Added AbstractGridFilter

pyrecest/filters/abstract_grid_filter.py

@@ -0,0 +1,39 @@
+import warnings
+import numpy as np
+from .abstract_filter_type import AbstractFilterType
+from pyrecest.distributions.abstract_grid_distribution import AbstractGridDistribution
+
+class AbstractGridFilter(AbstractFilterType):
+    def __init__(self, gd):
+        if not isinstance(gd, AbstractGridDistribution):
+            raise ValueError("gd must be an instance of AbstractGridDistribution")
+        self.filter_state = gd
+
+    @property 
+    def filter_state(self) -> AbstractGridDistribution:
+        """
+        Get the current estimate of the grid distribution.
+        Returns : AbstractGridDistribution : the current grid distribution
+        """
+        return self._filter_state
+
+    @filter_state.setter
+    def filter_state(self, new_state):
+        if not isinstance(new_state, AbstractGridDistribution):
+            warnings.warn("gd_ is not a GridDistribution. Transforming the distribution with a number of coefficients equal to that of the filter.", RuntimeWarning)
+            new_state = self.filter_state.from_distribution(new_state, len(self.filter_state.grid_values), self.filter_state.enforce_pdf_nonnegative)
+
+        elif np.shape(self.filter_state.grid_values) != np.shape(new_state.grid_values):
+            warnings.warn("New grid has a different number of grid points.", RuntimeWarning)
+
+        self._filter_state = new_state
+
+    def update_nonlinear(self, likelihood, z):
+        grid_vals_new = self.filter_state.grid_values * np.reshape(likelihood(z, self.filter_state.get_grid()), (-1, 1))
+        assert np.shape(grid_vals_new) == np.shape(self.filter_state.grid_values)
+
+        self.filter_state.grid_values = grid_vals_new
+        self.filter_state = self.filter_state.normalize(suppress_warning = True)
+
+    def plot_filter_state(self):
+        self.filter_state.plot_state()