Search

Currently, learning Bayesian Networks (BNs) from data has become a much attention-getting issue in fields of machine learning and data mining. While there exists few efficient algorithms for learning BNs in presence of incomplete data. In this paper, we present a scoring function based on mutual information for evaluating BN structures. To decrease computational complexity, we introduce MRMR criterion into the scoring function, which enables the computation of the scoring function to involve in only two-dimensional mutual information. When the dataset is incomplete, we use EMI method to estimate the Mutual Information (MI) from the incomplete dataset. As for whether a node ordering is manually given or not, we develop two versions of algorithms, named as MRMR-E1 and MRMR-E2 respectively and evaluate them through experiments. The experimental results on Alarm network show good accuracy and efficiency of our algorithms. [ABSTRACT FROM AUTHOR]

Although encouraging results have been reported, existing Bayesian network (BN) learning algorithms have some troubles on limited data. A statistical or information theoretical measure or a score function may be unreliable on limited datasets, which affects learning accuracy. To alleviate the above problem, we propose a novel BN learning algorithm MRMRG, Max Relevance and Min Redundancy Greedy algorithm. MRMRG algorithm applies Max Relevance and Min Redundancy feature selection technique and proposes Local Bayesian Increment (LBI) function according to the Bayesian Information Criterion (BIC) formula and the likelihood property of overfitting. Experimental results show that MRMRG algorithm has much better accuracy than most of existing BN learning algorithms when learning BNs from limited datasets. [ABSTRACT FROM AUTHOR]

Existing algorithms for learning Bayesian network (BN) require a lot of computation on high dimensional itemsets, which affects accuracy especially on limited datasets and takes up a large amount of time. To alleviate the above problem, we propose a novel BN learning algorithm OMRMRG, Ordering-based Max Relevance and Min Redundancy Greedy algorithm. OMRMRG presents an ordering-based greedy search method with a greedy pruning procedure, applies Max-Relevance and Min-Redundancy feature selection method, and proposes Local Bayesian Increment function according to Bayesian Information Criterion (BIC) formula and the likelihood property of overfitting. Experimental results show that OMRMRG algorithm has much better efficiency and accuracy than most of existing BN learning algorithms on limited datasets. [ABSTRACT FROM AUTHOR]

Bayesian networks (BNs) have been widely used for learning model structures of a domain in the area of data mining and knowledge discovery. This paper incorporates ensemble learning into BN structure learning algorithms and presents a novel ensemble BN structure learning approach. Based on the Markov condition and the faithfulness condition of BN structure learning, our ensemble approach proposes a novel sample decomposition technique and a components integration technique. The experimental results reveal that our ensemble BN structure learning approach can achieve an improved result compared with individual BN structure learning approach in terms of accuracy. [ABSTRACT FROM AUTHOR]

Existing algorithms for learning Bayesian network (BN) require a lot of computation on high dimensional itemsets, which affects accuracy especially on limited datasets and takes up a large amount of time. To alleviate the above problem, we propose a novel BN learning algorithm MRMRG, Max Relevance and Min Redundancy Greedy algorithm. MRMRG algorithm is a variant of K2 algorithm for learning BNs from limited datasets. MRMRG algorithm applies Max Relevance and Min Redundancy feature selection technique and proposes Local Bayesian Increment (LBI) function according to the Bayesian Information Criterion (BIC) formula and the likelihood property of overfitting. Experimental results show that MRMRG algorithm has much better efficiency and accuracy than most of existing BN learning algorithms when learning BNs from limited datasets. [ABSTRACT FROM AUTHOR]

This paper presents an improved particle swarm optimizer (IPSO) for solving truss structure optimization problems. The algorithm is based on the particle swarm optimizer with passive congregation (PSOPC) and a harmony search (HS) scheme. It handles the problem-specified constraints using a ‘fly-back mechanism' method and the variables' constraints using the harmony search scheme. The IPSO is tested on a planar truss structure optimization problem and is compared with the PSO and the PSOPC algorithm respectively. The result shows that the IPSO method presented in this paper is able to accelerate the convergence rate effectively and has the fastest convergence rate among these three other algorithms. [ABSTRACT FROM AUTHOR]

Receptor-mediated endocytosis can be exploited to achieve efficient cell-specific gene delivery. Our laboratory has used two approaches for targeted gene delivery. One uses polycation as a carrier for plasmid DNA and the other uses peptide nucleic acid (PNA) as a carrier. Targeted gene delivery using polycation carriers has been widely utilized with some success. This approach mainly suffers from large particle size and nonspecific interaction with blood components. These drawbacks have limited the use of this type of vector for in vivo applications. Using PNA as a carrier, on the other hand, allows for smaller particle size and less nonspecific interactions. The stability of this vector in the circulation may be a limiting factor. In addition, both types of vector lack mechanisms for endosome escape and nuclear transport. In this chapter, current developments and uses for targeted gene delivery of each approach (polycation vs. PNA) are reviewed. [ABSTRACT FROM AUTHOR]

This paper extends the idea of weighted distance functions to kernels and support vector machines. Here, we focus on applications that rely on sliding a window over a sequence of string data. For this type of problems it is argued that a symbolic, context-based representation of the data should be preferred over a continuous, real format as this is a much more intuitive setting for working with (weighted) distance functions. It is shown how a weighted string distance can be decomposed and subsequently used in different kernel functions and how these kernel functions correspond to real kernels between the continuous, real representations of the symbolic, context-based representations of the vectors. [ABSTRACT FROM AUTHOR]

Availability of a wide variety of Web services over the Internet offers opportunities of providing new value added services by composing existing ones. Service composition poses a number of challenges. A composite service can be very complex in structure, containing many temporal and data-flow dependencies between the component services. It is highly desirable therefore to be able to validate that a given composite service is well formed: proving that it will not deadlock or livelock and that it respects the sequencing constraints of the constituent services. In this paper, we propose an approach of composition analysis and verification based on Colored Petri nets (CP-nets), which is an extended version of Petri Nets, which have a sound mathematical semantics and a number of existing analysis tools. We provide translation rules of web composition language into CP-nets and a technique to analyze and verify effectively the net to investigate several behavioral properties. Our translation technique is essentially independent of which language we describe composition. As an example, to show the effectiveness of our technique, we pick up BPEL and translate the BPEL specification into CP-nets in a constructive way. These nets are analyzed and verified as prototypes of the specification. [ABSTRACT FROM AUTHOR]

In this paper we first investigate the problem of finding a delay- and degree-bounded maximum sum of nodes ALMT. We proved the problem is NP-hard, and testify its relationship with the well-studied degree-bounded minimum maximum-delay ALMT problem. We proposed a greedy algorithm FindOverlayTree with SRPF heuristic. The simulation results show that our design solution is a valid and cost-effective approach where compared to other three modified classic heuristics under several performance metrics. [ABSTRACT FROM AUTHOR]