Developing Novel Activation Functions Based Deep Learning LSTM for Classification

This study proposes novel Long Short-Term Memory (LSTM)-based classifiers through developing the internal structure of LSTM neural networks using 26 state activation functions as alternatives to traditional hyperbolic tangent (tanh) function. The have high performance in solving vanishing gradient problem that is observed recurrent networks. Performance investigations were carried out utilizing three distinct deep learning optimization algorithms evaluate efficiency proposed functions-based for two different classification tasks. simulation results demonstrate use Modified Elliott, Softsign, Sech, Gaussian, Bitanh1, Bitanh2 and Wave trump tanh-based terms accuracy. are encouraged be utilized tested other