It is often the case that in the development of a system-on-a-chip (SoC) design, a family of SystemC transaction level models (TLM) is created. TLMs in the same family often share common functionalities but differ in their timing, implementation, configuration and performance in various SoC developing phases. In most cases, all the TLMs in a family must be verified for the follow-up design activities. In our previous work, we proposed to call such family TLM product line (TPL), and proposed feature-oriented (FO) design methodology for efficient TPL development. However, developers can only verify TLM in a family one by one, which causes large portion of duplicated verification overhead. Therefore, in our proposed methodology, functional verification of TPL has become a bottleneck. In this paper, we proposed a novel TPL verification method for FO designs. In our method, for the given property, we can exponentially reduce the number of TLMs to be verified by identifying mutefeature-modules (MFM), which will avoid duplicated veri-fication. The proposed method is presented in informal and formal way, and the correctness of it is proved. The theoretical analysis and experimental results on a real design show the correctness and efficiency of the proposed method.

Industrial control systems (ICSs) are distributed complex cyber physical systems (CPSs). The core parts of ICSs are usually ultimate targets of attackers but being protected strictly and difficult to attack directly. Attackers tend to take cyber-physical attacks via multiple attack-steps from outer nodes with less protection. Then, threats propagate from one node to another until reaching the core assets or being detected. This paper develops a unified methodology which can model and analyze the cyber-physical attacks on ICSs quantitatively and automatically. We define the weighted colored Petri net and propose the basic cyber-physical attack models. Second, we propose the method to calculate the weights in attack models by modeling threat propagation between two nodes as mixed-strategy Bayesian attack-defense game with incomplete information and solving the refined Bayesian Nash Equilibrium. Results show that weights in the cyber-physical attack model (CPAM) are nearly stable even when parameters change. Moreover, we build threat propagation matrix and security state vector and further design the cyber-physical attack path analysis algorithm which can discover possible attack paths with specific attack losses. The case study verifies our methodology and evaluations show that our methodology has good time performance within limited node number. This work can help protect ICSs more efficiently.

With the development of integrated circuit design technology, soft errors have become an important threat to system reliability, and software-based fault-tolerant techniques are gradually attracting people's attention. In many cases, researchers use fault injection techniques that are less observable and less controllable to verify system reliability, and at this point, analysing where soft errors occur requires considerable work. In this paper, we present EBSCN, an error backtracking method. The EBSCN method sorts functions by suspiciousness by analysing the erroneous output results, which will help researchers reduce the amount of work required for analysis. The EBSCN method includes a feature extraction method based on clustering and a feature analysis method based on a deep neural network. This paper introduces the principle of the two methods as well as methods to improve and extend them with program-related information. We discuss the effect of the scale of the output result and the severity of the error on the EBSCN method through experiments and verify the effect of the EBSCN method. The results showed that the proportion of the function in which the soft error actually occurs in the ranking of the top 25% of the suspiciousness sequence is no less than 82%, and the proportion ranked in the top 50% is no less than 97%.

错误定位就是寻找程序错误的位置.现有的错误定位方法大多利用测试用例的覆盖信息,以标识一组导致程序失效的可疑语句,却忽视了这些语句相互作用导致失效的上下文.因此,提出一种增强上下文的错误定位方法Context-FL,以构建上下文的方式来优化错误定位性能.Context-FL利用动态切片技术构建数据与控制相关性的错误传播上下文,显示了导致失效的语句之间传播依赖关系;然后,基于可疑值度量来区分上下文片段中不同语句的可疑度;最后,Context-FL以标记可疑值的上下文作为定位结果.实验结果表明,Context-FL优于8种典型错误定位方法.

Most of the existing approaches of automatic key phrase extraction ignored the effects of sen-tences.And they usually ignored the effects of topics.That had negative effects on key phrase extraction. A new approach of automatic key phrase extraction based on graph and clustering was proposed.And it could consider the effects of the sentences and topics in articles.The experimental results showed that the new approach had better performance on key phrase extraction than all existing approaches.

机器阅读理解是自然语言处理领域一项得到广泛关注与研究的任务.该文针对中文机器阅读理解数据集DuReader,分析其数据集的特点及难点,设计了一种基于循环神经网络和自注意力机制的抽取式模型Mixed Model.通过设计段落融合等策略,该文提出的模型在DuReader测试集上达到了54.2的Rouge-L得分和49.14的Bleu-4得分.

