Split discrete and continuous Kalman filter implementations

benchmarks/filtsmooth.py

@@ -49,7 +49,9 @@ def setup(self, linearization_implementation):
         )
         self.regression_problem = regression_problem
 
-        self.kalman_filter = filtsmooth.gaussian.Kalman(prior_process=prior_process)
+        self.kalman_filter = filtsmooth.gaussian.DiscreteKalman(
+            prior_process=prior_process
+        )
 
     def time_filter(self, linearization_implementation):
         self.kalman_filter.filter(self.regression_problem)
@@ -99,7 +101,9 @@ def setup(self, linearization_implementation):
             initarg=regression_problem.locations[0],
         )
 
-        self.kalman_filter = filtsmooth.gaussian.Kalman(prior_process=prior_process)
+        self.kalman_filter = filtsmooth.gaussian.DiscreteKalman(
+            prior_process=prior_process
+        )
         self.filtering_posterior, _ = self.kalman_filter.filter(regression_problem)
 
     def time_smooth(self, linearization_implementation):
@@ -166,7 +170,9 @@ def setup(self, linearization_implementation, num_samples):
             initarg=regression_problem.locations[0],
         )
 
-        self.kalman_filter = filtsmooth.gaussian.Kalman(prior_process=prior_process)
+        self.kalman_filter = filtsmooth.gaussian.DiscreteKalman(
+            prior_process=prior_process
+        )
 
         self.filtering_posterior, _ = self.kalman_filter.filter(regression_problem)
         self.smoothing_posterior = self.kalman_filter.smooth(

docs/source/tutorials/filtsmooth/continuous_linear_gaussian_filtering_smoothing.ipynb

@@ -302,7 +302,7 @@
    "metadata": {},
    "outputs": [],
    "source": [
-    "kalman_filter = filtsmooth.gaussian.Kalman(prior_process)"
+    "kalman_filter = filtsmooth.gaussian.ContinuousKalman(prior_process)"
    ]
   },
   {

docs/source/tutorials/filtsmooth/continuous_nonlinear_gaussian_filtering_smoothing.ipynb

@@ -248,7 +248,7 @@
     "prior_process = randprocs.markov.MarkovProcess(\n",
     "    transition=linearised_dynamics_model, initrv=initial_state_rv, initarg=0.0\n",
     ")\n",
-    "kalman_filter = filtsmooth.gaussian.Kalman(prior_process)"
+    "kalman_filter = filtsmooth.gaussian.ContinuousKalman(prior_process)"
    ]
   },
   {

docs/source/tutorials/filtsmooth/discrete_linear_gaussian_filtering_smoothing.ipynb

@@ -292,7 +292,7 @@
    "metadata": {},
    "outputs": [],
    "source": [
-    "kalman_filter = filtsmooth.gaussian.Kalman(prior_process)"
+    "kalman_filter = filtsmooth.gaussian.DiscreteKalman(prior_process)"
    ]
   },
   {

docs/source/tutorials/filtsmooth/discrete_nonlinear_gaussian_filtering_smoothing.ipynb

@@ -350,7 +350,7 @@
    "metadata": {},
    "outputs": [],
    "source": [
-    "kalman_filter = filtsmooth.gaussian.Kalman(prior_process)"
+    "kalman_filter = filtsmooth.gaussian.DiscreteKalman(prior_process)"
    ]
   },
   {

src/probnum/diffeq/odefilter/init_routines/_non_probabilistic_fit.py

@@ -70,7 +70,7 @@ def _improve(self, *, data, prior_process):
         )
 
         # Infer the solution
-        kalman = filtsmooth.gaussian.Kalman(prior_process)
+        kalman = filtsmooth.gaussian.ContinuousKalman(prior_process)
         out, _ = kalman.filtsmooth(regression_problem)
         estimated_initrv = out.states[0]
         return estimated_initrv

src/probnum/filtsmooth/_kalman_filter_smoother.py

@@ -8,6 +8,11 @@
 
 __all__ = ["filter_kalman", "smooth_rts"]
 
+MATCH_FILTER = {
+    "continuous": gaussian.ContinuousKalman,
+    "discrete": gaussian.DiscreteKalman,
+}
+
 
 def filter_kalman(
     observations: ArrayLike,
@@ -100,7 +105,7 @@ def filter_kalman(
     prior_process = _setup_prior_process(
         F=F, L=L, m0=m0, C0=C0, t0=locations[0], prior_model=prior_model
     )
-    kalman = gaussian.Kalman(prior_process)
+    kalman = MATCH_FILTER[prior_model](prior_process)
     return kalman.filter(regression_problem)[0]
 
 
@@ -195,7 +200,7 @@ def smooth_rts(
     prior_process = _setup_prior_process(
         F=F, L=L, m0=m0, C0=C0, t0=locations[0], prior_model=prior_model
     )
-    kalman = gaussian.Kalman(prior_process)
+    kalman = MATCH_FILTER[prior_model](prior_process)
     return kalman.filtsmooth(regression_problem)[0]
 
 

src/probnum/filtsmooth/gaussian/__init__.py

@@ -1,19 +1,21 @@
 """Gaussian filtering and smoothing."""
 
-from ._kalman import Kalman
+from ._kalman import ContinuousKalman, DiscreteKalman
 from ._kalmanposterior import FilteringPosterior, KalmanPosterior, SmoothingPosterior
 
 # Public classes and functions. Order is reflected in documentation.
 __all__ = [
-    "Kalman",
+    "DiscreteKalman",
+    "ContinuousKalman",
     "KalmanPosterior",
     "FilteringPosterior",
     "SmoothingPosterior",
 ]
 
 # Set correct module paths (for superclasses).
 # Corrects links and module paths in documentation.
-Kalman.__module__ = "probnum.filtsmooth.gaussian"
+DiscreteKalman.__module__ = "probnum.filtsmooth.gaussian"
+ContinuousKalman.__module__ = "probnum.filtsmooth.gaussian"
 KalmanPosterior.__module__ = "probnum.filtsmooth.gaussian"
 FilteringPosterior.__module__ = "probnum.filtsmooth.gaussian"
 SmoothingPosterior.__module__ = "probnum.filtsmooth.gaussian"