evaluation. In the problem of thermal-aware floorplanning, the layout evaluation by the thermal model takes 99.5% of the computational time for the best floorplanning algorithm proposed so far. The contribution of this paper is to present a parallelization of this evaluation phase in a master-worker model to achieve a dramatic speed-up of the thermal-aware floorplanning process. Exhaustive experimentation was carried out over 3D integrated circuits, with 48 and 128 cores, outperforming previous published works. Copyright (c) 2012 John Wiley & Sons, Ltd.

A good smart contract can provide safe and reliable services to efficiently solve real-world problems. However, smart contracts written by developers create potential security vulnerabilities and can result in huge economic losses. Therein, detecting potential vulnerabilities in smart contracts plays an important role in maintaining the security of the blockchain. There are a series of schemes work on finding bugs in smart contracts, such as Oyente and ContractFuzzer, etc. However, existing solutions only address several types of security vulnerabilities on a single platform. In this paper, we propose a general framework to detect potential vulnerabilities of smart contracts. In order to abstract the execution logic of smart contracts, we need to convert them into control flow graphs. We use the Ullmann algorithm to discover vulnerabilities from these graphs. Also, we give a running example to illustrate how our framework can be used to detect the reentrancy problems written by solidity.

In this paper we propose a Sliding Mode Controller for a class of scalar bilinear systems with delay in both the input and the state. Such a class is considered since it has shown to be suitable for modelling and control of a class of turbulent flow systems. The stability and robustness analysis for the reaching phase in the controlled system are Lyapunov-based. However, since the sliding dynamics is infinite dimensional and described by an integral equation, we show that the stability and robustness analysis is simplified by using Volterra operator theory.

Due to the existing cryptographic schemes based on the satisfiability problem (SAT) lack the specific security model and the security analysis of the provable security theory, the security is questioned. For this problem, combining with the advantages of SAT and Elgamal scheme, we construct a secure, verifiable encryption scheme in this paper. Firstly, combining and improving the SRR(N, k, s) model based on the difficulty problem of the random regular formula near the phase transition point, we construct an encryption algorithm. Secondly, on the basis of the existing Damgârd Elgamal scheme, a more secure scheme based on SAT and variants of discrete logarithm problem is proposed. Lastly, the analysis results show that compared with the existing SAT-based hybrid cryptographic schemes, the private key of the hybrid encryption scheme proposed in this paper is shorter and the ciphertext satisfies the verifiability, and the scheme can secure against adaptive chosen ciphertext attacks (CCA2).

International audience; Mimicry and deferred imitation have often been considered as separate kinds of imitation. In this paper, we present a simple architecture for robotic arm control which can be used for both. The model is based on some dynamical equations, which provide a motor control with exploring and converging capacities. A visuo-motor map is used to associate positions of the end effector in the visual space with proprioceptive position of the robotic arm. It enables a fast learning of the visuo-motor associations without needing to embed a priori information. The controller can be used both for accurate control and interaction. It has been implemented on a minimal robotic setup showing some interesting emergent properties. The robot can reproduce simple gestures in mimicry situation and finalized actions in deferred imitation situation. Moreover, it can even show some “intention” recognition abilities. Finally, the experiment of deferred imitation which inherits from learning by demonstration, also provides a good basis for cooperative and interactive experiments.

Chemical reaction system, dynamic operation can significantly increase the average rate of reaction, improve the time-average selectivity of complex reactions and enhance the molecular weight distribution of certain free-radical polymerization reactions, overcome the thermodynamic limitations of reversible reactions. It even can be used as integrated means of exothermic/ endothermic reaction and catalytic reaction/catalyst regeneration, opens up new ways to strengthen and control the reaction process, reduce waste emissions, and increase economic and social benefits. Therefore, it has great significance to model, simulate and calculate the process of chemical reaction. In this paper, we propose a firefly algorithm with memory and adaptive step length to solve the optimal dynamic model of chemical reaction. The characteristics of the proposed algorithm are analyzed in detailed and the simulation results of the algorithm are given. It provides a feasible solution to solve such problems and the simulation results also show the effectiveness of the proposed algorithm.

