Search

Traditional recommendation systems model user preferences based on past historical behaviors, thus unable to obtain dynamic user preferences. The conversational recommendation system (CRS) combines the conversational module with the recommendation module and overcomes the limitations by directly asking the user's preference for attributes. However, the existing CRS methods lack effective information propagation among various modules, making the model lack the basis for making correct decisions. In this paper, we propose an Employer-Employee network, which decomposes the actions into two stages, which are completed by two networks respectively. The Employer Network is responsible for analyzing information and making decisions (query or recommend), and the Employee Network is responsible for collecting information and performing tasks. Our contributions can be highlighted in three aspects: We first emphasize the importance of information propagation among multiple modules in the conversational recommendation system. Secondly, we propose an Employer-Employee (EE) network, which transforms each turn of action into a two-stage decision-making task handed over to two networks to complete. Thirdly, we conduct experiments on multiple datasets, and the experimental results show that our model achieves competitive performance compared with state-of-the-art baselines.

Multi-turn response selection is an important task in natural language processing, which is designed for developing dialogue agents. Existing models on this task mainly extract semantic features of dialogue contexts and rely heavily on linguistic matching for response selection. However, these previous approaches simply consider contextual features and largely ignore the temporal information among utterances. In this paper, we propose a novel graph-based retrieval model to tackle the above problems. We first construct a temporal graph based on both dialogue contexts and utterance relations, and then leverage the gated graph convolutional networks to aggregate significant information from all neighboring utterances. Preciously, we exploit the proposed graph-based architecture to perform accurate reasoning over multi-turn dialogues, capturing semantic and temporal features simultaneously for selecting the appropriate response. Experimental results have shown that our model can achieve strong performance on multi-turn response selection compared to the baseline models. Additionally, ablation studies validate the effectiveness of different components in our model.

Hyperspectral unmixing has been one of the hot spots and has great potential in a wide range of hyperspectral applications. In recent years, the employment of the probabilistic topic model to acquire latent topics of hyperspectral image has been an effective way for spectral unmixing. However, these methods fail to fully exploit the potential of topic models in uncovering image semantics and need extra sparsity constraints, which greatly increases the complexity of the model. To solve these problems, a union framework with sparse topic relaxion and group clustering (USTRGC) is proposed. In USTRGC, the implied sparse prior constraints by the sparse topic model are introduced. Through the relaxation of model, the influence caused by the endmember variability on the unmixing accuracy can be alleviated. Furthermore, unmixing models of different category are united to alleviate the ill-posed nature of the model, thereby improving the fractional abundance. Comprehensive evaluations on two real hyperspectral images, and comparisons with some popular HU methods available in existing research verifies the effectiveness and superiority of the proposed methods.

This paper describes a novel electrostatically actuated microgripper with freeform geometries designed by a genetic algorithm. The novel, semi-automated design methodology can design near-optimal MEMS devices that are robust to fabrication tolerances. The use of freeform geometries significantly improved the performance of the microgripper. Experimental validation showed that the designed microgripper achieved a large displacement (i.e., 91.5 μm) with a low actuation voltage (47.5 V), which agreed well with the theory. The microgripper can handle micro objects with a size from 10 μm to 100 μm. The result confirmed the superior performance of the new design methodology enabling freeform geometries. The design method can also be extended to the design of many other MEMS devices.

Combined with the actual teaching situation of Java Web Programming course in Colleges and universities, this paper analyzes the problems existing in the teaching process of Java Web Programming course, introduces the concept of OBE, clarifies the teaching objectives, reconstructs the teaching content, focuses on the teaching evaluation and optimizes the assessment process, so as to comprehensively improve the students' knowledge, ability and quality.

In order to improve the accuracy of surface classification and meet the needs of sub-pixel-level target detection, spectral unmixing has been one of the hot spots in hyperspectral remote sensing research. The employment of the probabilistic topic model to acquire latent topics of hyperspectral image has been an effective way for spectral unmixing. However, this approach fails to consider the sparsity of the semantic representation and high computational complexity. In addition, the number of endmembers cannot be determined automatically. To solve the problem, in this paper, the novel spectral unmixing method based on semi-automatic fully sparse semantic modeling framework (SFSSM) is proposed. In SFSSM, modestly few arithmetic operations are required to identify the pure spectral signatures (endmembers) and the fractional abundances of the endmembers. Meanwhile, the sparsity and representativeness of the topics generated by SFSSM guarantee that the endmembers can be obtained automatically in low time consumption. The experimental results obtained with two real image confirm that the proposed method significantly improves the performance when compared with the other methods.

