surgery.prefix_condition()

Condition a (t+f)-length distribution on a t-length data prefix.

Usage

Source

surgery.prefix_condition(
    noise_dist,
    data,
)

For independent-over-time noise (the elementwise default) the conditional reduces to the forecast-horizon marginal, i.e. a time slice [t:]. Only registered elementwise families take this path; correlated families (e.g. MultivariateNormal) need a dedicated dispatch implementing a genuine Gaussian conditional, which is checked here rather than silently reduced to a slice (the R1 fix).

Parameters

noise_dist: dist.Distribution

The observation distribution over the full horizon (*batch, t+f, obs).

data: Array

The observed prefix with shape (*batch, t, obs).

Returns

dist.Distribution

The forecast-horizon distribution over (*batch, f, obs).

Raises

NotImplementedError

If noise_dist is not a registered elementwise family (and has no dedicated dispatch).