Neural network

enum nnp::NeuralNetwork::ActivationFunction

List of available activation function types.

Values:

enumerator AF_UNSET: Unset activation function.

enumerator AF_IDENTITY: \(f_a(x) = x\)

enumerator AF_TANH: \(f_a(x) = \tanh(x)\)

enumerator AF_LOGISTIC: \(f_a(x) = 1 / (1 + \mathrm{e}^{-x})\)

enumerator AF_SOFTPLUS: \(f_a(x) = \ln (1 + \mathrm{e}^x)\)

enumerator AF_RELU: \(f_a(x) = \max(0, x)\) (NOT recommended for HDNNPs!)

enumerator AF_GAUSSIAN: \(f_a(x) = \mathrm{e}^{-x^2 / 2}\)

enumerator AF_COS: \(f_a(x) = \cos (x)\)

enumerator AF_REVLOGISTIC: \(f_a(x) = 1 - 1 / (1 + \mathrm{e}^{-x})\)

enumerator AF_EXP: \(f_a(x) = \mathrm{e}^{-x}\)

enumerator AF_HARMONIC: \(f_a(x) = x^2\)