Intelligent Computing Methodologies

For the problem of leak inductance energy and voltage spike, given in the fly-back photovoltaic converter system (hereinafter abbreviated as PV), a new physical circuit of RCD (resistor-capacitor-diode) is proposed which based on the Boost circuit topology to recovering leak inductance energy and depressing main switch voltage spike during the commutation of power MOSFET. The circuit can recycle the leak inductance energy effectively [1−2], and suppress output capacitance (hereinafter abbreviated as Coss) electric discharge which caused the voltage spike during MOSFET switch-on. In addition, after detailed analysis, the new physical circuit can solve the problem of the energy loss generate from the Coss electric discharge which is not pay more attention by the other papers. The new design can clamp the drain-source voltage and implement the control policy of soft-switch. In this paper, we give the theoretical derivation of the circuit parameter and the implemented physically experimental curve of the physical circuit by analyzing the causes of voltage spike generated. Compared with the traditional method, improved the converter efficiency from 94.2 % to 96.2 %, and the MPPT efficiency to 99.4 %.

Identity-based batch verification (IBBV) scheme is very desirable to solve efficiency, security and privacy preservation issues for vehicular ad hoc network (VANET). In 2015, Tzeng et al. proposed an IBBV scheme which was published in IEEE Transaction on Vehicular Technology. Their scheme has superior performance than other existing similar schemes in terms of security, computation cost and transmission overhead by performance evaluations. However, one time signature verification of their scheme needs two bilinear pairing operations. As it is well known, bilinear pairing is one of the most time-consuming operation in modern cryptography. Therefore, some efforts can be made to prevent the appearance of pairing and obtain better efficiency. In this paper, we propose an improved scheme of Tzeng et al.’s IBBV. Our improved IBBV scheme needs not use bilinear pairing without the lack of security and privacy-preserving. The total computation cost for signing and verifying is the constant 1.2 ms for single message and n messages respectively, which is far better performance than Tzeng et al. scheme and other similar schemes. So our improved IBBV scheme is more suitable for practical use. Finally, we apply the recovering technology of the vehicle’s real identity of Tzeng et al.’s IBBV scheme to a public key authentication scheme for mobile Ad-hoc networks to address an improved pairing-free authentication scheme.

An increasing number of studies have found that one of the most important factors for emergence and development of complex diseases is the interactions between SNPs, that is to say, epistasis or epistatic interactions. Though many efforts have been made for the detection of SNP-SNP interactions, the algorithm of such studies is still ongoing due to the computational and statistical complexities. In this work, we proposed an algorithm IACO based on ant colony optimization and a novel introduced fitness function Svalue, which combined both Bayesian networks and mutual information, for detecting SNP-SNP interactions. Furthermore, a memory based strategy is also employed to improve the performance of IACO, which effectively avoids ignoring the optimal solutions that have already been identified. Experiments of IACO are performed on both simulation data sets and a real data set of age-related macular degeneration (AMD). Results show that IACO is promising in detecting SNP-SNP interactions, and might be an alternative to existing methods for inferring epistatic interactions. The software package is available online at http://www.bdmb-web.cn/index.php?m=content&c=index&a=show&catid=37&id=98.

Separating singing voice from music accompaniment for monaural recordings is very useful in many applications, such as lyrics recognition and singer identification. Based on non-negative matrix partial co-factorization (NMPCF), we propose an improved algorithm which restricts the activation coefficients of singing voice components to be temporal continuous and sparse in each frame. Temporal continuity is favored by using a cost term which is the sum of squared difference between the activation coefficients in adjacent frames, and sparsity is favored by penalizing nonzero values for each frame. For the separated singing voice, we quantify the performance of the system by the signal-to-noise ratio (SNR) gain and the accuracy of singer identification. The experiments show that the constraints of temporal continuity and sparsity criteria both can improve the performance of singing voice separation, especially the constraint of temporal continuity.

The purpose of supervised learning with temporal encoding for spiking neurons is to make the neurons emit arbitrary spike trains in response to given synaptic inputs. Recent years, the supervised learning algorithms based on synaptic plasticity have developed rapidly. As one of the most efficient supervised learning algorithms, the remote supervised method (ReSuMe) uses the conventional pair-based spike-timing-dependent plasticity rule, which depends on the precise timing of presynaptic and postsynaptic spikes. In this paper, using the triplet-based spike-timing-dependent plasticity, which is a powerful synaptic plasticity rule and acts beyond the classical rule, a novel supervised learning algorithm, dubbed T-ReSuMe, is presented to improve ReSuMe’s performance. The proposed algorithm is successfully applied to various spike trains learning tasks, in which the desired spike trains are encoded by Poisson process. The experimental results show that T-ReSuMe has higher learning accuracy and fewer iteration epoches than the traditional ReSuMe, so it is effective for solving complex spatio-temporal pattern learning problems.

A nation’s Critical Infrastructures (CI) is vital to the trustworthy functioning of the economic, health care, and social sectors of the nation. Any disruption to CI will adversely affect the economy, and peaceful functioning of the government. Above all, it will adversely affect the morale and confidence of the citizens. Hence, protecting CI of a nation must be given top priority. Fundamental to protection mechanisms are risk and vulnerability analysis. Based on their outcomes protection mechanisms can be planned, designed, and implemented. In this paper we offer a concise template representation for critical assets, and explain a methodology for vulnerability assessment and risk analysis. We point out the potential role of agents, and deep learning methods in the development and commissioning of future cyber defense solutions.

Gene expression profiles which represent the state of a cell at a molecular level could likely be important in the progress of classification platforms and proficient cancer diagnoses. In this paper, we attempt to apply imperialist competition algorithm (ICA) with parallel computation and faster convergence speed to select the least number of informative genes. However, ICA same as the other evolutionary algorithms is easy to fall into local optimum. In order to avoid the defect, we propose an improved binary ICA (IBICA) with the idea that the local best city (imperialist) in an empire is reset to the zero position when its fitness value does not change after five consecutive iterations. Then IBICA is empirically applied to a suite of well-known benchmark gene expression datasets. Experimental results show that the classification accuracy and the number of selected genes are superior to other previous related works.

This paper proposes and develops a novel multi-objective optimization scheme called MORSGO based on iterative adaptation of plant root growth behaviors. In MORSGO, the basic local and global search operators are designed deliberately based on auxin-regulated tropism of the natural root system, including branching, regrowing of different types of roots. The fast non-dominated sorting approach is employed to get priority of non-dominated solutions obtained during the search process, and the diversity over archived individuals is maintained by using dynamical crowded distance estimation strategy. Accordingly, Pareto-optimal solutions obtained by MORSGO have merits of better diversity and lower computation cost. The proposed MORSGO is evaluated on a set of bio-objective and tri-objective test functions taken from the ZDT benchmarks in terms of two commonly used metrics IGD and SPREAD, and it is compared with NSGA-II and MOEA/D. Test results verify the superiority and effectiveness of the proposed algorithm.

