Collective classification in networked data has become an important and active research topic, it has a wide variety of real world applications, such as hyperlinked document classification, protein interaction and gen...
详细信息
In this paper, we introduce a new tractable subclass of cardinal direction relations we called strong saturated-convex rectangle cardinal direction relations. We prove that reasoning in this subclass is a polynomial t...
详细信息
In this paper, we introduce a new tractable subclass of cardinal direction relations we called strong saturated-convex rectangle cardinal direction relations. We prove that reasoning in this subclass is a polynomial time problem and show that the path-consistency method is sufficient for deciding consistency.
The current GPM algorithm needs many iterations to get good process models with high fitness which makes the GPM algorithm usually time-consuming and sometimes the result can not be accepted. To mine higher quality mo...
详细信息
Image retrieval has been an active research topic due to its great importance in the field of image management, digital library and web searching. However, the performances of the existing approaches, including text-b...
详细信息
Collective classification, which is represented to classify unobserved nodes simultaneously in networked data is becoming an important research area with applications in several domains, such as the classification of ...
详细信息
Collective classification, which is represented to classify unobserved nodes simultaneously in networked data is becoming an important research area with applications in several domains, such as the classification of documents, image processing. Most algorithms are based on the hypothesis that nearby nodes tend to have the same label. However, there are many networks that do not necessarily satisfy this hypothesis. In this paper, we present a new method based on random walk and link pattern of the network. It adopts the pseudoinverse laplacian matrix of the graph as similarity measure to identify nearby nodes and assigns an initial label for each unlabeled node, then iteratively update the label of unlabeled nodes based on the link pattern. The experimental results on two real world datasets demonstrate that the proposed method outperforms the other state-of-art approaches for this problem.
Community mining has been the focus of many recent researches on dynamic social networks. In this paper, we propose a clustering based improved ant colony algorithm (CIACA) for community mining in social networks. The...
详细信息
Community mining has been the focus of many recent researches on dynamic social networks. In this paper, we propose a clustering based improved ant colony algorithm (CIACA) for community mining in social networks. The CIACA combines the local pheromone update rule with the global update rule and utilizes heuristic function to adjust the clustering solution dynamically, assisted by decay coefficient of dynamic network model. In order to improve clustering accuracy and convergence rate in the process of ant migration, a structure tightness between nodes based clustering centers initializing method is proposed, which can provide us initial clustering centers with certain clustering precision and high diversity. In addition, random number and specific parameter are used in the ant transition probability, which strengthens the search stochastic properties of CIACA effectively. The proposed CIACA is tested on some benchmark social networks, and is compared with current representative algorithms in community mining. Experimental results show the feasibility and validity of CIACA.
This paper presents a novel representation and matching method for deformable shapes. The proposed approach finds most expressive segments of a deformable shape category called similar and discriminative parts, which ...
详细信息
The reconfigurable design problem is to find the element that will result in a sector pattern main beam with side lobes. The same excitation amplitudes apply to the array with zero-phase that should be in a high direc...
详细信息
The reconfigurable design problem is to find the element that will result in a sector pattern main beam with side lobes. The same excitation amplitudes apply to the array with zero-phase that should be in a high directivity, low side lobe pencil shaped main beam. Multi-beam antenna arrays have important applications in communications and radar. This paper presents a new method of designing a reconfigurable antenna array with quantized phase excitations using a new evolutionary algorithm called differential evolution (DE). In order to reduce the effect of mutual coupling among the antenna-array elements, the dynamic range ratio is minimized. Additionally, compared with the continuous realization and subsequent quantization, experimental results indicate better performance of the discrete realization of the phase-excitation value of the proposed algorithm.
Recently evolutionary search has been investigated by lots of researchers as method for structure learning of Bayesian network. In order to avoid the problem of premature convergence and enhance the learning accuracy,...
详细信息
Recently evolutionary search has been investigated by lots of researchers as method for structure learning of Bayesian network. In order to avoid the problem of premature convergence and enhance the learning accuracy, in this paper we integrated evolutionary way of bees and proposed triple-population evolution strategies to learn Bayesian network structure. During evolution, the best Bayesian network individual is considered as queen bee, and there are three populations: one stores and evolves elitist individuals, another stores random individuals to guarantee the diversity of individuals, some individuals from that two populations form the third population in which every individual is considered as drone and mated with queen bee to generate offspring. The proposed method has both exploitative and explorative ability. Experimental results show that the proposed method can converge to more accurate Bayesian network than standard genetic algorithm based learning method.
Bayesian network is a popular tool for uncertainty process in Artificial Intelligence. In recent years, more and more attention has been paid to learning of Bayesian network. In this paper, we proposed a novel learnin...
详细信息
Bayesian network is a popular tool for uncertainty process in Artificial Intelligence. In recent years, more and more attention has been paid to learning of Bayesian network. In this paper, we proposed a novel learning algorithm for Bayesian network based on (μ, λ)-Evolution Strategy, we present the encoding scheme and fitness function, designed the evolutionary operators of recombination, mutation and selection. Theoretical analysis and experimental results all demonstrate that the proposed method can learn the Bayesian network from data effectively.
暂无评论