Deep neural networks (DNNs) are successful in many computer vision tasks. However, the most accurate DNNs require millions of parameters and operations, making them energy, computation and memory intensive. This impedes the deployment of large DNNs in low-power devices with limited compute resources. Recent research improves DNN models by reducing the memory requirement, energy consumption, and number of operations without significantly decreasing the accuracy. This paper surveys the progress of low-power deep learning and computer vision, specifically in regards to inference, and discusses the methods for compacting and accelerating DNN models. The techniques can be divided into four major categories: (1) parameter quantization and pruning, (2) compressed convolutional filters and matrix factorization, (3) network architecture search, and (4) knowledge distillation. We analyze the accuracy, advantages, disadvantages, and potential solutions to the problems with the techniques in each category. We also discuss new evaluation metrics as a guideline for future research., Accepted for publication at 2020 IEEE 6th World Forum on Internet of Things (WF-IoT), New Orleans, LA, USA 2020

The paper presents the concept of using a Protege environment to extract knowledge about the properties and features of discrete problems described by ALMM technology. The concept of an Intelligent ALMM system was proposed firstly by Dudek-Dyduch E. The main purpose of the ALMM system is not only to solve problems with discreet optimization, but also to help the user in choosing the right method and algorithm for solving them. The authors, based on earlier work connected with component structure based solvers using ALMM technology, proposed the exploitation of description of the properties of individual problems in order to create the corresponding knowledge base. This paper presents an example of a knowledge base structure for the problem of scheduling tasks on a single machine that can be used to solve various optimization tasks. Using the OWL language convention, the architecture of classes and related properties of solved problems: axioms and assertions were proposed. An on-line knowledge base representing specific classes of problems together with its own query language can be a scenario for pre-modeling the architecture of the final application. For the ontological architecture, the examples of knowledge extraction scenarios to solve the optimization problem of the tasks scheduling on a single machine are presented too.

In this paper, we investigate the adaptive video delivery for multiple users over time-varying and mutually interfering multi-cell wireless networks. The key research challenge is to jointly design the physical-layer resource allocation scheme and application-layer rate adaptation logic, such that the users’ long-term fair quality of experience (QoE) can be maximized. Due to the timescale mismatch between these two layers and the asynchrony of user requests, however, it is difficult to directly model the cross-layer stochastic control problem by using a reinforcement learning framework. To address this difficulty, we propose a novel two-level decision framework where an optimization-based beamforming scheme (performed at the base stations) and a deep reinforcement learning (DRL)-based rate adaptation scheme (performed at the user terminals) are, respectively, developed, such that a highly complex long-term multi-user QoE fairness problem is decomposed into some relatively simple problems and solved effectively. Our strategy represents a significant departure from the existing schemes with consideration of either a short-term multi-user QoE maximization or a long-term single-user point-to-point QoE maximization. Extensive simulations demonstrate that the proposed cross-layer design is effective and promising.

Identifying butterfly species by image patterns is a challenging task in computer vision and pattern recognition community due to many butterfly species having similar shape patterns with complex interior structures and considerable pose variation. In additional, geometrical transformation and illumination variation also make this task more difficult. In this paper, a novel image descriptor, named Gaussian convolution angle (GCA) is proposed for butterfly species classification. The proposed GCA projects the butterfly vein image function and intensity image function along a group of vectors that start from a common contour points and ends at the remaining contour points which results a group of vectors that capture the complex vein patterns and texture patterns of butterfly images. The Gaussian convolutions of different scales are conducted to the resulting vector functions to generate a multiscale GCA descriptors. The proposed GCA is not only invariant to geometrical transformation including rotation, scaling and translation, but also invariant to lighting change. The proposed method has been tested on a publicly available butterfly image dataset that has 832 samples of 10 species. It achieves a classification accuracy of 92.03% which is higher than the benchmark methods.

Training of Convolutional Neural Networks (CNNs) on long video sequences is computationally expensive due to the substantial memory requirements and the massive number of parameters that deep architectures demand. Early fusion of video frames is thus a standard technique, in which several consecutive frames are first agglomerated into a compact representation, and then fed into the CNN as an input sample. For this purpose, a summarization approach that represents a set of consecutive RGB frames by a single dynamic image to capture pixel dynamics is proposed recently. In this paper, we introduce a novel ordered representation of consecutive optical flow frames as an alternative and argue that this representation captures the action dynamics more effectively than RGB frames. We provide intuitions on why such a representation is better for action recognition. We validate our claims on standard benchmark datasets and demonstrate that using summaries of flow images lead to significant improvements over RGB frames while achieving accuracy comparable to the state-of-the-art on UCF101 and HMDB datasets., Comment: Accepted in WACV 2017