Inspired by the group discussion behavior of students in class, a new group topology is designed and incorporated into original particle swarm optimization (PSO). And thus, a novel modified PSO, called group discussion mechanism based particle swarm optimization (GDPSO), is proposed. Using a group discussion mechanism, GDPSO divides a swarm into several groups for local search, in which some smaller teams with a dynamic change topology are included. Particles with the best fitness value in each group will be selected to learn from each other for global search. To evaluate the performance of GDPSO, four benchmark functions are selected as test functions. In the simulation studies, the performance of GDPSO is compared with some variants of PSOs, including the standard PSO (SPSO), PSO-Ring and PSO-Square. The results confirm the effectiveness of GDPSO in some of the benchmarks.

Cancer genome projects can provide different types of data on the genetic level, which is significant for cancer research and biological processes in computational methods. Thus, computational methods used to identify cancer subtypes should fully focus on integrating these multidimensional data (e.g., DNA methylation data, mRNA expression data, etc.). Sparse reduced-rank regression (Srrr) method, a state-of-the-art multiple response linear regression method, can easily deal with high dimensional statistical data. In this paper, we introduced Srrr method combining iCluster (Srrr-cluster) to discovery cancer subtypes. Firstly, we used Srrr to estimate the coefficient matrix and then cancer subtypes were clustered by iCluster. Finally, we used our Srrr-cluster method to analyze glioblastoma and breast cancer data. The results show that our Srrr-cluster method is effective for cancer subtype identification.

Thyroid disease is extremely common and of concern because of the risk of malignancies and hyper-function and they may become malignant if not diagnosed at the right time. Ultrasound is one of the most often used methods for thyroid nodule detection. However, node detection is very difficult in ultrasound images due to their flaming nature and low quality. In this paper, an algorithm for the formalization of the contour of the nodule using the variance reduction statistic is proposed where cut points are determined, then a method of selecting the nearest neighbor points which form the shape of the nodule is generated, later B-spline method is applied to improve the accuracy of the curve shape. The extracted results are been compared with graph_cut and watershed methods for efficiency. Experiments show that the algorithm can improve the accuracy of the appearance of modality and maximum significance of data in the images is also protected.

Feature selection is to remove redundant and irrelevant features from original ones of exemplars, so that a sparse and representative feature subset can be detected for building a more efficient and accurate classifier. This paper presents a novel definition for the discernibility and independence scores of a feature, and then constructs a two dimensional (2D) space with the feature’s independence as y-axis and discernibility as x-axis to rank features’ importance. This new method is named FSDI (Feature Selection based on Discernibility and Independence of a feature). The discernibility score of a feature is to measure the distinguishability of the feature to detect instances from different classes. The independence score is to measure the redundancy of a feature. All features are plotted in the 2D space according to their discernibility and independence coordinates. The area of the rectangular corresponding to a feature’s discernibility and independence in the 2D space is used as a criterion to rank the importance of the features. Top-k features with much higher importance than the rest ones are selected to form the sparse and representative feature subset for building an efficient and accurate classifier. Experimental results on 5 classical gene expression datasets demonstrate that our proposed FSDI algorithm can select the gene subset efficiently and has the best performance in classification. Our method provides a good solution to the bottleneck issues related to the high time complexity of the existing gene subset selection algorithms.

Sparse representation (SR) can effectively represent structure features of images and has been used in image processing field. A new palmprint image classification method by using multiple kernel sparse representation (MKSR) is proposed in this paper. Kernel sparse representation (KSR) behaves good robust and occlusion like as sparse representation (SR) methods. Especially, KSR behaves better classification property than common sparse representation methods and used widely in pattern recognition task. In KSR based classification methods, the selection of a kernel function and its parameters is very important. Usually, the kernel selected is not the most suitable and can not contain complete information. Therefore, MKSR methods are developed currently and used widely in image classification task. Here, multiple kernel functions select the weighted of Gauss kernel and polynomial kernel. In test, all palmprint images are selected from PolyU palmprint database. The palm classification task is implemented by the extreme learning machine (ELM) classifier. Compared with methods of SR and single kernel based SR, experimental results show that our method proposed has better calcification performance.

For video communication, Region of Interest (ROI) coding offers an efficient way of improving the quality of videos. In this paper, we propose a novel rate control scheme for Region of Interest coding. We set the area covering people are interested in as the Region of Interest (ROI) to preserve better quality than the background. In addition, we set a coefficient ω to evaluate the significance of ROI, and then use it to calculate the mean absolute distortion (MAD) of the ROI and the background. Finally, the quantization parameter (QP) can be decided by the quadratic model. The difference of QP can be used to distinguish the background and the ROI with the QP which is obtained in advance, so the ROI mask cannot be coded and transmitted. Moreover, through adjusting the coefficient ω to control the degree of the protection of the ROI, our method can protect the ROI effectively and control the difference between the background and the ROI expediently. Experimental results show that the scheme works well in protecting the ROI, the objective quality of the ROI will be improved and the bit-rate can be controlled in a certain range.

In order to improve the accuracy and real-timelines of heart rate variability (HRV) estimation in electrocardiogram (ECG) signals interferences, a novel HRV estimation method based on photoplethysmography (PPG) signals is proposed. The short-time autocorrelation principle is used to detect interferences in ECG signals, then, the improving sliding window iterative Discrete Fourier Transform (DFT) is used to estimate HRV in ECG interferences from the synchronously acquisitioned PPG signals. The international commonly used MIT-BIT Arrhythmia Database/Challenge 2014 Training Set is used to verify the interferences detection algorithm and HRV estimation algorithm which are proposed. At the same time, the proposed algorithms are compared with recently existing representative interferences detection algorithm based on RR intervals and PRV directly replaced HRV algorithm, respectively. The results show that the proposed methods are more accurate and more real-time.

Amyotrophic lateral sclerosis (ALS) is a fatal neurodegenerative disease that involves the degeneration and death of the nerve cells in brain and spinal cord that control voluntary muscle movement. This disease can cause patients struggling with a progressive loss of motor function while typically leaving cognitive functions intact. This paper presents a novel predication method that combines a dimension reduction (integrating partial least square into random frog algorithm) with support vector regression to predict the progression of ALS in the next 3–12 months according to the data collected from the patients over the latest three months. The experiment on the actual data from the PRO-ACT database indicates that the proposed method is effective and robust and can predict the clinical outcome by means of the slope of ALS progression, as measured using the ALS functional rating scale (ALSFRS) and the score used for monitoring ALS patients. Especially, the features selected can effectively distinguish the clinical outcome targets. It is of great benefit to aid clinical care, identify new disease predictors and potentially significantly reduce the costs of future ALS clinical trials.

