Search

Product ratings and reviews can provide rich useful information of users and items, and are widely used in recommender systems. However, it is nontrivial to infer user preference according to the history behaviors since users always have different and complicated interests for different target items. Hence, precisely modelling the context interaction of users and target items is important in learning their representations. In this paper, we propose a unified recommendation method with mutual self attention mechanism to capture the inherent interactions between users and items from reviews and ratings. In our method, we design a review encoder based on mutual self attention to extract the semantic features of users and items from their reviews. The mutual self attention could capture the interaction information of different words between the user review and item review automatically. Another rating-based encoder is utilized to learn representations of users and items from the rating patterns. Besides, we propose a neural gated network to effectively fuse the review- and rating-based features as the final comprehensive representations of users and items for rating prediction. Extensive experiments are conducted on real world recommendation datasets to evaluate our method. The experiments show that our method can achieve better recommendation performance compared with many competitive baselines.

Roles in a complex network usually represent the local connectivity patterns of nodes, which reflect the functions or behaviors of corresponding entities. Role discovery has great meaning for understanding the formation and evolution of networks. While the importance of role discovery in networks has been realized gradually, a variety of approaches of role-oriented network representation learning are proposed. Almost all the existing approaches are dependent on manual high-order structural properties which are always fragmentary. They suffer from unstable performances and poor generalization ability, because their hand-craft structural features sometimes miss the characteristics of different networks. In addition, graph neural networks (GNNs) have great potential to automatically capture structural properties, but it is hard to be given the rein to for the difficulty of designing role-oriented unsupervised loss. To overcome these challenges, we provide an idea that leverage low-dimensional extracted structural features as guidance to train graph neural networks. Based on the idea, we proposed GAS, a novel graph auto-encoder guided by structural information, to learn role-oriented representations for nodes. Results of extensive experiments show that GAS has better performance than other state-of-the-art approaches.

Smart mobile devices have access to huge amounts of data appropriate to deep learning models, which in turn can significantly improve the end-user experience on mobile devices. But massive data collection required for machine learning introduce obvious privacy issues. To this end, the notion of federated learning (FL) was proposed, which leaves the training data distributed on the mobile devices, and learns a shared model by aggregating locally-computed updates. However, in many applications one or more Byzantine devices may suffice to let current coordination learning mechanisms fail with unpredictable or disastrous outcomes. In this paper, we provide a proof-of-concept for managing security issues in federated learning systems via blockchain technology. Our approach uses decentralized programs executed via blockchain technology to establish secure learning coordination mechanisms and to identify and exclude Byzantine members. We studied the performance of our blockchain-based approach in a collective deep-learning scenario both in the presence and absence of Byzantine devices and compared our results to those obtained with an existing collective decision approach. The results show a clear advantage of the blockchain approach when Byzantine devices are part of the members.

In recent years, computer binocular vision has been commonly utilized to provide depth information for autonomous vehicles. This paper presents an efficient binocular vision system implemented on an SoC TMS320C6678 DSP for real-time depth information extrapolation, where the search range propagates from the bottom of an image to its top. To further improve the stereo matching efficiency, the cost function is factorized into five independent parts. The value of each part is pre-calculated and stored in the DSP memory for direct data indexing. The experimental results illustrate that the proposed algorithm performs in real time, when processing the KITTI stereo datasets with eight cores in parallel.

Few-shot learning in computer vision is a very challenging task. Many approaches have been proposed to tackle the few-shot learning problem. Meta learning, a method of learning to learn, is introduced into few-shot learning problem and has achieved pretty good results. But there is still a very big gap between the machine and our human in the few-shot learning tasks. We think it’s because the existing methods do not make full use of global knowledge (similar to the priori knowledge of human understanding of images) thus lacking a world view of the task. In other words, they focus too much on local information and neglect the whole task. In this paper, we rethink about the few-shot learning problem, and propose that we should make full use of global knowledge. Seen data set is used to obtain a embedding function between images and feature vectors, the images are embedded onto a hypersphere in the manner of cosine embedding. By taking this embedding function as global knowledge, we train a classifier to classify the corresponding embedded vector of images. The experiment proved that our approach significantly outperforms both baseline models and previous state-of-the-art methods. It surpasses most existing methods in terms of flexibility, simplicity and accuracy. Codes are available at https://github.com/SongyiGao/OICEFFSL.

