Chain

class pmrf.stats.bijectors.Chain(bijectors: Sequence[AbstractBijector])

Bases: AbstractFwdLogDetJacBijector, AbstractInvLogDetJacBijector

Composition of a sequence of bijectors into a single bijector.

Bijectors are composable: if f and g are bijectors, then g o f is also a bijector. Given a sequence of bijectors [f1, …, fN], this class implements the bijector defined by fN o … o f1.

!!! warning “Bijectors are applied in reverse order”

Given a sequence [f, g], the Chain bijector computes f(g(x)), not g(f(x)). This matches the mathematical convention for function composition but may be counterintuitive when building layers sequentially.

!!! example

`python layers = [] layers.append(f) layers.append(g) bijector = distreqx.Chain(layers) y = bijector.forward(x)  # y = f(g(x)) `

Initializes a Chain bijector.

Arguments:

• bijectors: a sequence of bijectors to be composed into one. Each bijector

  can be a distreqx bijector or a callable to be wrapped by Lambda. The sequence must contain at least one bijector.

forward(x: Array) → Array

Computes y = f(x).

forward_and_log_det(x: Array) → tuple[Array, Array]

Computes y = f(x) and log|det J(f)(x)|.

inverse(y: Array) → Array

Computes x = f^{-1}(y).

inverse_and_log_det(y: Array) → tuple[Array, Array]

Computes x = f^{-1}(y) and log|det J(f^{-1})(y)|.

same_as(other: AbstractBijector) → bool

Returns True if this bijector is guaranteed to be the same as other.

bijectors: list[AbstractBijector]