Resorting to extraction text techniques for Chinese heritage documents becomes an increasing need. Historic documents such as Chinese calligraphy usually were handwritten or scanned in low contrast so that an automatic optical character recognition procedure for document images analysis is difficult to apply. In this paper, we present a historic document image threshold based on a combination of Bradley’s algorithm and K-means. An adaptive K-means cluster as a pre-processing methods for document image has been used for automatically grouping the pixels of a document image into different homogeneous regions. In Bradley’s methods, every image’s pixel is set to black if its brightness is T percent lower than the average brightness of surrounding pixels in the window of the specified size, otherwise it is set to white. Finally, text bounding boxes are generated by concatenating neighboring word clusters with mathematical morphology method. Experimental results show that this algorithm is robust in dealing with non-uniform illuminated, low contrast historic document images in terms of both accuracy and efficiency.

A weakly calibrated monocular and line structured light based pavement transverse profile measurement method is proposed. Combining the linearization approximation of the laser triangulation, the weakly correlation of the road roughness national standard and the system calibration parameters was deduced. The problem of complicate system calibration in existed methods was transferred into a single factor calibration process. A second-order Gaussian filter was designed to enhance the laser curve and suppress the disturbance of lane markings efficiently utilizing the unique tubular feature of the laser curve intensity distribution. The results of the virtual camera simulation and real image experiments validated the robustness of proposed method and the error can be limited in the 10−4 order of magnitudes.

This work proposes an innovative technique to solve the problem of tracking and following a generic human target by a drone in a natural, possibly dark scene. The algorithm does not rely on color information but mainly on shape information, using the HOG classifier, and on local brightness information, using the optical flow algorithm. We tried to keep the algorithm as light as possible, envisioning its future application on embedded or mobile devices. After several tests, performed modeling the system as a set of SISO feedback-controlled systems and calculating the Integral Squared Error as quality indicator, we noticed that the final performance, overall satisfactory, degrades as the background complexity and the presence of disturbance sources, such as sharp edges and moving objects that cross the target, increase .

Traditional least-squares temporal difference (LSTD) algorithms provide an efficient way for policy evaluation, but their performance is greatly influenced by the manual selection of state features and their approximation ability is often limited. To overcome these problems, we propose a multikernel recursive LSTD algorithm in this paper. Different from the previous kernel-based LSTD algorithms, the proposed algorithm uses Bellman operator along with projection operator, and constructs the sparse dictionary online. To avoid caching all history samples and reduce the computational cost, it uses the sliding-window technique. To avoid overfitting and reduce the bias caused by the sliding window, it also considers $$ L_{2} $$ regularization. In particular, to improve the approximation ability, it uses the multikernel technique, which may be the first time to be used for value-function prediction. Experimental results on a 50-state chain problem show the good performance of the proposed algorithm in terms of convergence speed and prediction accuracy.

Hyperspectral remote sensing has drawn great interests in earth observation, since contiguous spectrum can provide rich information of ground objects. However, such numerous bands also pose great challenges to efficient processing of hyperspectral image (HSI). Manifold learning, as a promising tool for nonlinear dimensionality reduction, has been widely used in feature extraction of HSI data to find meaningful representations of original spectrum. Nevertheless, lack of explicit and nonlinear mapping is still a limitation. In the paper, we propose a novel manifold learning based method for the classification of ground targets in HSI data, named as Polynomial Laplacian Embedding (PLAE). We first encode spatial information into spectral signal to obtain a fused representation. Then we model the local geometry of high-dimensional HSI data with graph Laplacian, and we introduce a nonlinear and explicit polynomial mapping to find a compact low-dimensional feature space, in which efficient classification of ground targets can be achieved. Experiments conducted on benchmark data sets demonstrate that high classification accuracy can be obtained by using the features extracted by PLAE.

Autoencoder has been successfully used as an unsupervised learning framework to learn some useful representations in deep learning tasks. Based on it, a wide variety of regularization techniques have been proposed such as early stopping, weight decay and contraction. This paper presents a new training principle for autoencoder based on denoising autoencoder and dropout training method. We extend denoising autoencoder by both partial corruption of the input pattern and adding noise to its hidden units. This kind of noisy autoencoder can be stacked to initialize deep learning architectures. Moreover, we show that in the full noisy network the activations of hidden units are sparser. Furthermore, the method significantly improves learning accuracy when conducting classification experiments on benchmark data sets.

Bayesian Learning Automata (BLA) are demonstrated to be as efficient as the state-of-the-art automaton in two action environments, and it has parameter-free property. However, BLA need the explicit computation of a beta inequality, which is time-consuming, to judge its convergence.In this paper, the running time of BLA is concerned and two approaches are proposed to accelerate the computation of the beta inequality. One takes advantage of recurrence relation of the beta inequality, the other uses a normal distributions to approximate the beta distributions. Numeric simulation are performed to verify the effectiveness and efficiency of those two approaches. The results shows these two approaches reduce the running time substantially.

We propose in this paper a modification of one of the modern state-of-the-art genetic programming algorithms used for data-driven modeling, namely the Bi-objective Genetic Programming (BioGP). The original method is based on a concurrent minimization of both the training error and complexity of multiple candidate models encoded as Genetic Programming trees. Also, BioGP is empowered by a predator-prey co-evolutionary model where virtual predators are used to suppress solutions (preys) characterized by a poor trade-off error vs complexity. In this work, we incorporate in the original BioGP an adaptive mechanism that automatically tunes the mutation rate, based on a characterization of the current population (in terms of entropy) and on the information that can be extracted from it. We show through numerical experiments on two different datasets from the energy domain that the proposed method, named BioAGP (where “A” stands for “Adaptive”), performs better than the original BioGP, allowing the search to maintain a good diversity level in the population, without affecting the convergence rate.

Based on the missing not at random assumption and central limit theorem, this paper presents a novel way to accelerate the iteration speed in the collaborative filtering models called Gaussian iteration. In the proposed model, adding the Gaussian distribution to the estimation error makes the falling direction more credible, which significantly reduces the running time with the ideal accuracy. For evaluation, we compare the performance of the proposed model with three existing collaborative filtering models on two kinds of Movielens datasets. The results indicate that the novel method outperforms the existing models and it is easy to implement and faster. Moreover, the proposed model is scalable to the analogous objective function in other models.

