Optimal parameter selection for unsupervised neural network using genetic algorithm
This work addresses parameter tuning for a specific neural network, which is an incremental improvement in clustering methods.
The paper tackles the problem of selecting optimal parameters for the K-FLANN unsupervised neural network by using a genetic algorithm, resulting in efficient parameter optimization from a large search space as tested on artificial and synthetic datasets.
K-means Fast Learning Artificial Neural Network (K-FLANN) is an unsupervised neural network requires two parameters: tolerance and vigilance. Best Clustering results are feasible only by finest parameters specified to the neural network. Selecting optimal values for these parameters is a major problem. To solve this issue, Genetic Algorithm (GA) is used to determine optimal parameters of K-FLANN for finding groups in multidimensional data. K-FLANN is a simple topological network, in which output nodes grows dynamically during the clustering process on receiving input patterns. Original K-FLANN is enhanced to select winner unit out of the matched nodes so that stable clusters are formed with in a less number of epochs. The experimental results show that the GA is efficient in finding optimal values of parameters from the large search space and is tested using artificial and synthetic data sets.