This paper proposes a deep learning based approach for multi- meter intelligent detection and recognition under complex background. Firstly, SSD is used to detect the meter images and get the category and boxes of meters. Then the key point detection model is used to get the center point, the zero point, full-scale point and the end point of the pointer. Finally the indication is calculated through those key points. The experimental results show that the method can simultaneously identify different pointer meters, the error of key point is 2.8252 pixels and the average relative error is only 1.37%, which has high robustness and recognition accuracy.

Generative Adversarial Network (GAN) is a kind of deep learning model and one of the most promising methods for unsupervised learning in complex distribution in recent years. Due to the need for higher hardware support, it has certain limitations in practical application. Therefore, how to use the fast-developing Web technology and Cloud technology to promote and share it is a difficult problem. To address this issue, the theme is based on the hot model of GAN, combined with the thinkPHP framework, using python and php language, then, by means of the Cloud computing technology and bootstrap technology, the interactive response system for image style transformation between the mobile terminal and the server terminal is subsequently realized, which conforms to the development trend and expansion characteristics of the mobile terminal. This scheme can be promoted to art creators or the general public to meet their needs for artistic inspiration or image style transformation and ultimately to achieve a complete chain of resource collection, information processing and service feedback.

This paper presents a method for detecting hovering UAVs in ground clutter background. The proposed method is based on both Range-Doppler (RD) map and micro-Doppler signatures, aiming to distinguish the hovering UAV from ground clutter. Frequency modulation characteristic induced by rotating blades is extracted as a unique micro-Doppler signature to discriminate between the hovering UAV and ground clutter. Simulation data is used to verify the performance of the proposed method and results demonstrate the effectiveness of the proposed method.

PCA algorithm is a typical data dimensionality reduction method, which projects high-dimensional data to a lower-dimensional space to obtain a low-dimensional data set that can maximally represent these characteristics of the original data set. The PCA algorithm can effectively achieve dimensionality reduction for high-dimensional data and is widely used in various fields. Aimed at the tedious calculation process of PCA algorithm and the time-consuming of processing massive stream data, this paper proposes a distributed parallel dimensionality reduction algorithm that called DP-PCA by improving the PCA algorithm. Based on the theory of PCA algorithm, DP-PCA algorithm includes three parts of improvement research. Firstly, the original data set is preprocessed by using the “mean” method. Secondly, the solution process of correlation coefficient matrix is improved. Thirdly, this paper designs a distributed parallel dimensionality reduction scheme for DP-PCA algorithm. In addition, this paper deploys DP-PCA algorithm on Storm platform to realize parallelization of the algorithm, and tests the DP-PCA algorithm. Experiments show that DP-PCA algorithm improves computational efficiency and reduces the dimensionality reduction time, and improves the speedup ratio.

Widely accessible biomedical data provides new opportunities to discover knowledge for quality healthcare. However, optimal selection of the most informative, representative and non-redundant features from the huge volume of datasets are to be better explored. To address this challenge, in this paper we propose an integrative clustering and supervised feature selection approach. In our framework, the unsupervised clustering contributes to reducing redundancy by exploring the correlation among features, while supervised learning selects informative and representative features by examining the relation between features and outputs. Our approach was evaluated on a high-dimensional biomedical image feature set and a clinical dataset for survival prediction. Experimental results demonstrated that the selected features had the higher discriminative ability and lower redundancy, and our feature selection outperformed two state-of-the-art methods in terms of feature significance and feature correlation. Furthermore, experiments revealed that redundancy reduction helped handle the overfitting problem.

In order to overcome the shortcomings of original Ant Colony algorithm, such as slow convergence speed and easily falling into the local optimum during solving path planning among multi-agents in dynamic environment, an improved ant colony algorithm is proposed by 1) adjusting the transition probability, 2) adapting the pheromone strength Q value adaptively and introduce the pheromone reduction factor to speed up the algorithm's iterative speed, 3) using the explosive algorithm of the fireworks algorithm to expand the search area to avoid trapping for the deadlock problem. In addition, the collision prediction and path transition evaluation function between Agents are given so as to avoid collisions between multi-agents. The experiment of simulation verified its validity of algorithm in terms of time and iterations.