Mixture models have been used as efficient techniques in the application of image segmentation. In order to segment images automatically without knowing the number of true image components, the framework of Dirichlet process mixture model (DPMM, also known as the infinite mixture model) has been introduced into conventional mixture models. In this paper, we propose a novel approach for image segmentation by considering the truncated Dirichlet Process of Student’s t-mixture model (tDPSMM). We also develop a novel Expectation Maximization (EM) algorithm for parameter estimation in our model. The proposed model is tested on the application of images segmentation with both brain MR images and natural images. According to the experimental results, our method can segment images effectively and automatically by comparing it with other state-of-the-art image segmentation methods based on mixture models.

In this paper, the mixture of Gaussian processes (MGP) is applied to model and predict the time series of stock prices. Methodically, the precise hard-cut expectation maximization (EM) algorithm for MGPs is utilized to learn the parameters of the MGP model from stock prices data. It is demonstrated by the experiments that the MGP model with the precise hard-cut EM algorithm can be successfully applied to the prediction of stock prices, and outperforms the typical regression models and algorithms.

The mixture of Gaussian process functional regressions (mix-GPFR) is a powerful tool for curve clustering and prediction. Unfortunately, there generally exist a large number of local maximums for the Q-function of the conventional EM algorithm so that the conventional EM algorithm is often trapped in the local maximum. In order to overcome this problem, we propose a deterministic annealing EM (DAEM) algorithm for mix-GPFR in this paper. The experimental results on the simulated and electrical load datasets demonstrate that the DAEM algorithm outperforms the conventional EM algorithm on parameter estimation, curve clustering and prediction.

This paper analyzes the basic structure of Cloud Computing Data Center. Then, the reliability block diagram and the Monte Carlo simulation method are used to analyze the availability of Cloud Computing Data center. Meanwhile, suggestions on improving the availability of Cloud Computing Data Center are proposed. This method can effectively complete the simulation analysis of the availability and operation and maintenance cost of the Cloud Computing Data Center. It also can provide the theoretical basis for the operation and maintenance of the Cloud Computing Data Center.

A fast rate of convergence means that a given fixed amount of data can achieve a high level of prediction performance. In a fully asynchronous system, it is considered that the performance of the learning model could be reduced, because of the collision in parallel gradient computations. In this work, we proposed an algorithm called GreedyAC to reduce the collision. We present the detailed design procedure of GreedyAC, and indicate that using GreedyAC can reduce the regret bound of asynchronous adaptive gradient (AsyncAdaGrad) with mini-batch training when the data is sparse. The experimental results show that the proposed algorithm improves the performance significantly as the number of workers (work nodes) increases, and we found that the results are roughly consistent with our theoretical results.

The current mining algorithm only consider the user’s access frequency, neglecting the interest of users in their visiting path. Compared to the current algorithms for mining user browsing preferred path, clustering algorithm combines the Jacques ratio coefficient and the longest public path coefficient multiplication. This proposed method can estimate the user similarity of page interest and website access structures matrix more accurately for the element value based on the “three tuple” model. Adopting an improved mining algorithm for preference and interest calculation, the bad impact of mining is removed due to pages idle and links. The experimental results showed that the algorithm had higher efficiency and accuracy in web log mining of big data.

A novel clustering method based on spectral clustering theory and spectral cut standard is proposed via analyzing the characteristics of short text and the defects of the existing clustering algorithms. First of all, a weighted undirected graph is created according to spectral clustering theory, similarity between node and node is calculated on graph, and a symmetrical documents similarity matrix is constructed, which provides all information for the clustering algorithm. Inspired by Greedy strategy, we utilize prim to develop PrimMAE algorithm for the purpose of partitioning graph into two parts, in which RMcut is termination condition of partitioning process, and then it is fed into CASC algorithm to cut the documents set iteratively. Ultimately, high quality clustering results demonstrate the effectiveness of the new clustering algorithm.

The most important issue of big data processing is the relevance of analytical data; thought of this paper is to analyze the data as a graph optimal partitioning problem. Computing all circuit graphics firstly, calculated frequent map and redrawing of the system structure according to the results, the core problem is the time complexity of the algorithm. To solve this problem, researching DEMIX algorithm in non-strongly connected graph and study on relationship between frequent node and adjacency matrix which is strongly connected branches. Gives the corresponding examples, and analyzes the algorithm complexity. On the time complexity of the proposed method DEMIX is retrieving effect faster, more accurate search results.

Currently, a few alarm correlation algorithms are based on a framework involving frequency and support-confidence. These algorithms often fail to address text data in alarm records and cannot handle high-dimensional data. This paper proposes an algorithm based on the similarity distance and deep networks. The algorithm first translates text data in alarms to real number vectors; second, it reconstructs the input, obtains the alarm features through a deep network system and performs dimension reduction; and finally, it presents the alarm distribution visually and helps the administrator determine the new fault. Experimental results demonstrate that it cannot only mine the correlation among alarms but also determine the new fault quickly by comparing the graphs of the alarm distribution.

Reordering model is the crucial component in statistical machine translation (SMT), since it plays an important role in the generation of fluent translation results. However, the data sparseness is the key factor that greatly affects the performance of reordering model in SMT. In this paper, we exploit synonymous information to alleviate the data sparseness and take Chinese-Mongolian SMT as example. First, a synonym-based reordering model with Chinese synonym is proposed for Chinese-Mongolian SMT. Then, we flexibly integrate synonym-based reordering model into baseline SMT as additional feature functions. Besides, we present source-side reordering as the pre-processing module to verify the extensibility of our synonym-based reordering model. Experiments on the Chinese-Mongolian dataset show that our synonym-based reordering model achieves significant improvement over baseline SMT system.

Question classification plays a key role in question answering systems as the classification result will be useful for effectively locating correct answers. This paper addresses the problem of question classification by syntactic structure. To this end, questions are converted into parsed trees and each corresponding parsed tree is represented as a multi-dimensional sequence (MDS). Under this transformation from questions to MDSs, a new similarity measurement for comparing questions with MDS representations is presented. The new measurement, based on the all common subsequences, is proved to be a kernel, and can be computed in quadratic time. Experiments with kNN and SVM classifiers show that the proposed method is competitive in terms of classification accuracy and efficiency.

