Search

Heterogeneous information network (HIN) embedding is fundamental but extremely challenging because of the existence of various types of nodes and links. Expressive metagraphs are at the core of HIN embedding methods seeking to match the desired semantics; however, they are typically expensive to obtain and have a high risk of missing meaningful paths among nodes. Instead of heavily relying on domain knowledge to enumerate metagraphs among exponential numbers of potential metagraphs, we propose a novel representation learning model that mines the complex semantics underlying HINs. First, a set of meaningful metagraphs are first constructed to model the core semantic relations. Then, we generate semantic sequences via random walks, which are guided by metagraphs. A complex semantic augmented method is suggested to concatenate sequences with shared nodes for capturing semantic paths between distant nodes, which is hard to extract at the level of metapaths and easily fail to mine without domain knowledge. A pretraining and fine-tuning framework employing adversarial learning that generates semantic-preserving and robust representations, with semantically augmented paths reflecting long-distant correlations of multiple types of nodes, is investigated. We conduct the experiments on a real-life HIN, and the results show that our model performs better than the state-of-the-art alternatives.

Most Cross-modal hashing methods do not sufficiently exploit the discrimination power of semantic information when learning hash codes, while often involving time-consuming training procedures for large-scale dataset. To tackle these issues, we first formulate the learning of similarity-preserving hash codes in terms of orthogonally rotating the semantic data to hamming space, and then propose a novel Fast Semantic Preserving Hashing (FSePH) approach to large-scale cross-modal retrieval. Specifically, FSePH introduces an orthonormal basis to regress the targeted hash codes of training examples to their corresponding reasonably relaxed class labels, featuring significantly reducing the quantization error. Meanwhile, an effective optimization algorithm is derived for modality-specific projection function learning and an efficient closed-form solution for hash code learning, which are computationally tractable. Extensive experiments have shown that the proposed FSePH approach runs sufficiently fast, and also significantly improves the retrieval performances over the state-of-the-arts.

In this paper, we proposed a cross-tier cooperation resource allocation (CCRA) scheme over the Ultra-dense heterogeneous network (UDHN) to improve the whole system throughput. First, we use the cooperation signal-to-interference-plus-noise ratio (SINR) threshold to classify three types of user equipments (UEs). And we use stochastic geometry to analyze the effect of the cooperation SINR threshold on the number of UEs. We focus on using the dedicated channel to decrease cross-tier interference. In order to enhance the total system throughput, an NP-hard problem is formulated. In order to reduce the computational complexity, we decompose this resource allocation problem into subcarrier allocation and power allocation and we use two algorithms to solve it. The simulation results show that the system throughput is significantly improved by using the proposed CCRA scheme in a dense environment.

With the high operating temperature in power electronics, intermetallic compounds (IMCs) formed between solders and substrates can rapidly grow and lead to integration failure, which requires a robust diffusion barrier at elevated temperature. Electroless Ni-Fe-P coating can be a good alternative due to its excellent electrical conductivity, thermal stability and diffusion barrier property. In this work, preparation of ternary Ni-Fe-P alloy and its quality were reported. Mechanical properties of Ni-Fe-P at elevated temperature were evaluated by nanoindentation tests. It was found that Er and H generally followed logarithmic relation with loading rates at room temperature and elevated temperature.

Behavior recognition from large available motion capture data has received wide attention in the computer animation community and is growing increasingly important in recent years. In this paper, we present an efficient motion capture behavior recognition approach via neighborhood preserving dictionary learning. First, we normalize all the motion sequences in the database to make the motion to be comparable. Then, the neighborhood preserving property is exploited using Iterative Nearest Neighbors algorithm and subsequently added as a constraint condition for discriminative dictionary learning, whereby the raw motion frame can be represented as a compact set of atoms consisting of neighborhood preserving characteristics. Finally, the recognition result can be efficiently obtained by sparse coding based classification scheme. Extensive experiments tested on publicly available motion capture databases have demonstrated the accuracy and effectiveness of the proposed approach.

Housing reverse mortgages, it is a new way of pension as the worship. It can solve the aging problem of those who have house but have no money. Through the related check about Chongqing's house price, we learned that the price trend of Chongqing is stable, and then we predict the future house price. Through empirical research of annuity payments, we can get the information that those elderly who involved in reverse mortgage can get each month. It can increase participation of elderly disposable income and establish guidance of consumption which can be expected and effectively improve the quality of life of elderly. At last, put forward recommendations of reverse mortgages in Chongqing.

In this paper, we formulate the Motion capture (MoCap) data denoising problem as the concatenation of piecewise motion matrix recovery problem, in which the moving trajectories of each piecewise motion always share the similar subspace representation. To this end, we present an automatic MoCap data denoising approach based on the filtered local subspace affinity (LSA) and low rank approximation. The proposed approach does not need any physical information about the underling structure of MoCap data or require auxiliary data sets for the training priors. The experiments have shown the promising results.

