transformation objects used internally in GPflowR
This module will likely be unexported and not formally documented in the near future. This is essentially a test run of the module documentation set up
# transform objects
t <- transforms$Identity()
t <- transforms$Exp(lower = 1e-6)
t <- transforms$Log1pe(lower = 1e-6)
t <- transforms$Logistic(a = 0, b = 1)
t <- transforms$positive(lower = 1e-6)
# transform object member functions
t$forward(x)
t$backward(y)
t$tf_forward(x)
t$tf_log_jacobian(x)
The minimum value that positive transforms can take. This helps stability during optimization, because aggressive optimizers can take overly-long steps which lead to zero in the transformed variable, causing an error.
The location (a) and scale (b) of the logistic
distribution used in the logistic transform.