In this paper a Modular Granular Neural Network (MGNN) optimization is proposed, where a grey wolf optimizer is proposed to design MGNNs architectures. The design of these architectures consists in to seek number of sub modules, number of hidden layers for each sub module with their respective number of neurons, learning method, error goal and percentage of data used for the training phase. This model is based on the percentage of data (in this work are images) used for the training phase to perform a selection of which are the optimal images to be used for that phase. The proposed method was applied to pattern recognition based on the iris biometrics.
展开▼
