Added HypercylindricalStateSpaceSubdivisionDistribution

Author: FlorianPfaff

pyrecest/distributions/cart_prod/hypercylindrical_state_space_subdivision_distribution.py

@@ -0,0 +1,47 @@
+from pyrecest.distributions.cart_prod.state_space_subdivision_distribution import StateSpaceSubdivisionDistribution
+from pyrecest.distributions.cart_prod.abstract_hypercylindrical_distribution import AbstractHypercylindricalDistribution
+from pyrecest.distributions.nonperiodic.custom_linear_distribution import CustomLinearDistribution
+from pyrecest.distributions.circle.circular_uniform_distribution import CircularUniformDistribution
+from scipy.integrate import quad
+import numpy as np
+
+class HypercylindricalStateSpaceSubdivisionDistribution(StateSpaceSubdivisionDistribution, AbstractHypercylindricalDistribution):
+
+    def __init__(self, gd_, lin_distributions):
+        StateSpaceSubdivisionDistribution.__init__(self, gd_, lin_distributions)
+
+    def plot(self, interpolate=False):
+        if interpolate:
+            return AbstractHypercylindricalDistribution.plot(self)
+        else:
+            return StateSpaceSubdivisionDistribution.plot(self)
+
+    def plot_interpolated(self):
+        return self.plot(interpolate=True)
+
+    def mode(self):
+        return StateSpaceSubdivisionDistribution.mode(self)
+
+    @staticmethod
+    def from_distribution(distribution, no_of_grid_points, grid_type='CartesianProd'):
+        return HypercylindricalStateSpaceSubdivisionDistribution.from_function(
+            distribution.pdf, no_of_grid_points, distribution.linD, distribution.boundD, grid_type)
+
+    @staticmethod
+    def from_function(fun, no_of_grid_points, dim_lin, dim_bound=1, grid_type='CartesianProd', int_range=(-np.inf , np.inf)):
+        assert dim_lin == 1, 'Currently, bounded dimension must be 1.'
+
+        gd = CircularGridDistribution.from_distribution(CircularUniformDistribution(), no_of_grid_points)
+        grid = gd.get_grid()
+        cds = [None] * no_of_grid_points
+
+        for i in range(no_of_grid_points):
+            fun_curr = lambda y: np.reshape(fun(np.vstack((grid[i] * np.ones_like(y), y))), np.shape(y))
+
+            # Obtain grid value via integral
+            gd.grid_values[i], _ = quad(fun_curr, int_range[0], int_range[1])
+
+            # Original function divided by grid value is linear
+            cds[i] = CustomLinearDistribution(lambda x: fun_curr(x) / gd.grid_values[i], 1)
+
+        return HypercylindricalStateSpaceSubdivisionDistribution(gd, cds)