In the field of space computing, DSP is more and more used for its high performance. Like other non-radiation-resistant chips, COTS DSP is easily affected by high-energy particle irradiation in the space environment, which can prone to transient faults, also known as soft errors. Soft error is one of the serious problems to the reliability of space computing. To evaluate the reliability of the software system based on COTS DSP, we designed MDSPFI, a new Fault Injection (FI) framework based on DSP C6678. As an emulation-based FI framework, MDSPFI is much more efficient in performance than pure SWIFI, and has higher controllability than pure hardware implemented FI tools. We designed FI script as autonomous emulation, thus, avoiding the intensive communication between the host and the emulation platform. Moreover, we accelerate the average per-experiment runtime by running on hardware emulator as well as conducting multithreaded execution with the multicore architecture, and reduces the FI experiment count by adopting the Fault pruning techniques. Experimental results show that MDSPFI has fairly well correctness and effectiveness.

Gated recurrent unit (GRU) networks perform well in sequence learning tasks and overcome the problems of vanishing and explosion of gradients in traditional recurrent neural networks (RNNs) when learning long-term dependencies. Although they apply essentially to financial time series predictions, they are seldom used in the field. To fill this void, we propose GRU networks and its improved version for predicting trading signals for stock indexes of the Hang Seng Indexes (HSI), the Deutscher Aktienindex (DAX) and the S&P 500 Index from 1991 to 2017, and compare the GRU-based models with the traditional deep net and the benchmark classifier support vector machine (SVM). Experimental results show that the two GRU models proposed in this paper both obtain higher prediction accuracy on these data sets, and the improved version can effectively improve the learning ability of the model.

In the space environment, computer is easily affected by high-energy particle irradiation, which may incur transient faults, also known as soft errors. Soft error is one of the serious problems for space computation, making the space computer system unreliable. Many efforts have been made in this research area. One of them is fault injection technique, an experimental method for software reliability evaluation. To evaluate the software reliability running on x86/x64 architecture, we have designed a fault injection framework SEInjector. Like many other fault injection tools, SEInjector has been designed as making statistically sampling by randomized selection of fault locations. But, in some cases, we need detail analysis for fault-tolerance related design decisions, and full fault space exploration is required. However, the vast fault space makes the complete fault space exploration impractical. In this paper, we propose several methods to prune the fault space. We prune the known-outcome faults by the registers usage information, and we prune faults leading to the same outcome by classify the faults into equivalence classes. Experiment and calculation show that using these methods, the full fault space coverage time consumption can be largely reduced.

Determining the validity of knowledge triples and filling in the missing entities or relationships in the knowledge graph are the crucial tasks for large-scale knowledge graph completion. So far, the main solutions use machine learning methods to learn the low-dimensional distributed representations of entities and relationships to complete the knowledge graph. Among them, translationmodels obtain excellent performance. However, the proposed translation models do not adequately consider the indirect relationships among entities, affecting the precision of the representation. Based on the long short-term memory neural network and existing translation models, we propose a multiple-module hybrid neural network model called TransP. By modeling the entity paths and their relationship paths, TransP can effectively excavate the indirect relationships among the entities, and thus, improve the quality of knowledge graph completion tasks. Experimental results show that TransP outperforms state-of-the-art models in the entity prediction task, and achieves the comparable performance with previous models in the relationship prediction task.

The deep neural named entity recognition model automatically learns and extracts the features of entities and solves the problem of the traditional model relying heavily on complex feature engineering and obscure professional knowledge. This issue has become a hot topic in recent years. Existing deep neural models only involve simple character learning and extraction methods, which limit their capability. To further explore the performance of deep neural models, we propose two character feature learning models based on convolution neural network and long short-term memory network. These two models consider the local semantic and position features of word characters. Experiments conducted on the CoNLL-2003 dataset show that the proposed models outperform traditional ones and demonstrate excellent performance.

To analyze the machine learning technology in the field of medical data mining. Firstly, the state-of-art technologies in medical data mining was analyzed in this paper. Secondly, the concept of machine learning and related technology were introduced, and the application of machine learning technology in the recuperation field was explained. Finally, the application of machine learning technology in the field of military recuperation was pointed out. The adoption of machine learning technology in medical data mining greatly helps to improve the utilization of military medical data, strengthen medical services, develop complete medical policies, and configure medical resources.

In this paper, based on our previous multi-pattern uniform resource locator (URL) binary-matching algorithm called HEM, we propose an improved multi-pattern matching algorithm called MH that is based on hash tables and binary tables. The MH algorithm can be applied to the fields of network security, data analysis, load balancing, cloud robotic communications, and so on-all of which require string matching from a fixed starting position. Our approach effectively solves the performance problems of the classical multi-pattern matching algorithms. This paper explores ways to improve string matching performance under the HTTP protocol by using a hash method combined with a binary method that transforms the symbol-space matching problem into a digital-space numerical-size comparison and hashing problem. The MH approach has a fast matching speed, requires little memory, performs better than both the classical algorithms and HEM for matching fields in an HTTP stream, and it has great promise for use in real-world applications.

