RegressionProblem

class probnum.problems.RegressionProblem(observations, locations, likelihood=None, solution=None)[source]

Bases: object

Regression problem.

Fit a stochastic process to data, given a likelihood (realised by a DiscreteGaussian transition). Solved by filters and smoothers in probnum.filtsmooth.

Parameters

  • observations (ndarray) – Observations of the latent process.

  • locations (ndarray) – Grid-points on which the observations were taken.

  • likelihood (Optional[DiscreteGaussian]) – Likelihood of the observations; that is, relation between the latent process and the observed values. Encodes for example noise.

  • solution (Optional[Callable[[ndarray], Union[float, ndarray]]]) – Closed form, analytic solution to the problem. Used for testing and benchmarking.

Examples

>>> obs = [11.4123, -15.5123]
>>> loc = [0.1, 0.2]
>>> rp = RegressionProblem(observations=obs, locations=loc)
>>> rp
RegressionProblem(observations=[11.4123, -15.5123], locations=[0.1, 0.2], likelihood=None, solution=None)
>>> rp.observations
[11.4123, -15.5123]

Attributes Summary

likelihood

solution

Attributes Documentation

likelihood: Optional[probnum.statespace.discrete_transition.DiscreteGaussian] = None
solution: Optional[Callable[numpy.ndarray, Union[float, numpy.ndarray]]] = None