LogSigmoid module

Applies the element-wise function: $$ \mbox{LogSigmoid}(x) = \log\left(\frac{ 1 }{ 1 + \exp(-x)}\right) $$

Usage

nn_log_sigmoid()

Shape

• Input: \((N, *)\) where * means, any number of additional dimensions

• Output: \((N, *)\), same shape as the input

Examples

if (torch_is_installed()) {
m <- nn_log_sigmoid()
input <- torch_randn(2)
output <- m(input)
}