Social micro-blogging platforms facilitate the emergence of citizen’s needs and desires which reflect a variety of intents ranging from daily life (e.g., food and drink) to leisure life (e.g., travel and physical exercise). Identifying user intents in microblog and distinguishing different types of intents are significant. In this paper, we propose a novel approach to classify user intents into three categories, namely Travel, Food&Drink and Physical Exercise. Our method exploits Wikipedia concepts as the intent representation space, thus, each intent category is represented as a set of Wikipedia concepts. The user intents can be identified through mapping the microblogs into the Wikipedia representation space. Moreover, we develop a Collaborative User Model, which exploits the user’s social connections to obtain a comprehensive account of user intents. The quantitative evaluations are conducted in comparison with state-of-the-art baselines, and the experimental results show that our method outperforms baselines in each intent category.

Entity relation extraction is a key task in information extraction. The purpose is to find out the semantic relation between entities in the text. An improved tree kernel-based method for relation extraction described in this paper adds the predicate verb information associated with entity, prunes the original parse tree, and removes some redundant structure on the basis of the Path-enclosed Tree. The experiment shows that the proposed method delivered better performance than existing methods.

The robotic impedance control is currently one of the main control methods, its main characteristic is that it can make manipulators move to the appointed position quickly and accurately. Due to the high complexity of the robot system, to adjust the impedance controller parameter is always difficult. The impedance controller multi-objective optimization design method is proposed, taking dynamic performances as the optimization objectives, a multi-objective optimization algorithm based on Pareto optimality is applied to the optimal design, obtain Pareto optimal solutions, and get some initial impedance controller adjustment rules, the satisfactory solution is selected in Pareto-optimal solutions according to the requirements of the present system. Simulation results indicate the effectiveness of the proposed algorithm.

This paper presents a high-frequency strategy based on Deep Neural Networks (DNNs). The DNN was trained on current time (hour and minute), and $$ n $$-lagged one-minute pseudo-returns, price standard deviations and trend indicators in order to forecast the next one-minute average price. The DNN predictions are used to build a high-frequency trading strategy that buys (sells) when the next predicted average price is above (below) the last closing price. The data used for training and testing are the AAPL tick-by-tick transactions from September to November of 2008. The best-found DNN has a 66 % of directional accuracy. This strategy yields an 81 % successful trades during testing period.

In this paper, an adaptive extended computed torque controller is proposed for trajectory tracking of 3 degree-of-freedom planar parallel manipulators. The dynamic model, including the modeling errors and uncertainties, is established in the joint space of 3 degree-of-freedom planar parallel manipulators. Based on the dynamic model, an adaptive extended computed torque control scheme is proposed in which a feed-forward neural network is combined with error compensators for adaptive compensating the unknown modeling errors and uncertainties of the parallel manipulators. The weights of the neural network are based on sliding functions and self-tuned online during the tracking control of system without any offline training phase. Using the combination of Sim Mechanics and Solid works, the comparative simulations are conducted for verifying the efficiency of the proposed control scheme.

Financial trend prediction is an interesting but also challenging research topic. In this paper, we exploit Twitter moods to boost next-day financial trend prediction performance based on deep network models. First, we summarize six-dimensional society moods from Twitter posts based on the profile of mood states Bipolar lexicon expanded by WordNet. Then, we combine Twitter moods and financial index by Deep Network models, and propose two methods. On the one hand, we utilize a Deep Neural Network of good fitting capability to evaluate and select predictive Twitter moods; On the other hand, we use a Convolutional Neural Network to explore temporal patterns of financial data and Twitter moods through convolution and pooling operations. Extensive experiments over real datasets are carried out to validate the performance of our methods. The results show that Twitter mood can improve prediction performance under the deep network models, and the Convolutional Neural Network based method performs best on most cases.

Research interest in robotic wheelchairs is driven in part by their potential for improving the independence and quality-of-life of persons with disabilities and the elderly. However the large majority of research to date has focused on indoor operations. In this paper, we aim to develop a smart wheelchair robot system that moves independently in outdoor terrain smoothly. To achive this, we propose a robotic wheelchair system that is able to classify the type of outdoor terrain according to their roughness for the comfort of the user and also control the wheelchair movements to avoid drop-off and watery areas on the road. An artificial neural network based classifier is constructed to classify the patterns and features extracted from the Laser Range Sensor (LRS) intensity and distance data. The overall classification accuracy is 97.24 % using extracted features from the intensity and distance data. These classification results can in turn be used to control the motor of the smart wheelchair.

Reasonable interest distribution and risk-sharing problem are known as the biggest obstacles for manufacturing enterprises and the third party logistics enterprises to choose whether to collaborate and evaluate the operation effect of the existing cooperative system. Firstly, based on the deeply exploration of the interest and risk relationship in the cluster environment, decision model of synergy profit distribution and risk sharing based on the Shapley value has been put forward; Secondly, this paper explores the 2-d curve fluctuations while the logistics cost is taken as the fixed variable and the value of interest distribution and risk quantification as the relative variable, and suggests the breakthrough point and the range of synergistic selection to help enterprises in decision-making; Finally, the case study has verified the fluctuation rule.

The necessities of the MNs are generally included in the cellular modules available in LTE/3G and Wi-Fi for high-speed Internet. Further, many people with Wi-Fi and LTE/3G use a cross-layer-based handoff as they move from one place to another. MN mobility management has been handled adequately; however, using the network-based mobility management described in this paper, carriers can manage and maintain a lower-cost network.

In this paper, we review the current status of PMIPv6 (Proxy Mobile IPv6) multicast listener support being standardized in the IETF and point out limitations of the current approach and we proposed a fast multicast handover procedure in inter-domain PMIPv6 network for network-based mobility management. The proposed fast multicast handover procedure in inter domain optimizes multicast management by using the context of the MNs. We evaluate the proposed fast multicast handover procedure compared to the one based through the developed analytical models and confirm that introduced fast multicast handover procedure provides reduced service interruption time and total network overhead compared to the one based during handovers.

This paper presents the design and validation of a differential braking (DB) controller for Sport Utility Vehicles (SUVs) using driver-in-the-loop real-time simulations. According to the driver’s logic, the desired vehicle states will be decided, while actual vehicle states deviate desired values, the control system is applied to improve the lateral stability of SUVs. To derive the controller design, driver-in-the-loop real-time simulations are conducted on the UOIT (University of Ontario Institute of Technology) vehicle simulator,the Fishhook maneuver and Double Lane Change test scenarios are simulated to examine the performance of the controller. The driver-in-the-loop real-time simulation results demonstrate the effectiveness of the proposed differential braking controller in the lateral rollover stability and maneuverability improvement of the SUV.