Existing Multi-Instance learning (MIL) methods for medical image classification typically segment an image (bag) into small patches (instances) and learn a classifier to predict the label of an unknown bag. Most of such methods assume that instances within a bag are independently and identically distributed. However, instances in the same bag often interact with each other. In this paper, we propose an Induced SelfAttention based deep MIL method that uses the self-attention mechanism for learning the global structure information within a bag. To alleviate the computational complexity of the naive implementation of self-attention, we introduce an inducing point based scheme into the self-attention block. We show empirically that the proposed method is superior to other deep MIL methods in terms of performance and interpretability on three medical image data sets. We also employ a synthetic MIL data set to provide an intensive analysis of the effectiveness of our method. The experimental results reveal that the induced self-attention mechanism can learn very discriminative and different features for target and non-target instances within a bag, and thus fits more generalized MIL problems.

User stories are descriptions of functionality that a software user needs. They play an important role in determining which software requirements and bug fixes should be handled and in what order. Developers elicit user stories through meetings with customers. But user story elicitation is complex, and involves many passes to accommodate shifting and unclear customer needs. The result is that developers must take detailed notes during meetings or risk missing important information. Ideally, developers would be freed of the need to take notes themselves, and instead speak naturally with their customers. This paper is a step towards that ideal. We present a technique for automatically extracting information relevant to user stories from recorded conversations between customers and developers. We perform a qualitative study to demonstrate that user story information exists in these conversations in a sufficient quantity to extract automatically. From this, we found that roughly 10.2% of these conversations contained user story information. Then, we test our technique in a quantitative study to determine the degree to which our technique can extract user story information. In our experiment, our process obtained about 70.8% precision and 18.3% recall on the information.

A key operational decision faced by a multi-facility health care diagnostic centre serving different patient categories (for example: Health Check-up Patient (HCP), Out-Patient (OP), Emergency Patient (EP), or In-Patient) is whom to serve next at a particular facility. In this paper, we model random arrival of these patients belonging to different categories and priorities at multiple diagnostic facilities over a finite planning horizon. We formulate a mathematical model for sequential decision-making under uncertainty using Markov Decision Process (MDP) with the objective of maximising net revenue and use dynamic programming (DP) to solve it. To address dimensionality and scalability issue of MDP, we provide a decentralised MDP (D_MDP) formulation. We develop simulation-based neighbourhood search algorithm to improve DP solution for D_MDP. We compare these solutions with three other rule-based heuristics using simulation.

As the success of deep models has led to their deployment in all areas of computer vision, it is increasingly important to understand how these representations work and what they are capturing. In this paper, we shed light on deep spatiotemporal representations by visualizing the internal representation of models that have been trained to recognize actions in video. We visualize multiple two-stream architectures to show that local detectors for appearance and motion objects arise to form distributed representations for recognizing human actions. Key observations include the following. First, cross-stream fusion enables the learning of true spatiotemporal features rather than simply separate appearance and motion features. Second, the networks can learn local representations that are highly class specific, but also generic representations that can serve a range of classes. Third, throughout the hierarchy of the network, features become more abstract and show increasing invariance to aspects of the data that are unimportant to desired distinctions (e.g. motion patterns across various speeds). Fourth, visualizations can be used not only to shed light on learned representations, but also to reveal idiosyncrasies of training data and to explain failure cases of the system.

Recent renewed interest in computational writer identification has resulted in an increased number of publications. In relation to historical musicology its application has so far been limited. One of the obstacles seems to be that the clarity of the images from the scans available for computational analysis is often not sufficient. In this paper, the use of the Hinge feature is proposed to avoid segmentation and staff-line removal for effective feature extraction from low quality scans. The use of an auto encoder in Hinge feature space is suggested as an alternative to staff-line removal by image processing, and their performance is compared. The result of the experiment shows an accuracy of 87 % for the dataset containing 84 writers’ samples, and superiority of our segmentation and staff-line removal free approach. Practical analysis on Bach’s autograph manuscript of the Well-Tempered Clavier II (Additional MS. 35021 in the British Library, London) is also presented and the extensive applicability of our approach is demonstrated.

