In this paper, the concept of adaptive multiobjective genetic algorithm (AMOGA) is applied for the structure optimization of radial basis function network (RBFN). The problem of finding the number of hidden centers remains a critical issue in the design of RBFN. The number of basis function controls the complexity and generalization ability of the network. The most parsimonious network obtained from the pareto front is applied in one of the challenging research area of proteomics and computational biology: Protein superfamily classification. The problem deals with predicting the family membership of a newly discovered amino acid sequence. The modification to the earlier approach of multiobjective genetic algorithm (MOGA) is done based on the two key controlling parameters such as probability of crossover and probability of mutation. These values are adaptively varied based on the performance of the algorithm i.e., based on the percentage of total population present in the best nondomination level. Principal component analysis (PCA) is used for dimension reduction and significant features are extracted from long feature vector of amino acid sequences. Numerical simulation results illustrates the efficiency of our approach in terms of faster convergence, optimal architecture and high level of classification accuracy.
Design of Cu-Zn-Al-Sc Oxide Catalyst for Methanol Synthesis Using Genetic Algorithm Based on Radial Basis Function Network as the Evaluation Function
