Soft Max Activation Function For Neural Network Multi Class Classifiers

This paper investigates the effectiveness of the Softmax function used as activation function in Neural networks for multi class classification problems. An Elman Neural Network is used with Softmax activation function. The Thyroid Diagnosis problem and Glass type identification problem are considered as multi class datasets which are obtained from UCI machine learning laboratories. The experimental results prove that the Softmax classifiers are more efficient and have higher accuracy than the other methods available in literature.