pygeoinf 1.0.8__py3-none-any.whl → 1.1.0__py3-none-any.whl

pygeoinf/symmetric_space/circle.py

@@ -1,8 +1,29 @@
 """
-Sobolev spaces for functions on a circle.
+Provides concrete implementations of function spaces on the circle (S¹).
+
+This module uses the abstract framework from the symmetric space module to create
+fully-featured `Lebesgue` (L²) and `Sobolev` (Hˢ) Hilbert spaces for functions
+defined on a circle.
+
+The core representation for a function is a truncated real Fourier series,
+and the module provides efficient methods to transform between the spatial domain
+(function values on a grid) and the frequency domain (Fourier coefficients)
+using the Fast Fourier Transform (FFT). This allows for the construction of
+differential operators and rotationally-invariant Gaussian measures on the
+circle, which are diagonal in the Fourier basis.
+
+Key Classes
+-----------
+- `CircleHelper`: A mixin class providing the core geometry, FFT machinery, and
+  plotting utilities.
+- `Lebesgue`: A concrete implementation of the L²(S¹) space of square-integrable
+  functions.
+- `Sobolev`: A concrete implementation of the Hˢ(S¹) space of functions with a
+  specified degree of smoothness.
 """
 
 from __future__ import annotations
+
 from typing import Callable, Tuple, Optional
 import matplotlib.pyplot as plt
 import numpy as np
@@ -12,117 +33,45 @@ from scipy.sparse import diags
 
 from matplotlib.figure import Figure
 from matplotlib.axes import Axes
-from pygeoinf.operators import LinearOperator
-from pygeoinf.forms import LinearForm
-from pygeoinf.gaussian_measure import GaussianMeasure
 
-from pygeoinf.hilbert_space import EuclideanSpace
-from pygeoinf.symmetric_space.symmetric_space import SymmetricSpaceSobolev
+from pygeoinf.hilbert_space import (
+    HilbertModule,
+    MassWeightedHilbertModule,
+)
+from pygeoinf.operators import LinearOperator
+from pygeoinf.linear_forms import LinearForm
+from .symmetric_space import (
+    AbstractInvariantLebesgueSpace,
+    AbstractInvariantSobolevSpace,
+)
 
 
-class Sobolev(SymmetricSpaceSobolev):
+class CircleHelper:
     """
-    Implementation of the Sobolev space H^s on a circle.
+    A mixin class providing common functionality for function spaces on the circle.
 
-    Functions on the circle are represented by their values on a grid of
-    equally spaced points, and the inner product is defined via their
-    Fourier coefficients.
+    This helper is not intended to be instantiated directly. It provides the core
+    geometry (radius, grid points), the FFT transform machinery, and plotting
+    utilities that are shared by both the `Lebesgue` and `Sobolev` space classes.
     """
 
-    def __init__(
-        self,
-        kmax: int,
-        order: float,
-        scale: float,
-        /,
-        *,
-        radius: float = 1.0,
-    ):
+    def __init__(self, kmax: int, radius: float):
         """
         Args:
-            kmax: The maximum Fourier degree to be represented in this space.
-            order: The Sobolev order, controlling the smoothness of functions.
-            scale: The Sobolev length-scale.
-            radius: The radius of the circle. Defaults to 1.0.
+            kmax: The maximum Fourier degree to be represented.
+            radius: Radius of the circle.
         """
         self._kmax: int = kmax
         self._radius: float = radius
 
-        super().__init__(
-            order,
-            scale,
-            2 * kmax,
-            self._to_components,
-            self._from_components,
-            self._inner_product,
-            self._to_dual,
-            self._from_dual,
-            vector_multiply=lambda u1, u2: u1 * u2,
-        )
-
-        self._fft_factor: float = np.sqrt(2 * np.pi * radius) / self.dim
+        self._fft_factor: float = np.sqrt(2 * np.pi * radius) / (2 * self.kmax)
         self._inverse_fft_factor: float = 1.0 / self._fft_factor
 
-        values = np.zeros(self.kmax + 1)
-        values[0] = 1
-        for k in range(1, self.kmax + 1):
-            values[k] = 2 * self._sobolev_function(k)
-
-        self._metric = diags([values], [0])
-        self._inverse_metric = diags([np.reciprocal(values)], [0])
-
-    @staticmethod
-    def from_sobolev_parameters(
-        order: float,
-        scale: float,
-        /,
-        *,
-        radius: float = 1.0,
-        rtol: float = 1e-8,
-        power_of_two: bool = False,
-    ) -> "Sobolev":
-        """
-        Creates an instance with `kmax` chosen based on Sobolev parameters.
-
-        The method estimates the truncation error for the Dirac measure and is
-        only applicable for spaces with order > 0.5.
-
-        Args:
-            order: The Sobolev order. Must be > 0.5.
-            scale: The Sobolev length-scale.
-            radius: The radius of the circle. Defaults to 1.0.
-            rtol: Relative tolerance used in assessing truncation error.
-                Defaults to 1e-8.
-            power_of_two: If True, `kmax` is set to the next power of two.
-
-        Returns:
-            An instance of the Sobolev class with an appropriate `kmax`.
-
-        Raises:
-            ValueError: If order is <= 0.5.
-        """
-        if order <= 0.5:
-            raise ValueError("This method is only applicable for orders > 0.5")
-
-        summation = 1.0
-        k = 0
-        err = 1.0
-        while err > rtol:
-            k += 1
-            term = (1 + (scale * k / radius) ** 2) ** -order
-            summation += 2 * term
-            err = 2 * term / summation
-            if k > 10000:
-                raise RuntimeError("Failed to converge on a stable kmax.")
-
-        if power_of_two:
-            n = int(np.log2(k))
-            k = 2 ** (n + 1)
-
-        return Sobolev(k, order, scale, radius=radius)
+    def _space(self):
+        return self
 
