In this paper, we analyze the rollback traffic in polar-orbit satellite networks that use the snapshot routing algorithm. The concept of diamond rollback links and polar rollback links are presented for the first time, and the numbers of diamond and polar rollback links in polar-orbit satellite networks are concisely formulated. Simulations are performed based on the Iridium and Teledesic system in NS2, and the results finally confirm our analysis. With this work, we can not only simplify the rollback loops avoidance scheme, but also provide guidance for future satellite network routing optimization and topology design.

Recent researches on mobile robots show that convolutional neural network (CNN) has achieved impressive performance in visual place recognition especially for large-scale dynamic environment. However, CNN leads to the large space of image representation that cannot meet the real-time demand for robot navigation. Aiming at this problem, we evaluate the feature effectiveness of feature maps obtained from the layer of CNN by variance and propose a novel method that reserve salient feature maps and make adaptive binarization for them. Experimental results demonstrate the effectiveness and efficiency of our method. Compared with state of the art methods for visual place recognition, our method not only has no significant loss in precision, but also greatly reduces the space of image representation.

Comparing with that of terrestrial networks, the location management in satellite networks is mainly restricted by three factors, i.e., the limited on-board processing (OBP), insufficient link resources and long propagation delay. Under these restrictions, the limited OBP can be smoothened by terrestrial gateway-based location management, the constraint from link resources demands the bandwidth-efficient management scheme and long propagation delay potentially lowers the management efficiency. Currently, the reduction of the management cost has always been the main direction in existing work which is based on the centralized management architecture. This centralized management has many defects, such as the non-optimal routing, scalability problem and single point of failure. To address these problems, this paper explores gateway-based distributed location management schemes for Low Earth Orbit (LEO) satellite networks. Three management schemes based on terrestrial gateways are proposed and analyzed: loose location management, precise location management, and the grouping location management. The analyses specifically analyze the cost of location queries and show their significant influence on the total cost which includes the location management and query. Starting from the above analysis, we speculate and prove the existence of the optimum scheme in grouping location management, which has the lowest total cost for the query frequency within given range. Simulation results validate the theoretical analysis on the cost and show the feature of latency in location queries, which provide a valuable insight into the design of the distributed location management scheme in satellite networks.

The large Doppler offset that exists in high dynamic environments poses a serious impediment to the acquisition of direct sequence spread spectrum (DSSS) signals. To ensure acceptable detection probabilities, the frequency space has to be finely divided, which leads to complicated acquisition structures and excessively long acquisition time at low SNR. A local frequency folding (LFF) method designed for combined application with established techniques dedicated to PN-code synchronization is proposed in this paper. Through modulating local PN-code block with a fixed waveform obtained by folding all frequency cells together, we eliminate the need for frequency search and ease the workload of acquisition. We also analyze the performance of LFF and find that the detection performance degradation from folding can be compensated by FFT-based coherent integration. The study is complemented with numerical simulations showing that the proposed method has advantages over unfolding methods with respect to detection probability and mean acquisition time, and the advantage becomes obvious but limited if the folded number gets larger.

Collaborative filtering with only implicit feedbacks has become a quite common scenario (e.g. purchase history, click-through log, and page visitation). This kind of feedback data only has a small portion of positive instances reflecting the user's interaction. Such characteristics pose great challenges to dealing with implicit recommendation problems. In this letter, we take full advantage of matrix factorization and relative preference to make the recommendation model more scalable and flexible. In addition, we propose to take into consideration the concept of covisitation which captures the underlying relationships between items or users. To this end, we propose the algorithm Integrated Collaborative Filtering for Implicit Feedback incorporating Covisitation (ICFIF-C) to integrate matrix factorization and collaborative ranking incorporating the covisitation of users and items simultaneously to model recommendation with implicit feedback. The experimental results show that the proposed model outperforms state-of-the-art algorithms on three standard datasets.

