Why tanh: choosing a sigmoidal function

By B. Kalman, S. Kwasny. 1992

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

Citations in the corpus

Financial Time Series Forecasting with Deep Learning : A Systematic Literature Review: 2005-2019 (2019, background): “Most commonly used nonlinear activation functions are: sigmoid (Equation 2) [40], hyperbolic tangent (Equation 3) [41], Rectified Linear Unit (ReLU) (Equation 4) [42], leaky-ReLU (Equation 5) [43], swish (Equation 6) [44], and softmax (Equation 7) [39].”
Deep Learning for Financial Applications : A Survey (2020, background): “Sigmoid [22], hyperbolic tangent [23], Rectified Linear Unit (ReLU) [24], leaky ReLU [25], swish [26], and softmax[27] are among the most preferred nonlinear activation functions in the literature.”
Development of a suitable model for characterizing photovoltaic arrays with shaded solar cells (2007, methodology): “All data sets are scaled down to the range { 1; 1} and a hyperbolic tangent sigmoid transfer function is used to be the activation function of the single hidden layer (Kalman and Kwasny, 1992).”
Forecasting daily streamflow using hybrid ANN models (2006, methodology): “However, Kalman and Kwasny (1992) argue that the hyperbolic tangent transfer function should be used, and the empirical results obtained by Maier and Dandy (1998) also indicate that not only is the training with the hyperbolic tangent function faster than the training with the logistic sigmoid…”
Artificial neural network simulation of hourly groundwater levels in a coastal aquifer system of the Venice lagoon (2012, unknown): “Hyperbolic tangents were selected as the activation functions for the hidden layer since they have been found to provide better accuracy with respect to other sigmoids (Kalman and Kwasny, 1992; Maier and Dandy, 1998).”
Neural networks for the prediction and forecasting of water resource variables: a review of modelling issues and applications (2000, result): “Kalman and Kwasny (1992) argue that the hyperbolic tangent transfer function should be used, which is in agreement with the empirical results obtained by Maier and Dandy (1998a).”