Skyline query processing in road networks has been investigated extensively in recent years. Skyline points for road network applications may be large while the query point may only interest the ones within a certain range. In this paper, we address the issue of efficient evaluation of Continuous Range Skyline Queries (CRSQ) in road networks. Due to the computation of network distance between objects in road networks is expensive and suffers the limitation of memory resources, we propose a novel method named Dynamic Split Points Setting (DSPS) dividing a given path in road networks into several segments. For each segment, we use Network Voronoi Diagrams (NVDs) based technique to calculate the candidate skyline interest points at the starting point of the segment. After that, when the query point moves, we dynamically set the spilt points by DSPS strategy to ensure that when the query point moves within a segment, skyline points remain unchanged and only need to be updated while moving across the split points. Extensive experiments show that our DSPS strategy is efficient compared with previous approaches.

For the sake of easily developing and improving performances of BDS receiver and BDS (Beidou navigation satellite system)/INS (inertial navigation system) integration system, the INS-aided SDR (software defined radio) is investigated. The performance of the SDR with an intermediate frequency signal collector for B1-band with and without INS-aided are analysed and experimentally validated, respectively. From the optimized acquiring scheme-PMF (partial matched filter)-FFT (fast Fourier transform) with INS aiding to the steadily tracking method, all important components of the system are introduced and analysed in detail. Then, based on the principle of minimizing the tracking error, an optimal adaptive bandwidth designing method is adopted. The results show that the presented PMF-FFT based scheme is capable of realizing effective BDS B1 signal acquisition and tracking with the INS information aiding. This SDR will be a useful tool for develop the BDS and others GNSS soft- or hardware receivers and integration system.

A fuzzy neural sliding mode controller (FNNSMC) is proposed for robot manipulators. Sliding mode controller is implemented based on two radial basic function neural networks and a fuzzy system. The first neural network is used to estimate the robot dynamic function. The second neural network combines with a fuzzy system to present the switching control term of sliding mode control. This combination resolves the chattering phenomenon. The stability of proposed controller is proven. Finally, simulation is done on a 2-link serial robot manipulator to verify the effectiveness.

In this paper a method is proposed called data consistency measure computing of compatibility based multi-sensor Kalman fusion system, so that the measured data is obtained from several sensors constructed at a certain position rather than traditional methods of combining only one gyroscope and one accelerometer applied to current self-balancing vehicle, in order to attain a more reliable system for real time online measure and also prolong the lifetime of product. According to the measure computing of compatibility based outcome of measured angle and angular velocity, data is fused to gain the input of a proportional differential (PD) control part, which is adopted to reach the balancing stability of a self-balancing vehicle.

PID controller is widely used in the many industrial process control fields because of its simple algorithm, good robustness and high reliability. On the basis of the summary on the tuning process, the principle and the respective characteristics of Z-N frequency response method, Cohen-Coon method, Astrom-Hagglund method and CHR method, the simulation experiments are carried out on the three typical controlled objects. According to the simulation results, the advantages and disadvantages of four engineering tuning methods used for setting the PID controller parameters are analyzed and compared.

The Hybrid Electric Vehicle (HEV) powertrain, and an efficient power conversion and energy transmission process, are significant factors to reduce conventional fuel consumption and vehicle gas emission. The scale of gas emission of an HEV depends on an efficient design process for the powertrain and an optimal energy management system. Therefore, this paper models an efficient powertrain and energy transmission process for both series and parallel HEVs, which can contribute to the emission reduction process. Different power conversion stages, energy transmission paths, emissions, and a systems response corresponding to the driver’s profile are analyzed systematically. Finally, the emission of the proposed system is compared with the European standard of vehicle gas emission .

To make the automatic person indexing of interview video in the TV news program, this paper proposes the method to detect the overlay name text line among the whole overlay texts in one frame. The proposed method is based on the identification of the beginning frame and the edge using Canny edge detector. The experimental results on Korean television news videos show that the proposed method efficiently detects and localizes the overlaid name text line.

Collaborative filtering recommendation currently serves as a key method when it comes to how to rapidly and effectively make proper personalized recommendation for customers. For the sake of above, it becomes a research hot pot in e-commerce system. By analyzing the defects of existing collaborative filtering algorithm, we proposed an improved collaborative filtering algorithm (ICF). In our construction, we firstly establish “users-items” interest degree matrix, and introduce the mechanism of singularity degree to generate the set of similar users, and then optimizes the method of neighbor set generation by recommend importance mechanism. Finally, our experiment proves that ICF has higher performance of accuracy and coverage than other existing algorithms.

This paper describes methods of improving TV-watching experience using Machine Learning for Linear TV recommendations. There is an overview of existing methods for video content recommendations and an attempt of developing new method that focused only on linear TV recommendations and takes into account all specifics around it. Recommendation system based on this approach was implemented in Russian pay TV provider ZOOM TV, and demonstrated two times churn rate reduction in comparison with same service without recommendation system. Existing methods and new method effectiveness compared with offered approach by analyzing real people content consumption during 1 year.

Probabilistic skyline queries as an aspect of queries on uncertain data have become an important issue. Previous work on uncertainty modeling for probabilistic skyline queries only lies within the data. However, attribute values of uncertain data are influenced by contexts in real applications while uncertainty is also along with contexts. Further, previous work on probabilistic skyline queries only retrieves those points whose skyline probabilities are higher than a given probabilistic threshold. In this paper, we develop a novel probabilistic skyline query on uncertain data over uncertain contexts called UC-PSkyline, where possible world semantics model is utilized to model uncertain contexts. To avoid unnecessary pair-wise dominance tests, we devise an in-memory tree structure ZB*-tree to process UC-PSkyline queries efficiently. We also develop preprocessing and pruning techniques that can efficiently improve performance of UC-PSkyline. Experiments show the effectiveness and efficiency of the proposed techniques on real and synthetic data sets.

In the real word, many multi-objective optimization problems are subject to dynamic changing conditions, which may occur in objectives, constraints and parameters. This paper provides a prediction strategy, called multi-direction prediction strategy (MDP), to enhance the performance of multi-objective evolutionary optimization algorithms in dealing with dynamic environments. Besides, the proposed prediction method makes use of multiple directions determined by several representative individuals. Our experimental results indicate that MDP can well tackle dynamic multi-objective problems.

This paper presents a new model predictive current control (MPCC) method in order to reduce common-mode voltage (CMV) for a three-phase voltage source inveter (VSI). By utilizing twenty-one vectors instead of eight vectors used in the conventional MPCC method,the proposed MPCC method can not only reduce the CMV but also improve load current performance. Simulation results are given to verify the effectiveness of the proposed MPCC method.