The main idea of filter methods in feature selection is constructing a feature-assessing criterion and searching for feature subset that optimizes the criterion. The primary principle of designing such criterion is to capture the relevance between feature subset and the class as precisely as possible. It would be difficult to compute the relevance directly due to the computation complexity when the size of feature subset grows. As a result, researchers adopt approximate strategies to measure relevance. Though these strategies worked well in some applications, they suffer from three problems: parameter determination problem, the neglect of feature interaction information and overestimation of some features. We propose a new feature selection algorithm that could compute mutual information between feature subset and the class directly without deteriorating computation complexity based on the computation of partitions. In light of the specific properties of mutual information and partitions, we propose a pruning rule and a stopping criterion to accelerate the searching speed. To evaluate the effectiveness of the proposed algorithm, we compare our algorithm to the other five algorithms in terms of the number of selected features and the classification accuracies on three classifiers. The results on the six synthetic datasets show that our algorithm performs well in capturing interaction information. The results on the thirteen real world datasets show that our algorithm selects less yet better feature subset.

The extraction of micro-motion parameters is deeply influenced by the precision of estimation on translational motion parameters. Based on the periodicity of micro-motion, the quadratic polynomial fitting is carried out among range delays to align envelope. The micro-motion component of phase information is eliminated by conjugate multiplication after which the translational motion parameters are estimated. Then the translational motion is precisely compensated through the third order polynomial fitting. Results of simulation demonstrate that the algorithm put forward here can realize the precise compensation for translational motion parameters even under an environment with low signal noise ratio (SNR).

Software-defined networking (SDN) is recognized as the next-generation networking paradigm. The software-defined architecture for underwater acoustic sensor networks (SDUASNs) has become a hot topic. However, the current researches on SDUASNs is still in its infancy, which mainly focuses on network architecture, data transmission and routing. There exists some shortcomings that the scale of the SDUASNs is difficult to expand, and the security maintenance is seldom dabble. Therefore, a scalable software-definition architecture for underwater acoustic sensor networks (SSDUASNs) is introduced in this paper. It realizes an organic combination of the knowledge level, control level, and data level. The new nodes can easily access the network, which could be conducive to large-scale deployment. Then, the basic security authentication mechanism called BSAM is designed based on our architecture. In order to reflect the advantages of flexible and programmable in SSDUASNs, security authentication mechanism with pre-push (SAM-PP) is proposed in the further. In the current UASNs, nodes authentication protocol is inefficient as high consumption and long delay. In addition, it is difficult to adapt to the dynamic environment. The two mechanisms can effectively solve these problems. Compared to some existing schemes, BSAM and SAM-PP can effectively distinguish between legal nodes and malicious nodes, save the storage space of nodes greatly, and improve the efficiency of network operation. Moreover, SAM-PP has a further advantage in reducing the authentication delay.

In this paper, we study the end-to-end TCP performance over a path deploying a High-Availability cluster, whose characteristics are highlighted by the failover procedure to remove single-point failure. This paper proposes an approach, called High-Availability Local Recovery (HALR), to enhance TCP performance in the face of a cluster failover. To minimize the latency of retransmission, HALR saves TCP packets selectively and resends them locally after the failover is finished. For better understanding, we further develop simple analytic models to predict the TCP performance in the aspect of flow latency under a range of failover times and the effects of HALR. Using simulation results, we validate our models and show that HALR improves the TCP performance significantly over a failover event as compared with the original TCP. Typically, HALR reduces the flow latency from 4.1 sec to less than 1.9 sec when the failover time equals to 500 ms. The simulation by real packet trace further demonstrates that the memory requirement of the proposed solution is not a concern for modern network equipments.