Over the past decade, a number of frameworks have been introduced to support different crowdsourcing tasks. However, complex creative tasks have remained out of reach for workflow modeling. Unlike typical tasks, creative tasks are often interdependent, requiring human cognitive ability and team collaboration. The crowd workers are required not only to perform typical tasks, but also to participate in the analysis and manipulation of complex tasks, hence the number and execution order of tasks are unknown until runtime. Thus, it is difficult to model this kind of complex tasks by using existing workflow approaches. Therefore, we propose a workflow modeling approach based on state machine to design crowdsourcing model that can be translated into SCXML code and executed by an open source engine. This approach and engine are embodied in SmartCrowd. Through two evaluations, we found that SmartCrowd can provide support for complex crowdsourcing tasks, especially on creative tasks. Moreover, we introduce a set of basic design patterns, and by employing them to compose complex patterns, our framework can support more crowdsourcing research.

The gravitational search algorithm (GSA) has proven to be a good optimization algorithm to solve various optimization problems. However, due to the lack of exploration capability, it often traps into local optima when dealing with complex problems. Hence its convergence speed will slow down. A clustering-based learning strategy (CLS) has been applied to GSA to alleviate this situation, which is called galactic gravitational search algorithm (GGSA). The CLS firstly divides the GSA into multiple clusters, and then it applies several learning strategies in each cluster and among clusters separately. By using this method, the main weakness of GSA that easily trapping into local optima can be effectively alleviated. The experimental results confirm the superior performance of GGSA in terms of solution quality and convergence in comparison with GSA and other algorithms.

Heavy Hitters refer to the set of flows that represent a significantly large proportion of the link capacity or of the active traffic. Identifying Heavy Hitters is of particular importance in both network management and security applications. Traditional methods are focusing on sampling in the middle box and analyzing those packets using streaming algorithms. The paradigm of Software Defined Network (SDN) simplifies the work of flow counting. However, continuously monitoring the network will introduce overhead, which needs to be considered as a tradeoff between accurate measurement in real-time. In this paper, We propose a novel method that stamps each suspicious flow with a weight based on an online learning algorithm. The granularity of measurement is dynamically changed according to the importance of each flow. We take advantage of history flows to make the procedure of finding a heavy hitter faster so that applications can make decisions instantly. Using real-world data, we show that our online learning method can detect heavy hitters faster with less overhead and the same accuracy.

Relation detection is a core component for Knowledge Base Question Answering (KBQA). In this paper, we propose a knowledge base (KB) relation detection model based on multi-view matching, which utilizes useful information extracted from questions and KB. The matching inside each view is through multiple perspectives to compare two input texts thoroughly. All these components are trained in an end-to-end neural network model. Experiments on SimpleQuestions and WebQSP yield state-of-the-art results on relation detection.

Team formation problem refers to finding a set of skillful individuals to accomplish given tasks as a team. A growing interest of recent researches on team formation is to concern the collaboration factor, following the general idea that effective communication among team members may contribute to better team performance. Previous studies have introduced a variety of ways to model the collaboration factor applying the concept communication cost, yet few have investigated the effectiveness of the proposed metrics. In this paper, an empirical study is conducted to evaluate the effectiveness of existing communication cost metrics in terms of the influence on team performance. We select real data from IMDb as an example for the study and apply statistical analysis. Based on the result of evaluation, we further propose modification for the communication cost metrics and demonstrate the feasibility. This empirical study is expected to provide suggestion and inspiration for both researchers and practitioners, while design or select models to solve team formation problem in different application scenarios.