For the digital signal processors (DSP), a new approach of detecting soft errors is proposed to overcome the transient fault, named EDIO. The goal of EDIO is to enhance the reliability of a DSP software system with reduced performance overhead, compared to former software fault tolerant techniques. EDIO employs diverse instructions for the sake of furthest exploring errors and increasing performance. The loop optimization delays the fault tolerance latency between errors detecting and errors handling, only in this way can EDIO schedule the software pipeline to reduce performance overhead significantly. We evaluate EDIO by the performance experiments and the ion irradiation experiments, which demonstrate exceptional fault coverage with a reasonable performance cost. Compared to the well-known software approach, EDIO demonstrates 6.4 times average speedup while remaining the ability to detect faults.

Soft errors, mainly caused by Single event upsets in space, is one of the most serious problems for space applications running in hash environments. For accurately assessing the reliability of these applications toward soft errors, researchers always apply fault injection techniques to simulate the effects aroused by SEU. This paper presents a new fault injection tool, SEInjector. Based on the dynamic instrument framework PIN, SEInjector can inject faults into applications for x86 at runtime, without revision or recompilation. This tool provides the following advantages: 1. It can confine the fault injection within a specified code segment, which provides a method to analysis the effectiveness of fault tolerance techniques towards that code segment. 2. It deletes the unused register file from the target, thus prunes the fault injection space. 3. The fault injection processes can be executed with multi-thread.

Fault localization is essential for solving the issue of software faults. Aiming at improving fault localization, this paper proposes a deep learning-based fault localization with contextual information. Specifically, our approach uses deep neural network to construct a suspiciousness evaluation model to evaluate the suspiciousness of a statement being faulty, and then leverages dynamic backward slicing to extract contextual information. The empirical results show that our approach significantly outperforms the state-of-the-art technique Dstar.

Existing word-frequency-based algorithms for keyword extraction do not consider the semantic relationships among words. Moreover, word-graph-based algorithms cannot distinguish multiple topics, and topic-model-based algorithms possess high time complexity. All of these keyword extraction algorithms exhibit limitations. This paper proposes a new word-embedding-based algorithm, namely, WEC, for keyword extraction. The algorithm incorporates word frequency, effects of word co-occurrence, and semantic relationship among contexts. The algorithm also estimates the final weights of words with cosine similarity and pointwise mutual information and extracts topics by clustering. Experimental results show that the WEC algorithm outperforms state-of-the-art keyword extraction methods on four datasets when tested under various evaluation metrics.

This article focuses on the study and improvement of fingerprint location algorithm and designs, and implements the wireless-free internet indoor location system based on an Android platform. We have attempted to improve the accuracy of indoor location results. Our approach consisted of two main algorithms: A sampling algorithm uses a received signal strength indicator vector to construct the location fingerprint database. We tested 20 times at each sample point and stored the average result in the location fingerprint database to avoid errors introduced by signal fluctuations. A matching algorithm analyzes the location of the test point based on a K-nearest neighbor (K-NN) algorithm. Here, we introduced penalty to improve the accuracy of the K-NN algorithm. We evaluated the system in a busy engineering building and demonstrated localization accuracies in a 5m x 5m area. The results show that our system is capable of completely controlling the error in distances under 1.5m, and is promising in contrast with the results of previous studies.

Existing graph-based ranking techniques for keyphrase extraction only consider the connections between words in a document, ignoring the impact of the sentence. Motivated by the fact that a word must be important if it appears in many important sentences, we propose to take full advantage of the reinforcement between words andsentences by melting three kinds of relationships between them. Moreover, a document is grouped with many topics. The extracted keyphrases should be synthetic in the sense that they should deal with all the main topics in a document. Inspired by this, we take topic model into consider. Experimental results show that our approach performs betterthan state-of-the-art keyphrase extraction method on two datasets under three evaluation metrics.

This paper proposes a DSP fault tolerant approach based on loop optimization known as DSP Loop Optimization Approach (DLOA). DLOA reduces the performance overhead incurred by traditional fault tolerance techniques while maintaining their fault tolerance capabilities. DLOA delays the fault tolerance latency between errors detecting and errors handling to scheduling the software pipeline, increasing performance significantly. The performance experiments and the ion irradiation experiments in the Heavy Ion Research Facility in Lanzhou (HIRFL) demonstrated that DLOA used in SWIFT achieved a 6.2 times average speedup and with its fault tolerance ability unaffected. Keywords: DSP; Transient fault; Fault tolerance; Loop optimization.