tfp.experimental.bijectors.inverse_log_det_jacobian_ratio

Computes p.ildj(x, ndims) - q.idlj(y, ndims), numerically stably.

p_kwargs and q_kwargs are passed to the registered ildj_ratio_fn. The fallback implementation passes them to the inverse_log_det_jacobian methods of p and q.

p A bijector instance.
x A tensor from the image of p.forward.
q A bijector instance of the same type as p, with matching shape.
y A tensor from the image of q.forward.
event_ndims The number of right-hand dimensions comprising the event shapes of x and y.
p_kwargs Keyword args to pass to p.
q_kwargs Keyword args to pass to q.

ildj_ratio log ((abs o det o jac p^-1)(x) / (abs o det o jac q^-1)(y)), i.e. in TFP code, p.inverse_log_det_jacobian(x, event_ndims) - q.inverse_log_det_jacobian(y, event_ndims). In some cases this will be computed with better than naive numerical precision, e.g. by moving differences inside of a sum reduction.