As traditional radar jamming methods face severe challenges, a new strategy for forwarding jamming using chaff was proposed. Its principle was introduced, its interference performance was calculated, and compound jamming effect and safety of jammer were calculated. A space model for long-distance support of compound jamming was established. The pseudo-random and ergodicity of the Tent chaotic sequence were introduced into the whale optimization algorithm to reduce the probability of the whale optimization algorithm falling into a local optimum, which was used to solve the model. Simulation verification was carried out and relevant airspace layout was obtained. Simulation shows that the model built is scientific and effective.

Nowadays, intrusion detection is a technology to effectively avoid a number of risks of network intrusion. The K-means algorithm is widely used in intrusion detection. But, the algorithm has some shortcomings, such as random selection of k value, sensitive selection of initial cluster centers, and low accuracy in clustering high-dimensional data. In order to make up for these shortcomings of the K-means algorithm, this paper proposes an AE-Kmeans architecture which combines an autoencoder with the improved K-means algorithm. The AE-Kmeans architecture realizes the dimension reduction and feature extraction of these original data by introducing an autoencoder, and uses the improved K-means algorithm to cluster these processed data. These improvements of the improved K-means algorithm mainly include two aspects: the first aspect, a new method is introduced to select initial cluster centers; the second aspect, the algorithm calculates the weight of each attribute with the coefficient of variation, and then uses these weights in Euclidean distance formula, resulting in a weighted Euclidean distance formula. Finally, the AE-Kmeans architecture uses KDD CUP99 data set for intrusion detection simulation experiments. The Experimental results show that the AE-Kmeans architecture not only enhances the ability to deal with high-dimensional data, but also improves the detection rate and reduces the error-detection rate compared with K-means algorithm.

Breast mass segmentation in digital mammography is one of the most significant methods of breast cancer prevention. An integrated approach for mammographic mass segmentation is proposed in this paper. Given a mammographic image, it is first eliminated interference and enhanced in the preprocessing states. Then, the preprocessed images are detected and segmented by level set method. A preliminary evaluation of the proposed method performs on a known public database, namely Mini Mammographic Image Analysis Society (MIAS) database. In order to present that our proposed algorithm is more excellent, we make comparison experiments to demonstrate that our proposed approach can potentially obtain better masses detection results.

The research of wearable instruments is a new research direction which accompanies with the development trends of measure instrument, such as miniaturization, virtualization and intelligence. One of the important research projects of the wearable instrument system design is offer convenient, flexible and reliable interactive input methods to engineering technicians who are cope with testing and maintaining task of huge and complex equipments. This study attempts to explore design patterns of the separation between the instrument panel and functions using the platform of the single-handed keyboard mouse interaction device, brings in inertial sensor measurement units, and puts forward a new kind of interaction method which combines keyboard stroke and gesture movements. All the functions of traditional instrument panel are realized based on this platform, and some conduct experiments are created to evaluate the interactive performance of this new interaction method.

This paper mainly discusses the structures in the laboratory of a fiber-optic WIM system. This system is based on the feature of the fiber that when it vertically pressed by heavy objects the light intensity of the fiber change with them. This experiment use CCD camera, splitter and other professional optical equipment to capture the light intensity changes video. And then, use computer to obtain the image from the video stream, grayscale process, binary process, corrosion process and expansion process the image to calculate the light intensity. The experimental results reflect some regular relationship of the experimental vehicle weight and the intensity of the fiber and speed of the vehicle.

Most university teachers pay attention to stimulating the students' learning motivation. Moreover, stimulating the students' learning motivation is a capability of IBSTPI (the Internet Board of Standards for Training, Performance and Instruction). In this context, this research aims to design and implement a plan of training activity to enhance university novice teachers' stimulating the students' learning motivation skills through school-based training. Finally, according to the classification of teachers' knowledge, we construct the training object and content system to design the activities based on the Network Platform of Moodle and Course review platform based on online video for university novice teachers.

Considering the expense, the security of communication and demand in practice, a strategy is proposed, which propagates replicated data based on message and transaction-log table to update and maintain the database. It emphasizes that updates to a replica copy should been executed locally as much as possible avoiding using two-phase commit protocol operated directly on a remote copy in order to improve system performance. The message mechanism can prevent conflicts when two or more users update the same copy, and the transaction-log table is established to log up-to-date write transactions on a copy for a update propagation to other copies, avoiding access the original copy with large amounts of data.

In order to improve the automatic control of water and electricity resource, an intelligent resource management system for community based on a sensor network is presented for the reason that sensor network has the advantage of low-power transmission. It is provided by a number of experiments that this system has successfully achieved the automatic control goal by rotational solar-energy heater, mechanical control techniques and the intelligent management of the entire system. The overall architecture for electrical control and mechanical control of the system is introduced in order to describe the intelligent management of the system. Finally, the theoretical and experimental analyses are presented to support the validity and advances of this system.

