Applied Mechanics and Materials Vols. 411-414

Abstract: To tackle the QoS-based service selection problem, an efficient ant colony service selection algorithm called CASS is proposed in this paper. In this algorithm, a skyline query process is used to filtering the candidates related each service class and a clustering based shrinking process is used to guide the ant search directions. We evaluate our approach experimentally using standard real datasets and synthetically generated datasets, and compared with the recently proposed related service selection algorithms. It reveals very encouraging results in terms of the quality of solution, and the processing time required.

Abstract: This paper studied the leader how to motivate the follower to maximize its interests for ill-posed linear bilevel programming problem. We first presented a coordination model and gived its corresponding penalty problem. Under some conditions, we established the result on the existence of the solution. Then, an algorithm was developed to obtain a coordination solution to the original bilevel programming problem. Finally, numerical results show that the proposed method is feasible.

Abstract: In human social life, it is often need to make comprehensive evaluation for person, thing , or a project to carry on the classification or evaluation. Analytic hierarchy process is relatively common and the most simple evaluation model. It draw up a series of evaluation index according to the evaluation object. The index may contain multiple child index, according to the relationship between the indexes or artificial factors to determine tie index weight, and get the overall evaluation for objects. The artificial factor is too much, it can not objectively reflect the real situation of the evaluation object. This paper proposed a evaluation method based on neural networks, based on the indexed of the analytic hierarchy process, through using the expert evaluation samples, by the BP neural network to study, so as to get the objective weight of the indexes, and then reflect the real situation of the evaluation objects.

Abstract: Among all improved BP neural network algorithms, the one improved by heuristic approach is studied in this paper. Firstly, three types of improved heuristic algorithms of BP neural network are programmed in the environment of MATLAB7.0. Then network training and simulation test are conducted taking a nonlinear function as an example. The approximation performances of BP neural networks improved by different numerical optimization approaches are compared to aid the selection of proper numerical optimization approach.

Abstract: An improved evolutionary algorithm (SCAGA) is proposed in this paper. The algorithm is based on new population initialization method and genetic operator. SCAGA adopts the crossover probability and mutation probability that vary with the increase of evolution generation in order to control genetic operations in an effective range. Meanwhile, SCAGA presents a new crossover strategy that restricts the cross of the chromosomes to some extent to protect good genes schema. The schema theorem is employed in the algorithm to analyze the working mechanism of SCAGA. According to experiment results for test functions and TSP problems, SCAGA is effective.

Abstract: Timetabling problem is important for efficient allocations of resources. University timetabling is to allocate a weekly schedule for instructors and students under a number of requirements and constraints. This paper purposes a solution based on a co-operative co-evolution multi-objective genetic algorithm (CCMOGA). The idea is to split the complexity into several simpler parts and co-operatively evolve them. Two definitions of chromosomes are studied: instructor-based and student-based. Experimental results with an actual situation show that the proposed method is able to give effective solutions, and the chromosome definition based on students yields higher quality schedules than does the definition based on instructors.

Abstract: This paper considers the parallel-machine scheduling with preference of machines so as to minimizing total tardiness. An immunoglobulin-based artificial immune (IAI) algorithm is constructed for searching the best production sequence of aluminum foil factory. The IAI algorithm has a systematic immune mechanism which mainly is built on somatic hypermutation and recombination methods. There are three categories of immunoglobulin in somatic hypermutation for constructing antibody proliferation mechanism to obtain superior antibodies. To avoid falling into local optimal solutions, receptor editing mechanism suppresses the worse solutions to accelerate convergence. And reverse mechanism is developed in somatic recombination to generate some new antibodies. The computational results show that the performance of IAI is significant improvement to the EDD based heuristic. Therefore, the proposed IAI algorithm is competitive for the parallel-machine scheduling problem.

Abstract: Rule acquisition is a hot topic in the field of data mining. And the inconsistent information systems are widespread nowadays. However, rules acquisition methods are always the difficulty of rough set theory application in inconsistent decision information systems; So the paper proposes a new rule acquisition method. Firstly, we use maximum distribution reduction method for knowledge reduction in single decision-making inconsistent information system and then we use decision-making resolution matrix and decision-making matrix function to get the decision rules. Finally, we mine the rules from inconsistent decision-making information systems.

Abstract: The GA(Genetic Algorithm) is used in this paper to solve the evaluation problem in the checkers game. The evaluation parameter is chosen based on six critical facts, such as the number of black pieces and the number of red pieces. The evaluation function is linear with different weights. Using the weights as the chromosome, along with the local hill-climbing method in GA and the mathematical statistical method to choose child generation, the algorithm realizes evolution automatically. With sufficient evolution times, new chromosome appears, leading to the formation of an optimized evaluation function. Based on the experimental data analysis, the algorithm could enhance the power of checkers program effectively.

Abstract: To solve the training problems for complex problems, a new learning algorithm based on continued fractional weight function for training neural networks is proposed, which is different from that of polynomial weight functions. The continued fractional weight functions using interpolation methods are more suitable for training the patterns obtained by some rational problems. The analysis of generalization is also presented in this paper. At last, to illustrate the power of the new learning algorithm, a simulation example is presented to show that the new algorithm proposed in this paper has good performance both on generalization and calculating precision.