Added HypercylindricalStateSpaceSubdivisionGaussianDistribution

Author: FlorianPfaff

pyrecest/distributions/cart_prod/hypercylindrical_state_space_subdivision_gaussian_distribution.py

@@ -0,0 +1,76 @@
+from pyrecest.distributions.cart_prod.state_space_subdivision_gaussian_distribution import StateSpaceSubdivisionGaussianDistribution
+from pyrecest.distributions.cart_prod.hypercylindrical_state_space_subdivision_distribution import HypercylindricalStateSpaceSubdivisionDistribution
+from pyrecest.distributions import GaussianDistribution
+import numpy as np
+
+
+class HypercylindricalStateSpaceSubdivisionGaussianDistribution(StateSpaceSubdivisionGaussianDistribution, HypercylindricalStateSpaceSubdivisionDistribution):
+
+    def __init__(self, gd_, gaussians):
+        HypercylindricalStateSpaceSubdivisionDistribution.__init__(self, gd_, gaussians)
+        StateSpaceSubdivisionGaussianDistribution.__init__(self, gd_, gaussians)
+
+    def mode(self):
+        return StateSpaceSubdivisionGaussianDistribution.mode(self)
+
+    def linear_mean(self):
+        return StateSpaceSubdivisionGaussianDistribution.linear_mean(self)
+
+    def linear_covariance(self):
+        return StateSpaceSubdivisionGaussianDistribution.linear_covariance(self)
+
+    def hybrid_mean(self):
+        return StateSpaceSubdivisionGaussianDistribution.hybrid_mean(self)
+
+    def hybrid_moment(self):
+        trig_mom_complex = self.gd.trigonometric_moment(1)
+        trig_mom_real = [trig_mom_complex.real, trig_mom_complex.imag]
+        return trig_mom_real + [self.linear_mean()]
+
+    @staticmethod
+    def from_distribution(distribution, no_of_grid_points, grid_type='CartesianProd'):
+        # Even for WN, the conditional is a HypercylindricalWN divided
+        # by a WN. No analytical formula is known to me
+        hcrbd_non_gauss = HypercylindricalStateSpaceSubdivisionDistribution.from_function(
+            distribution.pdf, no_of_grid_points, distribution.linD, distribution.boundD, grid_type)
+        # Superclass generates CustomLinearDistributions. Convert.
+        lin_dists = [GaussianDistribution.from_distribution(dist) for dist in hcrbd_non_gauss.linear_distributions]
+        return HypercylindricalStateSpaceSubdivisionGaussianDistribution(hcrbd_non_gauss.gd, lin_dists)
+
+    @staticmethod
+    def from_function(fun, no_of_grid_points, dim_lin, dim_bound=1, grid_type='CartesianProd', int_range=(-np.inf , np.inf)):
+        hcrbd_non_gauss = HypercylindricalStateSpaceSubdivisionDistribution.from_function(
+            fun, no_of_grid_points, dim_lin, dim_bound, grid_type, int_range)
+        return HypercylindricalStateSpaceSubdivisionGaussianDistribution(hcrbd_non_gauss.gd, lin_dists)
+
+
+"""
+class HypercylindricalStateSpaceSubdivisionGaussianDistribution:
+    def __init__(self, gd_, gaussians):
+        # In this class, the grid must be on the hypertorus.
+        self.gd_ = HypertoroidalGridDistribution(gd_)
+        self.gaussians = gaussians
+
+    def mode(self):
+        pass
+
+    def linear_mean(self):
+        pass
+
+    def linear_covariance(self):
+        pass
+
+    def hybrid_mean(self):
+        pass
+
+    def hybrid_moment(self):
+        pass
+
+    @classmethod
+    def from_distribution(cls, distribution, no_of_grid_points, grid_type='CartesianProd'):
+        pass
+
+    @classmethod
+    def from_function(cls, fun, no_of_grid_points, dim_lin, dim_bound=1, grid_type='CartesianProd', int_range=(-np.inf, np.inf)):
+        pass
+        """