Pulse diagnosis is an integral part in Traditional Chinese Medicine (TCM). However, the feeling of pulse condition is subjective and mysterious based on lots of practice and experience which brings many obstacles to pass on the skill. Additionally, in order to take full advantage of the limited resource of skillful TCM experts, remote TCM by Internet was proposed in recent years. Obviously, reproduction apparatus of pulse condition is of great importance in remote TCM system as well as pulse diagnosis education. In this paper, a pulse condition reproduction apparatus, which is used to reproduce corresponding pulse vibration according to the pulse condition data received from universal serial bus, was developed. To evaluate the performance of the apparatus, the reproduced pulse wave was measured and compared with the input pulse wave. Consequently, the correlation coefficients between the average output pulse waveforms and the input pulse waveforms is 0.9911 for normal pulse. The flowing sense of the pulse under fingers is indirectly demonstrated by the phase difference between adjacent fingers. Results indicated that the developed apparatus is available for pulse diagnosis education and remote palpation.

The ubiquity of cloud computing is shifting the deployment of Business Process Management Systems (BPMS) from traditional on-premise models to the Software-as-a-Service (SaaS) paradigm, thus aiming to deliver Business Process Automation as a Service to multiple tenants in the cloud. However, scaling up a traditional BPMS to cope with simultaneous demand from multiple organisations in the cloud is challenging, since its underlying system architecture has been designed to serve a single organisation with a single workflow engine. A typical SaaS often deploys multiple instances of its core applications and distributes workload to these application instances via load balancing. But, for stateful and often long-running process instances, standard stateless load balancing strategies are inadequate. In this paper, we propose a conceptual design of a scalable system architecture for deploying BPMS in the cloud. In our design, Object Role Modeling (ORM) is used to conceptualise the data requirements of the system and UML sequence diagrams are used to capture the interactions between system components. A prototypical implementation using an open-source traditional BPMS offers focused load balancing strategies and demonstrates improved capabilities for supporting large volumes of work in a multi-tenanted cloud environment.

Organizational mining aims at gaining insights for business process improvement by discovering organizational knowledge relevant to the performance of business processes. A key topic of organizational mining is the discovery of organizational models from event logs. While it is common for modern organizations to have employees sharing roles and responsibilities across different internal groups, most of the existing methods for organizational model discovery are unable to identify such overlaps. The overlapping resources are likely to be generalists in an organization. Existing findings in process redesign best practices have proven that generalists can help increase the flexibility of a business process (similarly to the flexibility of the role of “libero” in certain team sports). In this paper we propose an approach capable of discovering organizational models with overlaps and thus helping identify generalists in an organization. The approach builds on existing cluster analysis techniques to address the underlying technical challenges. Through experiments on real-life event logs the applicability and effectiveness of the proposed method are evaluated.

Whale optimization algorithm (WOA) is a newly proposed search optimization technique which mimics the encircling prey and bubble-net attacking mechanisms of the whale. It has proven to be very competitive in comparison with other state-of-the-art metaheuristics. Nevertheless, the performance of WOA is limited by its monotonous search dynamics, i.e., only the encircling mechanism drives the search which mainly focus the exploration in the landscape. Thus, WOA lacks of the capacity of jumping out the of local optima. To address this problem, this paper propose a memetic whale optimization algorithm (MWOA) by incorporating a chaotic local search into WOA to enhance its exploitation ability. It is expected that MWOA can well balance the global exploration and local exploitation during the search process, thus achieving a better search performance. Forty eight benchmark functions are used to verify the efficiency of MWOA. Experimental results suggest that MWOA can perform better than its competitors in terms of the convergence speed and the solution accuracy.

The traditional process-centric modeling approach often focuses on tasks with a fixed execution order. This modeling approach is difficult to describe business processes in which the number of steps and cases is varied depending on the case state and determined dynamically at run time. Motivated by modeling requirements from complex collaborative business scenarios in the internet, such as crowdsourcing, a concept of Business Object (BO) and BO-oriented Process Modeling (BOOPM) approach are proposed. We introduce the role element into standard Harel statechart to explicitly describe the life cycle of a BO whose state changes as tasks processed by different business roles. We also extend event communication mechanism into the statechart to support collaboration among multiple BOs to achieve a business goal. Meanwhile, the BO provides a suitable foundation to control the granularity of the process model properly. After that, the ability of BOOWM for process specification is evaluated based on a set of workflow patterns. Finally, we illustrate BOOPM approach using a crowdsourcing example and realize a prototype system to support the execution of a business process modeling by BOOPM approach.