     @property
-    def kmax(self) -> int:
+    def kmax(self):
         """The maximum Fourier degree represented in this space."""
         return self._kmax
 
@@ -134,7 +83,28 @@ class Sobolev(SymmetricSpaceSobolev):
     @property
     def angle_spacing(self) -> float:
         """The angular spacing between grid points."""
-        return 2 * np.pi / self.dim
+        return np.pi / self.kmax
+
+    @property
+    def spatial_dimension(self) -> int:
+        """The dimension of the space."""
+        return 1
+
+    @property
+    def fft_factor(self) -> float:
+        """
+        The factor by which the Fourier coefficients are scaled
+        in forward transformations.
+        """
+        return self._fft_factor
+
+    @property
+    def inverse_fft_factor(self) -> float:
+        """
+        The factor by which the Fourier coefficients are scaled
+        in inverse transformations.
+        """
+        return self._inverse_fft_factor
 
     def random_point(self) -> float:
         """Returns a random angle in the interval [0, 2*pi)."""
@@ -143,10 +113,40 @@ class Sobolev(SymmetricSpaceSobolev):
     def angles(self) -> np.ndarray:
         """Returns a numpy array of the grid point angles."""
         return np.fromiter(
-            [i * self.angle_spacing for i in range(self.dim)],
+            [i * self.angle_spacing for i in range(2 * self.kmax)],
             float,
         )
 
+    def laplacian_eigenvalue(self, k: int) -> float:
+        """
+        Returns the k-th eigenvalue of the Laplacian.
+
+        Args:
+            k: The index of the eigenvalue to return.
+        """
+        return (k / self.radius) ** 2
+
+    def trace_of_invariant_automorphism(self, f: Callable[[float], float]) -> float:
+        """
+        Returns the trace of the automorphism of the form f(Δ) with f a function
+        that is well-defined on the spectrum of the Laplacian.
+
+        Args:
+            f: A real-valued function that is well-defined on the spectrum
+               of the Laplacian.
+
+        Notes:
+            The method takes account of the Nyquist theorem for real fast Fourier transforms,
+            this meaning that element at k = -kmax is excluded from the trace.
+        """
+        trace = f(self.laplacian_eigenvalue(0))
+        if self.kmax > 0:
+            trace += f(self.laplacian_eigenvalue(self.kmax))
+        trace += 2 * np.sum(
+            [f(self.laplacian_eigenvalue(k)) for k in range(1, self.kmax)]
+        )
+        return float(trace)
+
     def project_function(self, f: Callable[[float], float]) -> np.ndarray:
         """
         Returns an element of the space by projecting a given function.
@@ -158,6 +158,14 @@ class Sobolev(SymmetricSpaceSobolev):
         """
         return np.fromiter((f(theta) for theta in self.angles()), float)
 
+    def to_coefficient(self, u: np.ndarray) -> np.ndarray:
+        """Maps a function vector to its complex Fourier coefficients."""
+        return rfft(u) * self.fft_factor
+
+    def from_coefficient(self, coeff: np.ndarray) -> np.ndarray:
+        """Maps complex Fourier coefficients to a function vector."""
+        return irfft(coeff, n=2 * self.kmax) * self._inverse_fft_factor
+
     def plot(
         self,
         u: np.ndarray,
@@ -218,154 +226,291 @@ class Sobolev(SymmetricSpaceSobolev):
         ax.fill_between(self.angles(), u - u_bound, u + u_bound, **kwargs)
         return fig, ax
 
-    def invariant_automorphism(self, f: Callable[[float], float]) -> "LinearOperator":
-        values = np.fromiter(
-            (f(k * k / self.radius**2) for k in range(self.kmax + 1)), dtype=float
-        )
-        matrix = diags([values], [0])
+    def _coefficient_to_component(self, coeff: np.ndarray) -> np.ndarray:
+        """Packs complex Fourier coefficients into a real component vector."""
+        # For a real-valued input, the output of rfft (real FFT) has
+        # conjugate symmetry. This implies that the imaginary parts of the
+        # zero-frequency (k=0) and Nyquist-frequency (k=kmax) components
+        # are always zero. We omit them from the component vector to create
+        # a minimal, non-redundant representation.
+        return np.concatenate((coeff.real, coeff.imag[1 : self.kmax]))
 