In 2019, the National Archives of Korea (NAK) developed a blockchain recordkeeping platform to conduct R&amp, D on recordkeeping approaches. This paper introduces two types of R&amp, D studies that have been conducted thus far. The first is the use of blockchain transaction audit trail technology to ensure the authenticity of audiovisual archives, i.e., the application of blockchain to a new system. The second uses blockchain technology to verify whether the datasets of numerous information systems built by government agencies are managed without forgery or tampering, i.e., the application of blockchain to an existing system. Government work environments globally are rapidly shifting from paper records to digital. However, the traditional recordkeeping methodology has not adequately kept up with these digital changes. Despite the importance of responding to digital changes by incorporating innovative technologies such as blockchain in recordkeeping practices, it is not easy for most archives to invest funds in experiments on future technologies. Owing to the Korean government’s policy of investing in digital transformation, NAK’s blockchain recordkeeping platform has been developed, and several R&amp, D tasks are underway. Hopefully, the findings of this study will be shared with archivists around the world who are focusing on the future of recordkeeping.

Web automation applications are widely used for different purposes such as B2B integration and automated testing of web applications. Most current systems build the automatic web navigation component by using the APIs of conventional browsers. While this approach has its advantages, it suffers performance problems for intensive web automation tasks which require real time responses and/or a high degree of parallelism. In this paper, we outline a set of techniques to build a web navigation component able to efficiently execute web navigation sequences. These techniques detect what elements and scripts of the pages accessed during the navigation sequence are needed for the correct execution of the sequence (and, therefore, must be loaded and executed), and what parts of the pages can be discarded. The tests executed with real web sources show that the optimized navigation sequences run significantly faster and consume significantly less resources.

This paper presents a control design for the irrigation station by sprinkling. The proposed method is applied in order to solve the problem of managing water sources and distributions systems. The main contribution is focused primarily on conceiving T-S fuzzy model for the irrigation station. In formalism of the fuzzy-coalescence algorithm, we derive sufficient conditions to determine the model parameters. Meanwhile, additional effort is developed to achieve PI control for the irrigation station model.

Next Point-of-interest (POI) recommendation has been recognized as an important technique in location-based services, and existing methods aim to utilize sequential models to return meaningful recommendation results. But these models fail to fully consider the phenomenon of user interest drift, i.e. a user tends to have different preferences when she is in out-of-town areas, resulting in sub-optimal results accordingly. To achieve more accurate next POI recommendation for out-of-town users, an adaptive attentional deep neural model HOPE is proposed in this paper for modeling user’s out-of-town dynamic preferences precisely. Aside from hometown preferences of a user, it captures the long and short-term preferences of the user in out-of-town areas using “Asymmetric-SVD” and “TC-SeqRec” respectively. In addition, toward the data sparsity problem of out-of-town preference modeling, a region-based pattern discovery method is further adopted to capture all visitor’s crowd preferences of this area, enabling out-of-town preferences of cold start users to be captured reasonably. In addition, we adaptively fuse all above factors according to the contextual information by adaptive attention, which incorporates temporal gating to balance the importance of the long-term and short-term preferences in a reasonable and explainable way. At last, we evaluate the HOPE with baseline sequential models for POI recommendation on two real datasets, and the results demonstrate that our proposed solution outperforms the state-of-art models significantly.

Three-dimensional head pose estimation has been an important and challenging task in computer vision partly because of its diverse applications. In this paper, we propose a new method to estimate head pose for the faces in the wild using deep neural network based on classification, rather than conventional regression. The network consists of three CNNs, corresponding to three head pose components, i.e. yaw, pitch and roll, respectively, and each CNN has 3 channels for color input. We have trained the network using the representative face datasets in the wild, such as AFLW, AFW and YouTube, and found that it outperforms the state-of-art algorithms with a substantial margin. We have also explored transfer learning among them: training the network with a dataset and testing it against others. It is found that there is directional difference between them. And yet, result obtained by transferring from AFLW to others is still better than those from recent studies.