In this paper, we proposed an intuitive teleoperation system combining surface-electromyogram (sEMG) with inertia measurement unit (IMU) sensors. Two IMU sensors were worn in the upper arm and forearm to capture arm motion. The results were sent as control command to the remote robotic manipulator in task space. Pattern recognition algorithm was employed to decode sEMG signals collected from the forearm to control a humanoid mechanical hand. As a validation of the proposed system, we evaluated the performance of the system which included a Universal Robot 10 (UR10), a homemade humanoid pro-prosthetic hand (SJT-6) and two homemade armbands. Final experiments showed that the success rate on moving spherical object can be up to 86.7%.

Current intelligent devices in Home Internet, such as routers and cameras, have suffered malicious attacks from hackers. Therefore, security for Home Internet appears particularly significant. In order to have a quantitative evaluation of security defense ability of Home Internet system, this paper proposes an improved vulnerability scoring method on Home Internet based on Information Security Technology Security Vulnerability Classification Guide. Compared to original scoring method which is mainly based on Internet, this improved scoring performs differently. It’s aimed to have a quantitative evaluation on security defense effectiveness of Home Internet system: higher vulnerability score indicates higher threaten degree and relatively weak defense ability. In this paper, the Home Internet system takes dynamically-enabled defense technology (randomly changes system status) to make defense. Through calculating vulnerability scores before and after random changes of system status, this paper succeeds in making a quantitative evaluation on security defense ability of Home Internet system.

Intrusion detection is extremely crucial to prevent computer systems from being compromised. However, as numerous complicated attack types have growingly appeared and evolved in recent years, obtaining quite high detection rates is increasingly difficult. Also, traditional heavily hand-crafted evaluation datasets for network intrusion detection have not been practical. In addition, deep learning techniques have shown extraordinary capabilities in various application fields. The primary goal of this research is utilizing unsupervised deep learning techniques to automatically learn essential features from raw network traffics and achieve quite high detection accuracy. In this paper, we propose a session-based network intrusion detection model using a deep learning architecture. Comparative experiments demonstrate that the proposed model can achieve incredibly high performance to detect botnet network traffics.

Acoustic Emission is a useful tool for on-line detection of tank bottom corrosion. The finite element software COMSOL is used to model the tank bottom. The rules of acoustic emission signal spread from corrosion are studied by analyzing signals from different directions in various positions. Digital filters and wavelet transforms are used to extract features of the acoustic emission signals. Simulation results showing acoustic emission signals spreading in different directions with different frequency spectrum characteristics and varies attenuation rules contribute significantly to the research on acoustic emission mechanisms of tank bottom corrosion and offer guidance for sensor selection.

The increasing gap between plentiful computing elements and limited memory bandwidth makes it increasingly difficult and sometimes even infeasible for HPC community to port more applications onto many-core processor architectures. The Sunway many-core processor SW26010 used to build the Sunway TaihuLight System contains a total of 260 heterogeneous cores. All these cores can be divided into 4 core groups (CGs). Each CG includes a Management Processing Element (MPE) core and 64 Computing Processing Elements (CPEs) cores. In this paper, we refactor an important molecular dynamics (MD) application GROMACS on the Sunway Taihulight system. By rewriting the compute-intensive kernel of GROMACS, we exploit a suitable parallelism for CPE cluster and implement pipelining computation between MPE and CPE cluster. Optimization strategies including the efficient use of scratchpad, the software-emulated cache and a hybrid parallel algorithm are adopted to solve the challenging memory bandwidth limitation. When comparing the refactored version using MPE and 64 CPEs with the original ported version using only MPE, we achieve a 16x speedup for the compute-intensive kernel. For simulating a molecule with 3 million atoms, we currently have managed to scale to 798,720 cores. Moreover, we analyze the adaptability of our mapping and optimization strategies for solving the memory bandwidth limitation when refactoring a real-world application on the Sunway heterogeneous many-core processor system.

