bijectors

Bijectors for transforming parameters.

These are re-exports from parax, which in turn takes them from the distreqx library, filling in any that the installed version of distreqx does not provide. The goal is to cover the most common applications; for more advanced use-cases, simply use distreqx directly instead.

Classes

AbstractBijector()	Differentiable bijection that knows to compute its Jacobian determinant.
Chain(bijectors)	Composition of a sequence of bijectors into a single bijector.
DiagLinear(diag)	Linear bijector with a diagonal weight matrix.
Exp([_is_constant_jacobian, ...])	Exponential bijector: y = exp(x).
Identity([_is_constant_jacobian, ...])	Identity bijector: y = x.
Inverse(bijector)	Inverted version of a given bijector.
Leafwise(bijectors)	Applies a pytree of bijectors to a pytree of inputs.
Permute(permutation)	Permutation bijector that reorders the elements of a 1D event.
R2ToComplex([_is_constant_jacobian, ...])	Maps a real array of shape (..., 2) to a complex array of shape (...).
ScalarAffine(shift[, scale, log_scale])	An affine bijector that acts elementwise.
Shift(shift)	Bijector that translates its input elementwise.
Sigmoid()	A bijector that computes the logistic sigmoid.
Softplus([_is_constant_jacobian, ...])	Transforms the real line to the positive domain using softplus y = log(1 + exp(x)).
Tanh()	A bijector that computes the hyperbolic tangent.
Transpose(permutation)	A bijector that transposes the dimensions of the input array.
TriangularLinear(matrix[, is_lower])	A linear bijector whose weight matrix is triangular.