Optimisation of the cascade correlation algorithm to solve the two-spiral problem by using CosGauss and Sigmoid activation functions
Sang Wha Lee · International Journal of Intelligent Information and Database Systems · 2014
This paper proposes a method of optimising the cascade correlation algorithm which uses CosGauss (cosine-modulated Gaussian function) and Sigmoid activation functions. In order to prove the effectiveness of the proposed method, this study selects two spirals as a benchmark problem, which is suggested as a method of solving the moving target problem, and applies them to the analysis of the condition of recognising input patterns. For the evaluation of the algorithm functions, each phase shows the two-spiral patterns which are recognised in input space. And graphics are used in this paper to demonstrate the number of hidden neurons to be produced in the process. As a result of the application of the proposed method, it is confirmed in this study that the capability of pattern recognition by the cascade correlation algorithm, which uses the optimisation method, has been remarkably improved.