OSGi-based cloud platform has the advantages of dynamic update, real-time monitoring. It brings convenience to cloud service. But the research of OSGi-based cloud platform doesn’t get the attention it deserves now and the service diagnosis and fault recovery for the OSGi-based Cloud Platform haven’t been researched systematically, especially in the resource conflicts among services, fault diagnosis and handling. These issues generally require users’ interaction, increase the users’ learning curve, which may turn out to be a waste of time. This paper divides those faults on the OSGi-based cloud platform into several models, and presents an approach to resolve them. Our method can detect service conflicts, handle exceptions, and diagnose service errors. And this method is proved by our experiment that can diagnose faults, repair faults, and send notifications to users, even at running time.

P2P systems provide peers a dynamic and distributed environment to share resource. Only if peers are voluntarily share with each other can system stably exist. However, peers in such systems are selfish and never want to share even with tiny cost. This can lead to serious free-riding problems. Incentive mechanisms based on evolutionary game aim at designing new strategies to distinguish defective peers from cooperative peers and induce them to cooperate more. Nevertheless, the behavior patterns of peers are versatile. Using only one certain strategy to depict peers’ behaviors is incomplete. In this paper, we propose an adaptive strategy which integrates advantages of 3 classic strategies. These 3 strategies form a knowledge base. Each time a peer with this strategy can select one adjusting to system status according to the adaptive function. Through experiments, we find that in structured system, this strategy can not only promote cooperation but also the system performance.

Meteorological factors, the main causes that impact the power load, have become a research focus on load forecasting in recent years. In order to represent the influence of weather factors on the power load comprehensively and succinctly, this paper uses PCA to reduce the dimension of multi-weather factors and get comprehensive variables. Besides, in view of a relatively low dynamic performance of BP network, a model for short-term load forecasting based on Elman network is presented. When adopting the BP algorithm, Elman network has such problems as being apt to fall into local optima, many iterations and low efficiency. To overcome these drawbacks, this paper improves the active function, optimizes its weights and thresholds using MEA, and formulates a MEA-Elman model to forecast the power load. An example of load forecasting is provided, and the results indicate that the proposed method can improve the accuracy and the efficiency.

The runner design uses the design principle of low specific speed Francis turbine’s blade, then hydraulic design, 3D modeling and ICEM meshing were carried on a pump-turbine with given parameters. Taking into account the factors of high strain rate and curvature over the stream surface, RNG k- model was used to solve N-S equations. The finite volume method was used to discrete and SIMPLEC algorithm was used to solve pressure - speed coupling equations. By simulating the model with the software of Fluent and correcting repeatedly, the efficiency of pump-turbine under design condition was predicted. Besides, the changes of the pressure field and velocity field in volute, runner and draft tube were analyzed during different operating conditions. This design method has provided a new reference for pump-turbine runner’s hydraulic design and made up for the lack of empirical data by using CFD technology to give a direct-viewing reflection of hydraulic performances simultaneously, Numerical experiment results show that the optimal efficiency of turbine’s operating condition can reach 91 % and the optimal efficiency of pump’s can reach 82 %.There will be a further improvement if modify the model repeatedly.

Rain and snow are often imaged as brighter streaks, which can not only confuse human vision but degrade efficiency of computer vision algorithm. Rain removal is very important technique in these fields such as video-surveillance and automatic driving. Most existing methods rely on optical flow algorithm to detect rain pixel and estimate motion field. However, it is extremely challenging for them to achieve real-time performance. In this paper, a LIDAR based algorithm is proposed, which is capable of achieving rain pixel robustly and efficiently from motion field. The motion objects (vehicles and human) are identified for separation by LIDAR (Sick LMS200) in this paper. Then rain pixels on moving objects are removed by bilateral filter which can preserve edge information instead of causing blurring artifacts around rain streaks. Experimental results show that our method significantly outperforms the previous methods in removing rain pixel and detecting motion objects from motion field.

Aiming at the problem that contours always are fractured and not precise in the segmentation of multiple visual objects, Distance Regularized Level Set Evolution(DRLSE) is used. Considering the characteristics of background difference image, an adaptive Otsu is proposed to segment difference image. In order to take full advantage of temporal information in videos, frame difference and background difference are introduced into energy function of DRLSE. Firstly, an adaptive Otsu and asymmetric morphological filtering based method are used to obtain better initial contours. Secondly, initial contours are evolved with improved DRLSE that frame difference and background difference are integrated in as priori knowledge, which can avoid that objects contours are evolved to background edge, and reduce over segmentation. The experimental results show that the contours of multiple video targets can be obtained more precisely and rapidly with the method proposed in this paper than with the existing methods.

This paper presents a new method for measuring human height from video frames. Based on coordinate transformation, the positional relationship between the coordinates of a person’s feature points on the image can be transformed into the person’s height with the knowledge of intrinsic parameters and rotation angle of the camera. In our method, the distance between the camera and the target is not a necessary parameter, which, in contrast, can be estimated by our algorithm. From experiments, we conclude that our method can be simply implemented to estimate a person’s height from video frames in a controllable error scale.

As the scene text detection and localization is one of the most important steps in text information extraction system, it had been widely utilized in many computer vision tasks. In this paper, we introduce a new method based on the maximally stable extremal regions (MSERs). First, a coarse-to-fine classier estimates the text probability of the ERs. Then, a pruning algorithm is introduced to filter non-text MSERs. Secondly, a hybrid method is performed to cluster connected components (CCs) as candidate text strings. Finally, a fine design classifier decides the text strings. The experimental results show our method gets a state-of-the-art performance on the ICDAR2005 dataset.

Compared with 3D mesh data, unstructured point cloud data lack adjacency relationship between points, which only contain geometric coordinates and little information. This paper focuses on the research of characteristics of unstructured point cloud detection algorithm. We put forward the multiscale 3D Harris feature point detection algorithm, which uses iteration strategy to select the optimal Harris response value in multiple scales. Compared with the classical 3D Harris feature point detection algorithm for mesh data, our algorithm can fully use the local information of point cloud models to detect feature point on point cloud models. It is very robust to rotation transformation of point clouds and noise.

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.

Target tracking is one of the most basic application in computer vision and it has attracted wide concern in recent years. Until now, to our best knowledge, most research focused on the tracking research with 2D images, including the Tracking-Learning-Detection (TLD), particle filter, Mean-shift algorithm, etc. While with the advanced technology and lower cost of sensors, 3D information can be used for target tracking problems in many researches and the data can be obtained by laser scanner, Kinect sensor and etc. As a new type of data description, depth information can not only obtain the spatial position information of target but also can protect privacy and avoid the influence of illumination changes. In this paper, a depth-sensitive Mean-shift method for tracking is proposed, which use the depth information to estimate the range of people’s movement and improve the tracking efficiency and accuracy effectively. What’s more, it can adjust kernel bandwidth to adapt to the target size according to the distance between target and the depth camera. In the designed system, Kinect2.0 sensor is not only used to get the depth data and track the target but also can be mobilized by steering gear flexibly when tracking. Experimental results show that these improvements make Mean-shift algorithm more robust and accurate for handling illumination problems during tracking and it can achieve the purpose of real-time tracking.

