Exploring solving strategies for incompatible problem is a challenging task. Although extension strategy generating method (ESGM) can address incompatible problem by expanding reasoning and extension transformations, the solving process often suffers a combination explosion of computational cost. In order to overcome this shortcoming, a new approach to performing extension transformations based on gene expression programming (GEP) is proposed. The method is able to establish superior operations of extension transformations heuristically and iteratively, which avoids the combination explosion effectively. In order to make GEP adapt to such applying requirement, chromosome architecture, decoding mode, individual selection and convergence criteria are restudied. The proposed method is illustrated with the application of ESGM to a self-guided touring route design problem. Numerical results verify that the proposed method helps provide extension strategies efficiently and has a huge potential for more complex incompatible problem solving.

Adopting gene expression programming to generate extension strategies for incompatible problem

In this article we investigate some approaches to the artificial neural networks training with use of hybrid algorithms. Algorithms which are based on the back propagation algorithm and the ant colony algorithm are considered in detail. The article describes the application of the artificial neural network with the authors’ hybrid training algorithm. The preliminary studies have shown that the algorithm improves the efficiency of the problems on standard test databases. The application of the algorithm for practical problems solution in the field of medicine, namely the definition of danger level determination of tuberculosis carriers is described. It was shown that the accuracy of the hybrid algorithm is up to 22% higher than of the classical one.

Eugene V. Kotliarov

Papers

Hybrid neural network for classification problem solving