Review on Methods of Selecting Number of Hidden Nodes in Artificial Neural Network?

Artificial Neural Networks are most effective and appropriate for pattern recognition and many other real world problems like signal processing, Classification problems. Superior results in pattern recognition can be directly provided in the forecasting, classification and data analysis. To bring proper results, ANN requires correct data preprocessing, architecture selection and network training but still the performance of a neural network depends on the size of network. Selection of hidden neurons in neural network is one of the major problems in the field of Artificial Neural Network. The random selections of hidden neurons may cause the problem of either Underfitting or Overfitting. Overfitting arises because the network matches the data so closely as to lose its generalization ability over the test data. The proposed method finds the near to optimal number of hidden nodes after training the ANN from real world data. The advantage of proposed method is that it is not approximately calculating number of hidden nodes but based on similarity between input data, they are calculated.