As for the burst traffic in DTN (Delay-tolerant Networks), an adaptive message ferry routing algorithm - ADMF (Adaptive Message Ferry) is proposed. Normal node with highest power is elected as elector-ferry, combining with ferry to store-carry-forward messages. ADMF algorithm can dynamically chose single-ferry mechanism and multiple-ferry mechanism according to variable network traffic. Simulation model is created based on The ONE software, and the conclusion is: when the traffic is higher, ADMF algorithm can improve network throughput, while reduce delay and packet loss ratio.

CMOS device degradations such as negative bias temperature instability (NBTI) and hot carrier injection (HCI) have been extensively investigated. However, the relationship between the device degradation and the circuit reliability is not well established. In this paper, we propose a model for the frequency degradation of the ring oscillator (RO) after static stress based on the device degradations of NBTI. The model is demonstrated by using the SMIC 65nm technology. We found that the frequency degradation is much significant for the RO consisting of short (60 nm) channel devices as compared to that of long (130 nm) channel devices. The difference is essentially due to the HCI which dominates the degradation of the RO consisting of short channel devices. Combine the proposed model and the frequency degradation of the RO under dynamic stress, the contribution of NBTI and HCI is distinguished.

Delay Tolerant Networks (DTNs) are occasionally-connected networks. They have high latency, long queuing time, limited resources and intermittent connectivity, which are different from traditional networks. Message Ferry (MF) scheme provides efficient data delivery for DTNs. In the MF scheme, nodes move proactively to send and receive messages. In this paper, we propose a ferry routing scheme for DTNs called Multi-Constraint Ferry Routing (MCFR). By making full use of the traffic between nodes, our routing scheme can improve network performance. Simulation results show that our proposal is better than the MFR in network performance because the MCFR has a higher delivery ratio and a lower end-to-end delay.

Improvement of no-load voltage waveforms and reduce the temperature of rotor are important problems for optimizing quality requirement of electric power and ensuring the operation safety of generator and power system. A 15 MW tubular hydro-generator is proposed, the no-load voltage waveform of which is poor, heat of rotor is severe. The multi-slice moving electromagnetic field-circuit coupling model of the hydro-generator and 3D temperature field FE model of the rotor are built up, the no-load voltage waveform and rotor heat are analyzed. The original design and 2 different optimized design schemes are compared. The results show that, slot skew of stator can optimize the no-load voltage, however the temperature of rotor is reduced rarely. Besides the above measure, increase the damper bar pitch, airgap, diameter of damper bar, and decrease the damper bar assembling airgap, the no-load voltage waveform can be optimized, and minimize the loss and temperature of rotor parts, such as damper bars, iron core, and exciting windings. The research is helpful for improve the design standard and enhance the operation reliability of the large hydro-generator and electric network.

In this paper, we present an efficient active contour with a joint distribution metric for semi-supervised natural image segmentation. Firstly, we project an RGB image into two-dimensional subspace and draw a polygon curve around the Region of Interest (ROI) as the initial evolving curve. Then, we model the regional statistics in terms of joint probability distributions and propose an effective distribution metric to regularize the active contours for evolution. Subsequently, we convert the resultant zero level set function into binary pattern and find all the 8-connected regions. Finally, the largest region is selected as the desired ROI and smoothed with a circular averaging filter so that the corresponding final segmentation result can be obtained. Meanwhile, the proposed approach also features fast convergence and easy implementation in comparison with the traditional methods, which need a laborious process of re-initializing the zero level set in terms of a sign distance function (SDF) periodically. The experiments show the promising results.

In this paper, we present the efficient region-based active contours with a novel distribution metric for complex object segmentation problems. Unlike most conventional approaches, we model the regional statistics using probability distribution function and propose a simple but effective distribution metric to drive the active contours. Subsequently, the proposed approach speeds up the segmentation process without initializing the zero level set in terms of a sign distance function (SDF) and re-initializing it periodically during the evolution as used in the traditional methods. Some challenging synthetic and real-world images are utilized to evaluate the proposed segmentation algorithm. The experiments show its promising result in comparison with the existing methods.

Effects of different farming patterns on soil organic carbon distribution in profile and oxidation stability in purple paddy soil of long term experiment site (since 1990) were studied. Results showed that the content of soil organic carbon (SOC) decreased gradually with the increase in depth among different kind of treatments. The SOC content, reaching 20.69 g kg−1, in 0–10 cm soil layer with the treatment of NRR(no-tillage and ridge, rice and reap) was the highest, while with the treatment of RRR(Rotation of waterloging rice and dry reap) the content was the lowest being only around 14.93 g kg−1. The organic carbon density in 0–20 cm and 0–60 cm soil layer with the treatment of NRR were obviously higher than that with CRWF(Conventional tillage, rice and winter fallow) and RRR. It indicated that carrying out long-term NRR may lead to the enrichment of SOC in macroaggregates of the surface soil layer. The readily oxidizable organic carbon (ROOC) content with the treatments of CRWF and NRR decreased gradually along with increasing soil depth. The general trend of RRR showed a slight increase at first and then gradually went down. The content of ROOC of CRWF was evidently higher than those of NRR and RRR in 0–10 cm and 10–20 cm soil layer and the oxidation stability coefficient(Kos) of CRWF was lower than those of NRR, suggesting that the treatment of NRR had positive effect on the stability of soil organic carbon in the cultivable layer. In conclusion, farming patterns have a greater impact on the soil organic carbon accumulation and transformation.