This paper presents a novel optimal redundant inertial sensors configuration in SINS with overall considerations of reliability, navigation accuracy, size and cost of system. The optimal number of redundant sensors is determined by considering the reliability and cost. Based on D-optimality criteria and real code genetic algorithm (RCGA), a novel optimal redundant configuration scheme is established. The experiment results show that the proposed redundant sensors configuration achieves required accuracy with smaller size which is 26.4% of the widely adopted redundant dodecahedron configuration. The optimal experiment method proposed in this paper has been proved to be simple and efficient for designing redundant sensor configuration.

This paper is concerned with Radial Basis Function (RBF) for restoration of damaged images. We introduce the RBF into the image inpainting, and the 2D image inpainting problem is naturally converted to 3D implicit surface reconstruction problem,which has been used successfully. Otherwise, we do some experiments with several kinds of damaged images, the results showed that the use of the RBF method for image processing applications is feasible.

Quality assessment of Multiobjective Optimization Evolutionary Algorithms (MOEA) has been a major concern in scientific field during the last few years. The entropy metric is introduced and highlighted in computing the diversity of MOEA. Existing performance metrics are briefly reviewed. The advantages of entropy approach over other diversity metrics are classified here. Besides, the definition of the entropy metric and the approach of measuring diversity based on entropy are presented, in which section a detailed explanation about how to choose appropriate parameters in entropy measurement and programming techniques are given. Above all, some significant conclusions arised from the further study of this entropy metric are summarized, which may have a meaningful impact on the high-efficient comparisons of MOEA performance on a quantitative basis. Last but not the least, it will help researchers to determine whether or not a population of solutions has reached maturity, in order to end the optimization process evaluation process.

Based on circular finned heat pipe has a big resistance and is easy to be deposited in the air-conditioning heat recovery system, Zhuang Chen et al [1] proposed to apply the separating longitudinal finned heat pipe to the air-conditioning heat recovery system. In this paper, the authors made a three-dimensional numerical simulation of a basic unit of 500 mm long longitudinal finned bundle and circular finned bundle using Fluent software. Through the comparison of heat transfer coefficient, power coefficient and other parameters, the authors conclude that the longitudinal finned heat pipe used in the air-conditioning heat recovery system has certain advantages.

Coastal zone detection is an important task in sustainable development and environmental protection. For coastline zone monitoring, extraction and analysis of coastline changes are an important task. Detection of coastline changes by satellite images and remote sensing methods became increasingly important over the recent decades. This study extracted mean high tide lines of the Yellow River estuary from Landsat MSS, TM and ETM+ data from 1976 to 2005, totaling 27 scenes. Results show that: The general pattern of accretion-erosion of the entire estuary was divided into four stages: The rapid accretion stage (1976–1986), accretion-erosion adjustment stage (1985–1996), slow erosion stage (1997–2003), and slow accretion stage (2004-present). The relation between the area and length is presented in Fig.7. There is a clear quadratic trend between the area and length. The relationship has a correlation coefficient, R=0.862, and significant level, P=0.001. For the Q8 course estuary, the total area increased by 8.52 km2 with a mean net accretion rate of 0.95 km2/a during the period from 1996 to 2005. Due to the low runoff and sediment load, the Yellow River delta could be eroded naturally for several years in the future.

Wireless sensor network is composed of hundreds of sensor nodes involving in limited energy, efficient and low-energy consuming routing algorithm is the crucial problem to the routing design. In this paper, the LEACH protocol, which is a typical clustering routing protocol, has been researched and improved based on energy, the new algorithm is different from the LEACH in the method of determining the number of clusters heads and the criterion that chooses cluster-heads. Simulation results indicate that the new algorithm has the advantages of reducing energy consuming and prolonging the lifetime of the sensor network.

Ecological restoration has been coming to be an important task in China, especially in resource-exhausted areas. Landscape restoration regionalization is a necessary part of it. This paper discussed the aim and meaning of landscape restoration regionalization firstly, ascertained the process for it. Then taking the abandoned mining areas of Mentougou District (MTG) of Beijing as an example, the author partitioned the landscape restoration regions of MTG based on GIS. The suitable function of each unit was confirmed, and the landscape units and structure which need to be restored was found. In this paper, 16 landscape restoration regions were partitioned, and the restoration orientation and restoration manner were also confirmed.