To increase the performance Motor imagery (MI) EEG classification, a multi-scale window length algorithm was proposed in this paper. Under this algorithm, eight different time windows with the length of 600–2000 ms were selected to classify the MI data. The confidence level of the classification results from the eight windows were calculated simultaneously, to determine the weights for the outputs of each pre-classifier. The final classification results of MI tasks were generated by the weighted results from the eight windows. Using the proposed algorithm, the response time of MI classification was decreased by 11.4%, comparing to the single window method. On the same level of response time, the classification accuracy was increased by 2.8% by the multi-scale window algorithm. These results show that the proposed algorithm can speed up the detection of MI tasks with reliable classification accuracy.

With the rapid growth in quantity and quality of remote sensing images, extracting the useful information in them effectively and efficiently becomes feasible but also challenging. Convolutional neural network (CNN) is a suitable method to deal with such challenge since it can effectively represent and extract the information. However, the CNN can release their potentials only when enough labelled data provided for the learning procedure. This is a very time-consuming task and even infeasible for the applications with non-cooperative objects or scenes. Unsupervised CNN learning methods, which relieve the need for the labels in the training data, is a feasible solution for the problem. In this work, we investigate a real-world motivated sparsity based unsupervised deep CNN learning method. At first, the method formulates a balanced data driven population and lifetime sparsity prior and thus construct the unsupervised learning method through a layerwise mean. Then we further perform the method on the deep model with multiple CNN layers. Finally, the method is used for the remote sensing image representation and scenes classification. The experimental results over the public UC-Merced Land-use dataset demonstrate that the developed algorithm obtained satisfactory results compared with the recent methods.

Engineering secure software is challenging. Software development organizations leverage a host of processes and tools to enable developers to prevent vulnerabilities in software. Code reviewing is one such approach which has been instrumental in improving the overall quality of a software system. In a typical code review, developers critique a proposed change to uncover potential vulnerabilities. Despite best efforts by developers, some vulnerabilities inevitably slip through the reviews. In this study, we characterized linguistic features—inquisitiveness, sentiment and syntactic complexity—of conversations between developers in a code review, to identify factors that could explain developers missing a vulnerability. We used natural language processing to collect these linguistic features from 3,994,976 messages in 788,437 code reviews from the Chromium project. We collected 1,462 Chromium vulnerabilities to empirically analyze the linguistic features. We found that code reviews with lower inquisitiveness, higher sentiment, and lower complexity were more likely to miss a vulnerability. We used a Naive Bayes classifier to assess if the words (or lemmas) in the code reviews could differentiate reviews that are likely to miss vulnerabilities. The classifier used a subset of all lemmas (over 2 million) as features and their corresponding TF-IDF scores as values. The average precision, recall, and F-measure of the classifier were 14%, 73%, and 23%, respectively. We believe that our linguistic characterization will help developers identify problematic code reviews before they result in a vulnerability being missed.

Cuckoo search algorithm (CSA) is a nature-inspired meta-heuristic based on the obligate brood parasitic behavior of cuckoo species, and it has shown promising performance in solving optimization problems. Chaotic mechanisms have been incorporated into CSA to utilize the dynamic properties of chaos, aiming to further improve its search performance. However, in the previously proposed chaotic cuckoo search algorithms (CCSA), only one chaotic map is utilized in a single search iteration which limited the exploitation ability of the search. In this study, we consider to utilize multiple chaotic maps simultaneously to perform the local search within the neighborhood of the global best solution found by CSA. To realize this, three kinds of multiple chaotic cuckoo search algorithms (MCCSA) are proposed by incorporating several chaotic maps into the chaotic local search parallelly, randomly or selectively. The performance of MCCSA is verified based on 48 widely used benchmark optimization functions. Experimental results reveal that MCCSAs generally perform better than CCSAs, and the MCCSA-P which parallelly utilizes chaotic maps performs the best among all 16 compared variants of CSAs.