The motion segmentation is to divide the original motion sequence into several motion fragments with specific semantic, which plays an important role in the motion compression, motion classification, motion synthesis. This paper presents a motion segmentation algorithm based on the central distance features and low-pass filter for the human motion capture data. The proposed approach mainly includes three steps. Firstly, a set of central distance features from the center joint ROOT to limbs was extracted, and those features were divided into the upper and lower limbs norms. Then, PCA method was used to get the one dimension principal component, which can better represent the original motion. Furthermore, the low-pass filter is utilized to get the denoising signal. Consequently, the segmental points set can be obtained. Experimental results show the promising performance of our algorithm.

Due to the deficiency in traditional Virtual Simulation of port logistics system which affects the efficiency of the port work, the port logistics system was based on distributed virtual simulation. The paper describes the object model template development which based on the port logistics system, FOM / SOM object class, interaction class and its federal "public" and "ordering" information relationships which between members of the design, and finally realized this system.

Construction of address code is one of the key techniques in optical code division multiple access (OCDMA) system. Aiming at the fact that (F, K, 1) optical orthogonal code (OOC) possesses good performance but capacity is small, and it can't meet the demands of all-optical networks. Although the capacity of (F, K, 2) OOC is significantly increased compared with (F, K, 1) OOC, system synchronization will be weaker than (F, K, 1) OOC. So, (F, K, 1, 2) OOC possesses big capacity and holds good synchronization. In this paper, the method of constructing big capacity (F, K, 1, 2) OOC based on block design is discussed. The algorithm of construction of OOC from block design is presented. The auto-correlation and cross-correlation restricts are set strictly. The algorithm is simulated; several groups of (F, K, 1, 2) OOC are gained. The results show that the algorithm has good astringency and simplicity. It can construct (F, K, 1, 2) OOC effectively. It is feasible.

A simple model of layered structural geological bodies already have more mature solutions, while the complex structure modeling and interpolation algorithm of three-dimensional space are becoming the key difficulties. Discussion in the paper not only realize the special geological interpolation processing, also makes the local Kriging interpolation in a variety of geological body model of the scope of application in practice which uses the drawing tool OpenGL.

Based on the investigation and analysis of oilfields in different regional area of China, the character of micro-ecosystem including biotic and abiotic factors and its general influence to bioremediation of oil hydrocarbon polluted soils was investigated. And then, the microbial properties, especially the restriction by water and oil content were discussed and the bioremediation feasibility was revealed in this paper. Results showed that the high oil content in soils would suppress the microbial action and community. On the other hand, the column chromatography and GC-MS revealed that more than 50% pollutants were bioavailable and low eco-toxic to the microbes. Generally, constant nutrient in oil polluted soils was lower than normal level, while trace nutrient metal were plentiful for microbial growth and metabolism. There was certain amount and active microbe in the oil contaminated soils, which applied the functional subject for bioremediation. Total microorganisms were almost as widely distributed from 107 to 109cell/g dry soil, while oil degrading microbe 105-106 cell/g dry soil. Simultaneously, the microbial density and activity were restricted by water and oil content. Generally, the high water quantity (>6%) and optimal oil pollutant could stimulate the proliferation, and promote the bioremediation efficiency.

Memory nonlinearities exist in electronic devices or transmission links. A blind identification method is proposed to estimate the Volterra inverse of memory nonlinearities for multiband system by minimizing the strong out-of-band spurs above the preset threshold based on domain transformation and spur classification. The inverse model is employed to recover the noisy sampled signal or improve the spurious-free dynamic range of memory nonlinear system by applying post-distortion or pre-distortion compensation. It can also used as multiband digital filter with varying intermediate frequency or bandwidth. In time-varying system, the filter coefficients can be updated adaptively by iterate search. The experiment results demonstrate the effectiveness.

The spectral purity of input signal is extremely important for exciting the device-under-test in high-speed broadband applications. To improve spurious-free dynamic range of signal generator, a dynamic nonlinear pre-compensation method based on the Volterra pre-distortion filter is developed to compensate memory nonlinearities of the signal generator. The filter coefficients are iteratively identified and updated to track the changes of the system’s time-varying characteristics by employing an adaptive Volterra kernel identification method. With the increase of spectrum components contained in stimulus signal, the Volterra filter needs the longer memory and the lower order, and vice versa. So the amount of the calculation will keep in an acceptable level by proper setting. The results show good performance, moderate complexity and high availability.

The Volterra series behavioral approach has been largely applied in recent years for modeling of nonlinear dynamic systems. Only recently it was used to mathematically model time-invariable systems such as analog to digital devices (ADCs). In this paper, Volterra series modeling methods for ADC nonlinearity are briefly introduced according to a classification criterion based on the main research trends. Some possible directions in this area are also given.