Border Gateway Protocol (BGP), with its advantages in routing isolation support and mature application, is a promising candidate to integrate satellite systems into the terrestrial IP network. However, with more and more ground stations accessing satellites by BGP, a significant amount of routing overhead can be produced on limited satellite links, especially for geostationary satellite networks with thousands of accessing terminals in extremely large areas. To solve this challenge, multicast transport of BGP was proposed, which takes advantage of the inherent broadcast property of wireless channels. However, its performance can be seriously degraded when interfered with the environment. In this paper, NCSR (Network Coding for Satellite network BGP Routing transport) [1] is explored in depth. Unlike existing counterparts, NCSR pays more attention to the lossy space links and can achieve reliability with more bandwidth savings. A greedy based coding algorithm is proposed to realize the network coding operation. To demonstrate the efficiency of NCSR, we conduct theoretical analyses and extensive simulations in typical scenarios of satellite systems. Simulation results show that NCSR can greatly reduce the bandwidth usage while achieving comparable latency. Discussions on practical considerations when applying network coding method for reliability assurance are also presented in detail.

In this paper, a new video broadcast protocol is proposed for video-on-demand (VoD) in shared environment. The new protocol is developed by modifying the first segment delivery scheme for the skyscraper protocol using the idea of patching. The results show that the start-up latency for users is greatly reduced when using our new protocol.

Sketch face recognition is to match sketch face images to photo face images. The main challenge of sketch face recognition is learning discriminative feature representations to ensure intra-class compactness and inter-class separability. However, traditional sketch face recognition methods encouraged samples with the same identity to get closer, and samples with different identities to be further, and these methods did not consider the intra-class compactness of samples. In this paper, we propose triplet-margin-center loss to cope with the above problem by combining the triplet loss and center loss. The triplet-margin-center loss can enlarge the distance of inter-class samples and reduce intra-class sample variations simultaneously, and improve intra-class compactness. Moreover, the triplet-margin-center loss applies a hard triplet sample selection strategy. It aims to effectively select hard samples to avoid unstable training phase and slow converges. With our approach, the samples from photos and from sketches taken from the same identity are closer, and samples from photos and sketches come from different identities are further in the projected space. In extensive experiments and comparisons with the state-of-the-art methods, our approach achieves marked improvements in most cases.

Sequences with good correlation properties are widely used in engineering applications, especially in the area of communications. Among the known sequences, cyclotomic families have the optimal autocorrelation property. In this paper, we decide the cross-correlation function of the known cyclotomic sequences completely. Moreover, to get our results, the relations between the multiplier group and the decimations of the characteristic sequence are also established for an arbitrary difference set.

The extended binary phase shift keying (EBPSK) transmission system with ultra narrow bandwidth has excellent BER performance, which raises many doubts with the researchers. Therefore, on the premise of the existence of a special filter that can transform the modulated phase information into amplitude information, the theoretical BER formula of EBPSK system in Additive White Gaussian Noise (AWGN) channel has been deduced. This paper gives the theoretical values of the parameters in the above BER formula and discusses the effects of parameters on BER firstly. Then the paper shows that the special impacting filter satisfies the above assumption, therefore, in the frame of binary detection theory, the excellent performance of high-efficiency EBPSK system can be explained and the correction of the theoretical BER formula can be validated.

A large-range switchable RF signal generator is demonstrated using a triple-wavelength fiber laser with uneven-frequency-spacing. Due to the birefringence characteristics of the triple-wavelength fiber laser, switchable dual-wavelength operation can be obtained by adjusting a polarization controller. Therefore, we can achieve a stable RF signals at microwave or millimeter-wave band.

To provide the processor arrays with adequate fault-tolerant capabilities, a number of spare or redundant processors are prepared within the arrays. For such processor arrays, reconfiguration should be executed to bypass faulty processors. Concerning reconfiguration of processor arrays, Melhem presented a minimization problem (called the SPA problem). The SPA problem is to find an assignment of spare processors to faulty processors that minimizes the number of dangerous processors. Here, the dangerous processors are processors, for which there remains no longer any spare processor to be assigned when one more faults occur. In this paper, we present a more rigorous definition of the SPA problem, in which input parameters are n2 ordinary processors, 2n spare processors and m (mn2) faulty processors, and the output is an optimal assignment of spare processors to faulty processors, in the sense that the number of dangerous processors is minimum. Then, we develop an efficient algorithm based on the necessary and sufficient conditions, which allows highly efficient computation of spare processor assignment. The worstcase time complexity of the proposed algorithm is O(n2).