To design a high performance bandgap reference circuit in nanoscale CMOS technology becomes a great challenge. Many negative effects of nanoscale CMOS technology in high performance bandgap reference design are discussed in this chapter. A bandgap reference circuit design with both voltage output and current output is presented also. In this design, low threshold voltage MOSfet have been utilized in this design to ensure the circuit at suitable DC operation points under extreme low temperature. Operational transconductance amplifier (OTA) has been used to achieve high DC power supply rejection rate (PSRR). Two opposite temperature coefficient resistors have been connected in series to obtain a one-order temperature independent resistor which also achieves a weak curvature compensation effect for reference voltage’s generation. This bandgap reference is implemented in a 65 nm CMOS technology, occupies 0.75 × 0. 67 mm including bond pads, Measured results show that this circuit can operate at supply voltage from 1.1 to 1.3 V, and the temperature coefficient of voltage output is 30 ppm/°C with 60 dB PSRR (DC, 30°C), and the temperature coefficient of current output is 52 ppm/°C with 70 dB PSRR (DC, 30°C), among −55°C to 125°C without any trimming or calibration.

Productivity can be greatly improved by converting the traditional assembly line to a seru system, especially in the business environment with short product life cycles, uncertain product types and fluctuating production volumes. Line-seru conversion includes two decision processes, i.e., seru formation and seru load. For simplicity, however, previous studies focus on the seru formation with a given scheduling rule in seru load. We select ten scheduling rules usually used in seru load to investigate the influence of different scheduling rules on the performance of line-seru conversion. Moreover, we clarify the complexities of line-seru conversion for ten different scheduling rules from the theoretical perspective. In addition, multi-objective decisions are often used in line-seru conversion. To obtain Pareto-optimal solutions of multi-objective line-seru conversion, we develop two improved exact algorithms based on reducing time complexity and space complexity respectively. Compared with the enumeration based on non-dominated sorting to solve multi-objective problem, the two improved exact algorithms saves computation time greatly. Several numerical simulation experiments are performed to show the performance improvement brought by the two proposed exact algorithms.

This paper proposes a background subtraction method for moving camera. The method relies on motion compensation to transfers the background model from the previous frame to the current frame. This motion compensation is carried out using homography transformation where the homography matrix is estimated from the set of point correspondences between previous and current frame. In order to achieve a fast processing speed, optical-flows from grid-based key-points are calculated to define the point correspondences. The background segmentation itself consists of 3 components: background model, candidate background model, and candidate age. Those 3 parameters are used to define the stable pixels which are considered as the background pixels. The proposed method was tested on a public benchmark system and achieved promising result as shown in the experimental report. Moreover, the method is able to work on real time with 56 fps of processing speed.

Intrusion detection is facing great challenges as network attacks producing massive volumes of data are increasingly sophisticated and heterogeneous. In order to gain much more accurate and reliable detection results, machine learning and visualization techniques have been respectively applied to intrusion detection. In this paper, we review some important work related to machine learning and visualization techniques for intrusion detection. We present a collaborative analysis architecture for intrusion detection tasks which integrate both machine learning and visualization techniques into intrusion detection. We also discuss some significant issues related to the proposed collaborative analysis architecture.

The tendency for high-dimension of universal steganalysis characteristics toward intensifying, and lead to the rapid rise in complexity of algorithm in time and space domain. So maintain the level of detection rates, and reduce the dimension of features at the same time, have significance in research of steganalysis. This paper determines the optimal dimension of feature vectors by principal component analysis; using the concept of Fisher linear discriminant, with the degree of “aggregations within class” and “discreteness between classes” to evaluate the ability of each dimension features to distinguish natural and hidden carrier, and then select the optimal subset. The analysis directs at the mainstream universal steganalysis model–SPAM model, and the simulation results show that optimal subset has a good detection and low computational complexity.

