TransportMaps.jl

Appearance

Rectifier Functions ​

Index ​

Types ​

TransportMaps.Softplus Type
julia
Softplus(β::Float64=1.0)

Smooth rectifier function: g(ξ) = log(1 + exp(βξ)) / β. Ensures monotonicity with smooth approximation to ReLU. Higher β values create sharper transitions.

source
TransportMaps.ShiftedELU Type
julia
ShiftedELU()

Rectifier combining exponential and linear behavior: g(ξ) = exp(ξ) for ξ ≤ 0, g(ξ) = ξ + 1 for ξ > 0. Ensures monotonicity with different behavior for negative and positive inputs.

source
TransportMaps.IdentityRectifier Type
julia
IdentityRectifier()

No-op rectifier: g(ξ) = ξ. Does not ensure monotonicity! Only use when partial derivatives are guaranteed to be positive by other means.

source
TransportMaps.ExpRectifier Type
julia
ExpRectifier()

Exponential rectifier: g(ξ) = exp(ξ). Ensures strict positivity and monotonicity. Can lead to extreme values for large |ξ|.

source