When a group of developers are involved in the distributed development of some software product, they must communicate with one another frequently to exchange information about the product. To reduce the penalty of communication, the support environment should provide developers with their necessary information and update the information automatically while the product is modified by developers. Furthermore, the environment must meet the following requirements despite of workstation failures: whether a specific information is correct or not should always be decidable; as much information as possible should be updated correctly and efficiently. This paper presents a framework to construct such a fault-tolerant environment based on attribute grammars. In the framework, a product is represented by an attributed tree, which is partitioned into several subtrees {T1,,Tm}. Attribute values in each subtree Ti(1im) express the information about the product required by a developer. We introduce a set of redundant data and algorithms to meet the fault-tolerance requirements mentioned above. The correctness of an attribute value in Ti can then be decided in O(mn0log n) time, where n0n, and n is the number of attribute instances in Ti. All available attribute values can be updated with time complexity O(m2n1 log n) and communication complexity O(m2), where n1 is the number of attribute instances that must be reevaluated.

In recent years, deep learning has been widely applied in relation extraction task. The method uses only word embeddings as network input, and can model relations between target named entity pairs. It equally deals with each relation mention, so it cannot effectively extract relations from the corpus with an enormous number of non-relations, which is the main reason why the performance of relation extraction is significantly lower than that of relation classification. This paper designs a deep reinforcement learning framework for relation extraction, which considers relation extraction task as a two-step decision-making game. The method models relation mentions with CNN and Tree-LSTM, which can calculate initial state and transition state for the game respectively. In addition, we can tackle the problem of unbalanced corpus by designing penalty function which can increase the penalties for first-step decision-making errors. Finally, we use Q-Learning algorithm with value function approximation to learn control policy π for the game. This paper sets up a series of experiments in ACE2005 corpus, which show that the deep reinforcement learning framework can achieve state-of-the-art performance in relation extraction task.

Software defect prediction (SDP) plays a significant part in allocating testing resources reasonably, reducing testing costs, and ensuring software quality. One of the most widely used algorithms of SDP models is Naive Bayes (NB) because of its simplicity, effectiveness and robustness. In NB, when a data set has continuous or numeric attributes, they are generally assumed to follow normal distributions and incorporate the probability density function of normal distribution into their conditional probabilities estimates. However, after conducting a Kolmogorov-Smirnov test, we find that the 21 main software metrics follow non-normal distribution at the 5% significance level. Therefore, this paper proposes an improved NB approach, which estimates the conditional probabilities of NB with kernel density estimation of training data sets, to help improve the prediction accuracy of NB for SDP. To evaluate the proposed method, we carry out experiments on 34 software releases obtained from 10 open source projects provided by PROMISE repository. Four well-known classification algorithms are included for comparison, namely Naive Bayes, Support Vector Machine, Logistic Regression and Random Tree. The obtained results show that this new method is more successful than the four well-known classification algorithms in the most software releases.

A typical fault-tolerance technique of systolic arrays is to include redundant processors and links so that the array is reconfigurable when some processors fail. Another typical technique is to implement each processor by a majority voter and N (N3) copies of processors so that the faults of up to N-2 copies of processors can be masked without reconfiguration. This paper proposes a systolic linear array called reconfigurable modular redundant linear array (RMA) that combines these techniques with N4. When up to 2 copies of each processor fail in RMA, the faults can be masked without reconfiguration. When some voters or more than 2 copies of a processor fail, RMA can be reconfigured by specifying a new switch pattern. In order to perform reconfiguration efficiently, we present a reconfiguration algorithm with time complexity O (n), where n is the number of processors in RMA.