Prediction: Forward Simulation

HybridModel.predict and HybridModel.simulate cover one model. predict_ensemble predicts with several and returns the member mean as a PredictionResult, with the per-channel spread in each experiment's std and each member's own result in members. evaluate_predictor calls one predictor outside any ODE.

Quick links

• predict_ensemble
• evaluate_predictor

---

predict_ensemble()

from jaxhybridmodels.prediction import predict_ensemble  ·  also re-exported as jaxhybridmodels.predict_ensemble

predict_ensemble(
    models: 'Sequence[HybridModel | FitResult]',
    data: 'Data',
    ts: 'ArrayLike | None' = None,
    n_points: 'int | None' = None,
) -> PredictionResult

Predict with every model and return the member mean, with the spread.

outputs (and buckets) hold the mean across members, so compute_metrics scores the ensemble mean. Each experiment's std holds the per-channel standard deviation across members (population, ddof=0; zero for one member), and members holds each model's own PredictionResult.

models holds HybridModel\ s or FitResult\ s, so the list fit_ensemble returns can be passed as is. They must share output_channel_names. data, ts and n_points are as for HybridModel.predict.

Source

---

evaluate_predictor()

from jaxhybridmodels.prediction import evaluate_predictor  ·  also re-exported as jaxhybridmodels.evaluate_predictor

evaluate_predictor(predictor: 'Any', inputs: 'Mapping[str, ArrayLike]') -> 'Array'

Evaluate a predictor at named inputs, vectorised over array-valued ones.

The inputs are broadcast against each other, so scalars, a 1-D sweep, or a 2-D grid built from x[:, None] and y[None, :] all work. The predictor is called once per broadcast point with a dict of scalars, the way a simulate_fn calls it, and the result has shape broadcast_shape + output_shape. A one-output BoundedPredictor has output shape (), so a sweep over 50 temperatures gives [50].

Examples

::

hm.evaluate_predictor(predictors["rate"], {"T": 300.0})  # shape ()
hm.evaluate_predictor(predictors["rate"], {"T": temps, "pH": 6.5})  # [len(temps)]

Returns

Item	Type	Description
Array		Call float(...) on a scalar result for a Python number.

Source