Unsupervised manifold learning has become accepted as an important tool for reducing dimensionality of a dataset by finding its meaningful low-dimensional representation lying on an unknown nonlinear subspace. Most manifold learning methods only embed an existing dataset but do not provide an explicit mapping function for novel out-of-sample data, thereby potentially resulting in an ineffective tool for classification purposes, particularly for iterative methods, such as active learning. To address this issue, out-of-sample extension methods have been introduced to generalize an existing embedding of new samples. In this paper, a novel out-of-sample method is introduced by utilizing high dimensional model representation (HDMR) as a nonlinear multivariate regression with the Tikhonov regularizer for unsupervised manifold learning algorithms. The proposed method was extensively analyzed using illustrative datasets sampled from known manifolds. Several experiments with 3D synthetic datasets and face recognition datasets were also conducted, and the performance of the proposed method was compared to several well-known out-of-sample methods. The results obtained with locally linear embedding (LLE), Laplacian Eigenmaps (LE), and t-distributed stochastic neighbor embedding (t-SNE) showed that the proposed method achieves competitive even better performance than the other out-of-sample methods.

In the field of modern bidding, electronic bidding leads a new trend of development, convenience and efficiency and other significant advantages effectively promote the reform and innovation of China’s bidding field. Nowadays, most systems require a strong and trusted third party to guarantee the integrity and security of the system. However, with the development of blockchain technology and the rise of privacy protection, researchers has begun to emphasize the core concept of decentralization. This paper introduces a decentralized electronic bidding system based on blockchain and smart contract. The system uses blockchain to replace the traditional database and uses chaincode to process business logic. In data interaction, encryption techniques such as zero-knowledge proof based on graph isomorphism are used to improve privacy protection, which improves the anonymity of participants, the privacy of data transmission, and the traceability and verifiable of data. Compared with other electronic bidding systems, this system is more secure and efficient, and has the nature of anonymous operation, which fully protects the privacy information in the bidding process.

In this paper we propose to utilize a new concept of discrete directional Gabor frames for automatic image registration. The directional Gabor representations have been shown to provide more accurate feature extraction than directional wavelet transforms for images where texture is the dominant feature. Initial experimental results are presented here which indicate that discrete directional Gabor frames exhibit strong correlations, which indicates that they are likely to improve the existing image registration toolbox.

Object tracking and avoidance is a vital functionality in autonomous vehicles. Object tracking techniques must be both fast and accurate for this safety-critical application. Neither of the two factors mentioned can be compromised in this safety critical application. Much of the work done in this field uses the point clouds themselves for object matching. This ensures greater accuracy at the expense of run time. This paper posits that such a fine level of registration is not required for comparable levels of accuracy. The object tracking technique proposed uses a coarse technique for object matching via voxels which reduces run time without compromising on accuracy. It also allows for evaluation of quality of matches.

Embeddings of medical concepts such as medication, procedure and diagnosis codes in Electronic Medical Records (EMRs) are central to healthcare analytics. Previous work on medical concept embedding takes medical concepts and EMRs as words and documents respectively. Nevertheless, such models miss out the temporal nature of EMR data. On the one hand, two consecutive medical concepts do not indicate they are temporally close, but the correlations between them can be revealed by the time gap. On the other hand, the temporal scopes of medical concepts often vary greatly (e.g., \textit{common cold} and \textit{diabetes}). In this paper, we propose to incorporate the temporal information to embed medical codes. Based on the Continuous Bag-of-Words model, we employ the attention mechanism to learn a "soft" time-aware context window for each medical concept. Experiments on public and proprietary datasets through clustering and nearest neighbour search tasks demonstrate the effectiveness of our model, showing that it outperforms five state-of-the-art baselines., 7 pages. IJCAI-ECAI-2018

Mobile vehicles, equipped with low-cost sensors, can provide unprecedented opportunities for urban monitoring with a wide coverage. The quality of collected data is very important for many applications. The low-cost mobile sensors can, however, suffer from limited accuracy, high instability, and sensor drift. Therefore, they need to be frequently calibrated to preserve a good data quality. Frequent sensor calibration can be achieved by deploying static high-precision sensors (called reference sensors) and exploiting meeting points (i.e., rendezvous) between the reference sensors and mobile sensors. A calibrated mobile sensor can also be used to calibrate an uncalibrated mobile sensor when they meet, referred to as multihop calibration. In this paper, we introduce a novel concept, k-hop calibratability, in the context of multihop calibration: a sensor is k-hop calibratable if the calibration path is no larger than k. We consider the problem of how to deploy the reference sensors to ensure that all sensors in the network are k-hop calibratable. To address this problem, we formally define the rendezvous connection graph to precisely describe the relationship between mobile sensors and (virtual) reference sensors. Based on this definition, we formulate the reference sensor deployment problem as a set cover problem. We also extend our problem to consider practical deployment requirements. We propose efficient algorithms and conduct an extensive evaluation in the context of mobile air quality monitoring. We present a detailed prototype implementation of the mobile sensors and reference sensors. Evaluations using real-world datasets show the effectiveness of the algorithms.

