softmax {multinma} | R Documentation |
Softmax transform
Description
The softmax transform is a multivariate generalisation of the logit
transform. softmax()
maps a vector of K-1
values on the real line to a
K
-simplex (i.e. values between 0 and 1, that sum to 1). inv_softmax()
provides the inverse transform, mapping a K
-simplex vector to a vector of
K-1
real values.
Usage
softmax(x)
inv_softmax(p)
Arguments
x |
|
p |
|
Value
softmax()
returns a vector of length K
that is a simplex.
inv_softmax()
returns a vector of reals of length K-1
.
Examples
x <- c(-1, 3, -0.5, 2)
(p <- softmax(x))
sum(p)
inv_softmax(p)
[Package multinma version 0.8.1 Index]