functional.models.predict_glm()

Register GLM-style observation/forecast sites from a latent predictor.

Usage

Source

functional.models.predict_glm(
    h,
    obs_dist_fn,
    latent,
)

The generalized-linear counterpart of predict(): instead of a zero-centered noise distribution shifted by a mean, the caller supplies a link obs_dist_fn that maps the full-horizon latent predictor to the observation distribution directly (e.g. lambda eta: Poisson(jnp.exp(eta))). The prefix/suffix mirroring is identical to predict(): while training the observation is observed; while forecasting the in-sample prefix is observed and the forecast suffix is sampled and exposed as the "forecast" deterministic site. The observation distribution must support time-axis surgery (~numpyro_forecast.surgery.slice_time() / ~numpyro_forecast.surgery.prefix_condition()), i.e. an elementwise family.

Parameters

h: Horizon

The horizon for the current model call (see Horizon).

obs_dist_fn: Callable[[Array], dist.Distribution]

Link mapping the full-horizon latent to the observation distribution (time at axis -2, shape (*batch, duration, obs)).

latent: Array

The deterministic latent predictor over the full horizon, time at axis -2.

Raises

RuntimeError

If forecasting (future > 0) but no observed data is available.

ValueError

If the observation distribution has discrete support but h.data is not integer-dtyped (the usual mistake for count models).