Literature has indicated that accurate dietary assessment is very important for assessing the effectiveness of weight loss interventions. However, most of the existing dietary assessment methods rely on memory. With the help of pervasive mobile devices and rich cloud services, it is now possible to develop new computer-aided food recognition system for accurate dietary assessment. However, enabling this future Internet of Things-based dietary assessment imposes several fundamental challenges on algorithm development and system design. In this paper, we set to address these issues from the following two aspects: (1) to develop novel deep learning-based visual food recognition algorithms to achieve the best-in-class recognition accuracy; (2) to design a food recognition system employing edge computing-based service computing paradigm to overcome some inherent problems of traditional mobile cloud computing paradigm, such as unacceptable system latency and low battery life of mobile devices. We have conducted extensive experiments with real-world data. Our results have shown that the proposed system achieved three objectives: (1) outperforming existing work in terms of food recognition accuracy; (2) reducing response time that is equivalent to the minimum of the existing approaches; and (3) lowering energy consumption which is close to the minimum of the state-of-the-art.

P2P botnets represent another escalation level in the race of arms between criminals and the research community. By utilizing a distributed P2P architecture they are resilient against random failures and attacks and overcome the limitations of a central command and control server. For this reason, it is important to monitor them to gather information for potential takedown attempts. In this paper, we introduce our high-frequency crawling tool Strobo-Crawler that can carry out a fine-grained node enumeration. Furthermore, we propose mechanisms to derive accurate snapshots of the botnet graph on the basis of restricted monitoring data. We applied Strobo-Crawler in a two week crawling campaign in the P2P botnets Sality and ZeroAccess and describe the results along with a careful evaluation of our graph reconstruction. Furthermore, we provide a thorough analysis of the resulting botnet graphs and also provide these graphs to the public. Our results indicate that they are highly resilient against node churn, but also against targeted attacks. Bots are highly interconnected and the graphs are characterized by a high clustering coefficient, high density, and low diameter.

Purpose The purpose of this paper is to answer the question whether it is possible to recognise the gender of a web browser user on the basis of keystroke dynamics and mouse movements. Design/methodology/approach An experiment was organised in order to track mouse and keyboard usage using a special web browser plug-in. After collecting the data, a number of parameters describing the users’ keystrokes, mouse movements and clicks were calculated for each data sample. Then several machine learning methods were used to verify the stated research question. Findings The experiment showed that it is possible to recognise males and females on the basis of behavioural characteristics with an accuracy exceeding 70 per cent. The best results were obtained while using Bayesian networks. Research limitations/implications The first limitation of the study was the restricted contextual information, i.e. neither the type of web page browsed nor the user activity was taken into account. Another is the narrow scope of the respondent group. Future work should focus on gathering data from more users covering a wider age range and should consider the context. Practical implications Automatic gender recognition could be used in profiling a user to create personalised websites or as an additional feature in automatic identification for security reasons. It might be also considered as a confirmation of declared gender in web-based surveys. Social implications As not all users perceive personalised ads and websites as beneficial, this application requires the analysis of a user perspective to provide value to the consumer without privacy violation. Originality/value Behavioural characteristics, such as mouse movements and keystroke dynamics, have already been used for user authentication and emotion recognition, but applying these data to gender recognition is an original idea.

In context of Arabic, the authorship attribution (AA) problem is not addressed well comparing with other natural languages such English, Chinese and Dutch. This paper addresses the attribution problem in context of Islamic fatwā’. To the best of our knowledge, this is the first study of its kind that addresses this problem in such domain. In term of attribution methods, three machine-learning classifiers namely, the locally weighted learning (LWL) classifier, decision tree C4.5, and Random Forest (RF) are used. The experiment is performed with a selected list of stylomatric features. To extract the most discriminating features, various feature selection techniques are used. The experimental results show that the classifiers have different behaviour respect each feature reduction techniques. Among the used classifiers, the C4.5 method gives the best accuracy.

