## Uses of Classorg.nd4j.linalg.activations.BaseActivationFunction

• Packages that use BaseActivationFunction
Package Description
org.nd4j.linalg.activations.impl
• ### Uses of BaseActivationFunction in org.nd4j.linalg.activations.impl

Subclasses of BaseActivationFunction in org.nd4j.linalg.activations.impl
Modifier and Type Class and Description
`class ` `ActivationCube`
f(x) = x^3
`class ` `ActivationELU`
f(x) = alpha * (exp(x) - 1.0); x < 0 = x ; x>= 0 alpha defaults to 1, if not specified
`class ` `ActivationGELU`
GELU activation function - Gaussian Error Linear Units
`class ` `ActivationHardSigmoid`
f(x) = min(1, max(0, 0.2*x + 0.5))
`class ` `ActivationHardTanH`
⎧ 1, if x > 1 f(x) = ⎨ -1, if x < -1 ⎩ x, otherwise
`class ` `ActivationIdentity`
f(x) = x
`class ` `ActivationLReLU`
Leaky RELU f(x) = max(0, x) + alpha * min(0, x) alpha defaults to 0.01
`class ` `ActivationMish`
https://arxiv.org/ftp/arxiv/papers/1908/1908.08681.pdf
`class ` `ActivationPReLU`
/** Parametrized Rectified Linear Unit (PReLU) f(x) = alpha * x for x < 0, f(x) = x for x >= 0 alpha has the same shape as x and is a learned parameter.
`class ` `ActivationRationalTanh`
Rational tanh approximation From https://arxiv.org/pdf/1508.01292v3 f(x) = 1.7159 * tanh(2x/3) where tanh is approximated as follows, tanh(y) ~ sgn(y) * { 1 - 1/(1+|y|+y^2+1.41645*y^4)} Underlying implementation is in native code
`class ` `ActivationRectifiedTanh`
Rectified tanh Essentially max(0, tanh(x)) Underlying implementation is in native code
`class ` `ActivationReLU`
f(x) = max(0, x)
`class ` `ActivationReLU6`
f(x) = min(max(input, cutoff), 6)
`class ` `ActivationRReLU`
f(x) = max(0,x) + alpha * min(0, x) alpha is drawn from uniform(l,u) during training and is set to l+u/2 during test l and u default to 1/8 and 1/3 respectively Empirical Evaluation of Rectified Activations in Convolutional Network
`class ` `ActivationSELU`
https://arxiv.org/pdf/1706.02515.pdf
`class ` `ActivationSigmoid`
f(x) = 1 / (1 + exp(-x))
`class ` `ActivationSoftmax`
f_i(x) = exp(x_i - shift) / sum_j exp(x_j - shift) where shift = max_i(x_i)
`class ` `ActivationSoftPlus`
f(x) = log(1+e^x)
`class ` `ActivationSoftSign`
f_i(x) = x_i / (1+|x_i|)
`class ` `ActivationSwish`
f(x) = x * sigmoid(x)
`class ` `ActivationTanH`
f(x) = (exp(x) - exp(-x)) / (exp(x) + exp(-x))
`class ` `ActivationThresholdedReLU`
Thresholded RELU f(x) = x for x > theta, f(x) = 0 otherwise. theta defaults to 1.0