Evolutionary algorithms (EAs) are population-based general-purpose optimization algorithms, and have been successfully applied in real-world optimization tasks. However, previous theoretical studies often employ EAs with only a parent or offspring population and focus on specific problems. Furthermore, they often only show upper bounds on the running time, while lower bounds are also necessary to get a complete understanding of an algorithm. In this paper, we analyze the running time of the (\(\mu +\lambda \))-EA (a general population-based EA with mutation only) on the class of pseudo-Boolean functions with a unique global optimum. By applying the recently proposed switch analysis approach, we prove the lower bound \(\varOmega (n \ln n\,+\,\mu \,+\,\lambda n\ln \ln n/ \ln n)\) for the first time. Particularly on the two widely-studied problems, OneMax and LeadingOnes, the derived lower bound reveals that the (\(\mu +\lambda \))-EA will be slower than the (1 + 1)-EA when the population size \(\mu \) or \(\lambda \) is above a moderate order. Our results imply that the increase of population size, while usually desired in practice, bears the risk of increasing the lower bound of the running time and thus should be carefully considered.

In many real-world reinforcement learning problems, an agent needs to control multiple actions simultaneously. To learn under this circumstance, previously, each action was commonly treated independently with other. However, these multiple actions are rarely independent in applications, and it could be helpful to accelerate the learning if the underlying relationship among the actions is utilized. This paper explores multi-action relationship in reinforcement learning. We propose to learn the multi-action relationship by enforcing a regularization term capturing the relationship. We incorporate the regularization term into the least-square policy-iteration and the temporal-difference methods, which result efficiently solvable convex learning objectives. The proposed methods are validated empirically in several domains. Experiment results show that incorporating multiaction relationship can effectively improve the learning performance.

In the clustering algorithms, it is a bottleneck to identify clusters with arbitrarily. In this paper, a new method DAPPFC (density-based affinity propagation for parameter free clustering) is proposed. Firstly, it obtains a group of normalized density from the unsupervised clustering results. Then, the density is used for density clustering for multiple times. Finally, the multiple-density clustering results undergo a two-stage synthesis to achieve the final clustering result. The experiment shows that the proposed method does not require the user’s intervention, and it can also get an accurate clustering result in the presence of arbitrarily shaped clusters with a minimal additional computation cost.

For the difficult situation of the node application upgrade in the distributed control system, this paper designs a remote online programming method. This method uses STM32F407 microcontroller and In-Application Programming (IAP) techniques. The upper network is based on Ethernet and http protocol, the lower network is based on UART and CAN. This paper plans the STM32 flash memory based on the lAP characteristics, and introduces the design principle of the distributed IAP. This paper also designs the IAP program and PC user interface program. This method has easy operation, high reliability and good stability, it has a very good value in the distributed system.

Social commerce is where e-commerce meets social networks. Links between social network users are being leveraged to generate and propagate word-of mouth about a products, services and brands, creating new business opportunities, and more specifically, a new channel for online business. However, a close look at the academic and practitioner literature on e-commerce and social networks reveals a clear lack of consensus on social commerce concepts and implementation mechanisms. We argue that without such consensus, social commerce might not reach its full potential. Therefore this paper starts by providing an overview of social commerce research and practice in light of the wide attention it has drawn recently. We then propose a social commerce framework consisting of three key parts, namely implementors, enablers and activities. Using Facebook APIs and plugins, we design, develop and deploy a set of social commerce applications as proof-of-concept of our proposed framework.

Diversifying the results retieved is an effective approach to tackling users’ information needs in Twitter, which typically described by query phrase are often ambiguous and have more than one interpretation. Due to tweets being often very short and lacking in reliable grammatical sytle, it reduces the effectiveness of traditional IR and NLP techniques. However, Twitter, as a social media, also presents interesting opportunies for this task (for example the author information such as the number of statuses). In this paper, we firstly address diversitication of the search results in Twitter with a learning method and explore a series of diversity features describing the relationship between tweets which include tweet content, sub-topic of tweet and the Twitter specific social information such as hashtags. The experimental results on the Tweets2013 datasets demonstrate the effectiveness of the learning approach. Additionally, the Twitter retrieval task achieves improvement by taking into account the diversity features. Finally, we find the sub-topic and Twitter specific social features can help solve the diversity task, especially the post time, hashtags of tweet and the location of author.