Most distributed data mining algorithms can efficiently manage and mine complete data from distributed resources. However, for an incomplete data some modifications are required in order to perform distributed data mining techniques and maintaining the privacy of the sensitive information to provide pretty good results of data mining. Classification is important tasks of data mining aimed at discovering knowledge and classify new instances. SVM is classified as one of the most important algorithm used for classification problems in several various spheres. In this paper, we proposed a new distributed privacy-preserving protocol with multiple imputations of missing or incomplete data. More so, a multiple imputations based on multivariate imputation by chained equations is used for missing data and Paillier cryptosystem for maintaining the privacy of the participants. Finally we constructed a global SVM model by introducing a third party (semi-honest approach) over vertical partition data based in Gram matrix without revealing the privacy of the data and used to classify new instances. The performance evolution of the proposed protocol was investigated while using accuracy metric on the distributed and centralized data. Results of our experiments reveal that the accuracy is the same as centralized data and achieve better results with imputed data while compared with omitted data. The performance of distributed data on our protocol achieves better processing time compared with centralized data.

When used as an interface in the context of Ambient Assisted Living (AAL), a social robot should not just provide a task-oriented support. It should also try to establish a social empathic relation with the user. To this aim, it is crucial to endow the robot with the capability of recognizing the user's affective state and reason on it for triggering the most appropriate communicative behavior. In this paper we describe how such an affective reasoning has been implemented in the NAO robot for simulating empathic behaviors in the context of AAL. In particular, the robot is able to recognize the emotion of the user by analyzing communicative signals extracted from speech and facial expressions. The recognized emotion allows triggering the robot's affective state and, consequently, the most appropriate empathic behavior. The robot's empathic behaviors have been evaluated both by experts in communication and through a user study aimed at assessing the perception and interpretation of empathy by elderly users. Results are quite satisfactory and encourage us to further extend the social and affective capabilities of the robot.

International audience; To face fire it is crucial to understand its behaviour in order to maximize fighting means. To achieve this task, the development of a metrological tool is necessary for estimating both geometrical and physical parameters involved in forest fire modelling. A key parameter is to estimate fire positions accurately. In this paper an image processing tool especially dedicated to an accurate extraction of fire from an image is presented. In this work, the clustering on several colour spaces is investigated and it appears that the blue chrominance Cb from the YCbCr colour space is the most appropriate. As a consequence, a new segmentation algorithm dedicated to forest fire applications has been built using first an optimized k-means clustering in the Cb-channel and then some properties of fire pixels in the RGB colour space. Next, the performance of the proposed method is evaluated using three supervised evaluation criteria and then compared to other existing segmentation algorithms in the literature. Finally a conclusion is drawn, assessing the good behaviour of the developed algorithm.

People with sleep apnea (SA) are at increased risk of having stroke and cardiovascular diseases. Polysomnography (PSG) is used to detect SA. This paper conducts feature selection from PSG signals and uses a support vector machine (SVM) to detect SA. To analyze SA, the Physionet Apnea Database was used to obtain various features. Electrocardiography (ECG), oxygen saturation (SaO2), airflow, abdominal, and thoracic signals were used to provide various frequency-, time-domain and non-linear features (n = 87). To analyse the significance of these features, firstly, two evaluation measures, the rank-sum method and the analysis of variance (ANOVA) were used to evaluate the significance of the features. These features were then classified according to their significance. Finally, different class feature sets were presented as inputs for an SVM classifier to detect the onset of SA. The hill-climbing feature selection algorithm and the k-fold cross-validation method were applied to evaluate each classification performance. Through the experiments, we discovered that the best feature set (including the top-five significant features) obtained the best classification performance. Furthermore, we plotted receiver operating characteristic (ROC) curves to examine the performance of the SVM, and the results showed the SVM with Linear kernel (regularization parameter = 1) outperformed other classifiers (area under curve = 95.23%, sensitivity = 94.29%, specificity = 96.17%). The results confirm that feature subsets based on multiple bio-signals have the potential to identify patients with SA. The use of a smaller subset avoids dimensionality problems and reduces the computational load.