Knowledge graph (KG) completion aims at predicting the unknown links between entities and relations. In this paper, we focus on this task through embedding a KG into a latent space. Existing embedding based approaches such as TransH usually perform the same operation on head and tail entities in a triple. Such way could ignore the different roles of head and tail entities in a relation. To resolve this problem, this paper proposes a novel method for KGs embedding by preforming bi-directional projections on head and tail entities. In this way, the different information of an entity could be elaborately captured when it plays different roles for a relation. The experimental results on multiple benchmark datasets demonstrate that our method significantly outperforms state-of-the-art methods.

Safety performance assessment is the measurement of a petrochemical enterprise’s achievement in safety management. In order to receive a comprehensive and objective evaluation result, it is necessary to consider all evaluation factors and the numerical uncertainty caused by fuzziness when safety performance assessment is conducted. To improve conventional safety performance evaluation, an evaluation index system is established, and the interval number is used in this study by using interval number to quantify scores and calculate the safety level of petrochemical enterprise safety. A case of petrochemical enterprise is used to illustrate the effectiveness of the method and system. This method is applied to the comprehensive evaluation of petrochemical enterprise safety to achieve good results.

This paper studies the problem of knowledge acquisition to classify service systems. We define a set of attributes and characteristics in order to classify the service systems. To state the interactions between attributes and characteristics we propose a non-linear optimization model and an adaptive neuro-fuzzy inference system (ANFIS) approach. We compare both approaches in terms of mean root square error in a data test based in International Standard Industrial Classification. Our results present a better performance of ANFIS approach over a set of data collected about ISIC classification in Colombia industries.

Automatic detection of epileptic seizure plays an important role in the diagnosis of epilepsy for it can obtain invisible information of epileptic electroencephalogram (EEG) signals exactly and reduce the heavy burdens of doctors efficiently. Current automatic detection technologies are almost shallow learning models that are insufficient to learn the complex and non-stationary epileptic EEG signals. Moreover, most of their feature extraction or feature selection methods are supervised and depend on domain-specific expertise. To solve these problems, we proposed a novel framework for the automatic detection of epileptic EEG by using stacked sparse autoencoder (SSAE) with a softmax classifier. The proposed framework firstly learns the sparse and high level representations from the preprocessed data via SSAE, and then send these representations into softmax classifier for training and classification. To verify the performance of this framework, we adopted the epileptic EEG datasets to conduct experiments. The simulation results with an average accuracy of 96 % illustrated the effectiveness of the proposed framework.

Currently when university students are going into job market, it is found a lot of challenges to help the students and also employers to help their efficiency in finding matching degree. However, during the job seeking peak time, for example, an event called “job fair” in China, it is found very challenging for employer to quickly filter potentially qualified applicants since an employer will probably receive huge number of resumes in a very short period. To solve this problem, in this research we proposed a student file based employer oriented job recommendation framework. In this system, a student is firstly modelled by the personal features and also academic features. Afterwards, different similarity mechanism between the fresh students with those recruited in the target employers are designed to help recommend students. Furthermore, a dynamic recruitment size aware strategy is also proposed to further polish the recommendation results. The experimental study on a Chinese university’s real recruitment data has shown its potential in real applications.

In this paper we analyze feasibility conditions for Fuzzy Linear Programming problems (FLP) and a special case where its constraints are composed by fuzzy numbers bounded by crisp numbers. We analyze three cases: strong, weak feasibility, and unfeasible FLPs, where strong feasibility is much more desirable than weak one since it generalizes feasible solutions.

This paper presents a simulation study on Quasi Type-2 fuzzy Markov chains. Its main objective is to identify its stationary behavior using two methods: the Greatest Eigen Fuzzy Set and the Powers of a Fuzzy Matrix based on other reports which shows that most of fuzzy Markov chains does not have an ergodic behavior. To do so, we simulate Quasi Type-2 fuzzy Markov chains at different sizes to collect some interesting statistics.

The term comorbidity refers to the coexistence of multiple diseases or disorders along with a primary disease in a patient. Hence, the prediction of disease comorbidity can identify the comorbid diseases when dealing with a primary disease. Unfortunately, since the records of comorbidity in clinic are far from complete, we can’t get enough knowledge to understand the reason for comorbidity. Though many researches have been done to predict disease comorbidity, the accuracy of these algorithms need to be improved. By investigating the factors underlying disease comorbidity, we found that a number of comorbidities are caused by common modules comprised by proteins. Thus, we here propose a novel algorithm to identify disease comorbidity by integrating different types of datasets ranging from properties to functions of protein. Results on real data of comorbidity display that our algorithm can perform better than previous approaches, and some of our new predictions are reported in literature, which can prove the effectiveness of our algorithm, and help deeply explain the molecular mechanism of disease comorbidity.

HOT regions, short for high occupied target regions, bound by many transcription factors (TFs) are considered to be one of the most intriguing findings of the recent large-scale sequencing studies. Recent researches have reported that HOT regions are enriched with so many biological processes and functions, which are related with promoters, enhancers and fraction of motifs. Hence, there are a lot of studies focused on the discovery of HOT regions with TFs datasets. Unfortunately, because of the limited TFs datasets from next generation sequencing (NGS) technology and huge time consuming, the HOT regions in each cell line of human genome can’t be fully marked. Here, unlike the previous jobs, we have made an identification of HOT regions by means of machine learning algorithms in 14 different human cell-lines with chromatin modification datasets. The outperform results of these cell-lines can prove the effectiveness and precision of our assumption enough. In addition, we have discovered the cell-type specific HOT regions (CSHRs) of each cell line, which is used to elucidate the associations with cell-type specific regulatory functions.

Most quantitative trait loci (QTL) mapping experiments typically collect phenotypic data on single traits. However, Research complex correlated traits may provide more available information. We develop a novel algorithm for multiple traits quantitative trait loci mapping by using Bayesian Variable Selection Regression, or BVSR, that allows a new robust genetic models for different and correlated traits. We develop computationally efficient Markov chain Monte Carlo (MCMC) algorithms for performing joint analysis. Taken together, these factors put a premium on having interpretable measures of confidence for individual covariates being included in the model. We conduct extensive simulation studies to assess the performance of the proposed methods and to compare with the conventional single-trait model and existing multiple-trait model. More generally, we demonstrate that, despite the apparent computational challenges, our proposed new algorithm can provide useful inferences in quantitative trait loci mapping.