Shuffled frog-leaping algorithm based neural network and its using in big data set

In this paper, the stochastic optimization algorithm shuffled frog-leaping algorithm (SFLA) is investigated. Big data challenge requires an efficient optimization algorithm to explore the potential data structures with deep neural network. First, the neural network classifier is introduced and compared with support vector machine. Neural network is suitable for large data set and it has complex ability to extract high level abstraction data. Second, the big data set is introduced and covering cancer data and speech data. Both data sets have a large number of samples with complex low level variance. Third, the neural network parameter is optimized using a modified version of SFLA. SFLA is efficient and robust to local minimums. Experimental results indicate that the proposed algorithm has good future application on big data set classification. The modified SFLA can effectively optimize the neural network parameters.