In this paper, an improved Particle Swarm Optimization Algorithm (GCPSO) is proposed to solve the shortcomings of the existing Particle Swarm Optimization Algorithm (PSO) which has low convergence precision, slow convergence rate and is easy to fall into local optimum when performing high-dimensional optimization in the late iteration. First, the whole particle swarm of the algorithm was divided into three sub-groups, and different ranges of inertia weight ω are set for balances global search and local search in each sub-group, which improves the algorithm’s ability to explore. Then we add Gaussian perturbation with the greedy strategy to PSO to avoid the algorithm falling into local optimum and improve the convergence speed. And finally, the proposed algorithm is compared with Genetic Algorithm (GA), PSO and Grasshopper Optimization Algorithm (GOA) to analyse its performance and speed. Through experimental analysis, GCPSO has a significant improvement at convergence speed, convergence accuracy and stability.

Machine learning (ML) algorithms provide a good solution for many security sensitive applications, they themselves, however, face the threats of adversary attacks. As a key problem in machine learning, how to design robust feature selection algorithms against these attacks becomes a hot issue. The current researches on defending evasion attacks mainly focus on wrapped adversarial feature selection algorithm, i.e., WAFS, which is dependent on the classification algorithms, and time cost is very high for large-scale data. Since mRMR (minimum Redundancy and Maximum Relevance) algorithm is one of the most popular filter algorithms for feature selection without considering any classifier during feature selection process. In this paper, we propose a novel adversary-aware feature selection algorithm under filter model based on mRMR, named FAFS. The algorithm, on the one hand, takes the correlation between a single feature and a label, and the redundancy between features into account; on the other hand, when selecting features, it not only considers the generalization ability in the absence of attack, but also the robustness under attack. The performance of four algorithms, i.e., mRMR, TWFS (Traditional Wrapped Feature Selection algorithm), WAFS, and FAFS is evaluated on spam filtering and PDF malicious detection in the Perfect Knowledge attack scenarios. The experiment results show that FAFS has a better performance under evasion attacks with less time complexity, and comparable classification accuracy.

Exascale computing is one of the major challenges of this decade, and several studies have shown that communications are becoming one of the bottlenecks for scaling parallel applications. The analysis on the characteristics of communications can effectively aid to improve the performance of scientific applications. In this paper, we focus on the statistical regularity in time-dimension communication characteristics for representative scientific applications on supercomputer systems, and then prove that the distribution of communication-event intervals has a power-law decay, which is common in scientific interests and human activities. We verify the distribution of communication-event intervals has really a power-law decay on the Tianhe-2 supercomputer, and also on the other six parallel systems with three different network topologies and two routing policies. In order to do a quantitative study on the power-law distribution, we exploit two groups of statistics: bursty vs. memory and periodicity vs. dispersion. Our results indicate that the communication events show a “strong-bursty and weak-memory” characteristic and the communication event intervals show the periodicity and the dispersion. Finally, our research provides an insight into the relationship between communication optimizations and time-dimension communication characteristics.

Based on the analysis of the status of archives social media information at home and abroad, it is not only a strategic step in the construction of archives informatization, but also an inevitable choice for its effective construction and utilization. This paper draws lessons from the international standards of document management system and the reference model of open archives information system, and designs the framework of archives social media integration system at the top level. It also takes the heterogeneous data of various archives social media information as the analysis source, and excavates the contents in depth. The LDA model is improved and extended by utilizing certain attribute information in microblog messages. At the same time, the relevant features of data sources are fused into the traditional document sorting function, which significantly improves the retrieval performance of multi-source social media information.

In this paper, we explore building classifiers to detect Salsa dance step primitives in choreographies available in the Huawei 3DLife data set. These can collectively be an important component of dance tuition systems that support e-learning. A dance step is reasoned as the shortest possible extract of bodily motion that can uniquely identify a particularly repeatable movement through time. The representation of dance steps adopted is a concatenation of vectorized matrices involving the 3D coordinates of tracked body joints. Under this modeling context, a Salsa dance performance is seen as an ordered sequence of Salsa dance steps, requiring a multiple of the variables allocated in the representation of a single step. Following a previous work by Masurelle & Essid that discusses the classification of six Salsa dance steps from 3DLife, we show that it is possible to obtain better classifiers under a similar experimental protocol in terms of both test accuracy and F-measure. By carefully re-annotating the data in 3DLife, we refocus on the six-step classification problem and then extend the protocol to the case of 20 dance steps. In comparison to common classifiers of the trade operating on full-dimensions, we show that it is possible to produce more accurate models by computing a subspace of the data. At the same time it is possible to reduce problematic bias in resulting models due to the uneven distribution of samples across step data classes. We provide and discuss experimental findings to support both hypotheses for the two experimental settings.