-        def mapping(u: np.ndarray) -> np.ndarray:
-            coeff = self.to_coefficient(u)
-            coeff = matrix @ coeff
-            return self.from_coefficient(coeff)
+    def _component_to_coefficient(self, c: np.ndarray) -> np.ndarray:
+        """Unpacks a real component vector into complex Fourier coefficients."""
+        # This is the inverse of `_coefficient_to_component`. It reconstructs
+        # the full complex coefficient array that irfft expects. We re-insert
+        # the known zeros for the imaginary parts of the zero-frequency (k=0)
+        # and Nyquist-frequency (k=kmax) components, which were removed to
+        # create the minimal real-valued representation.
+        coeff_real = c[: self.kmax + 1]
+        coeff_imag = np.concatenate([[0], c[self.kmax + 1 :], [0]])
+        return coeff_real + 1j * coeff_imag
 
-        return LinearOperator.formally_self_adjoint(self, mapping)
 
-    def invariant_gaussian_measure(
+class Lebesgue(CircleHelper, HilbertModule, AbstractInvariantLebesgueSpace):
+    """
+    Implementation of the Lebesgue space L² on a circle.
+
+    This class represents square-integrable functions on a circle. A function is
+    represented by its values on an evenly spaced grid. The L² inner product
+    is correctly implemented by accounting for the non-orthonormality of the
+    real Fourier basis functions.
+    """
+
+    def __init__(
         self,
-        f: Callable[[float], float],
+        kmax: int,
         /,
         *,
-        expectation: Optional[np.ndarray] = None,
-    ) -> "GaussianMeasure":
-        values = np.fromiter(
-            (np.sqrt(f(k * k / self.radius**2)) for k in range(self.kmax + 1)),
-            dtype=float,
-        )
-        matrix = diags([values], [0])
+        radius: float = 1.0,
+    ):
+        """
+        Args:
+        kmax: The maximum Fourier degree to be represented.
+        radius: Radius of the circle. Defaults to 1.0.
+        """
 
-        domain = EuclideanSpace(self.dim)
-        codomain = self
+        if kmax < 0:
+            raise ValueError("kmax must be non-negative")
 
-        def mapping(c: np.ndarray) -> np.ndarray:
-            coeff = self._component_to_coefficient(c)
-            coeff = matrix @ coeff
-            return self.from_coefficient(coeff)
+        self._dim = 2 * kmax
 
-        def formal_adjoint(u: np.ndarray) -> np.ndarray:
-            coeff = self.to_coefficient(u)
-            coeff = matrix @ coeff
-            return self._coefficient_to_component(coeff)
+        CircleHelper.__init__(self, kmax, radius)
 
-        covariance_factor = LinearOperator(
-            domain, codomain, mapping, formal_adjoint_mapping=formal_adjoint
+        values = np.fromiter(
+            [2 if k > 0 else 1 for k in range(self.kmax + 1)], dtype=float
         )
+        self._metric = diags([values], [0])
+        self._inverse_metric = diags([np.reciprocal(values)], [0])
 
-        return GaussianMeasure(
-            covariance_factor=covariance_factor,
-            expectation=expectation,
-        )
+    @property
+    def dim(self) -> int:
+        """The dimension of the space."""
+        return self._dim
 
-    def dirac(self, point: float) -> "LinearForm":
-        coeff = np.zeros(self.kmax + 1, dtype=complex)
-        fac = np.exp(-1j * point)
-        coeff[0] = 1.0
-        for k in range(1, coeff.size):
-            coeff[k] = coeff[k - 1] * fac
-        coeff *= 1.0 / np.sqrt(2 * np.pi * self.radius)
-        coeff[1:] *= 2.0
-        cp = self._coefficient_to_component(coeff)
-        return LinearForm(self, components=cp)
+    def to_components(self, u: np.ndarray) -> np.ndarray:
+        """Converts a function vector to its real component representation."""
+        coeff = self.to_coefficient(u)
+        return self._coefficient_to_component(coeff)
 
-    def to_coefficient(self, u: np.ndarray) -> np.ndarray:
-        """Maps a function vector to its complex Fourier coefficients."""
-        return rfft(u) * self._fft_factor
+    def from_components(self, c: np.ndarray) -> np.ndarray:
+        """Converts a real component vector back to a function vector."""
+        coeff = self._component_to_coefficient(c)
+        return self.from_coefficient(coeff)
 
-    def from_coefficient(self, coeff: np.ndarray) -> np.ndarray:
-        """Maps complex Fourier coefficients to a function vector."""
-        return irfft(coeff, n=self.dim) * self._inverse_fft_factor
+    def to_dual(self, u: np.ndarray) -> "LinearForm":
+        """Maps a vector `u` to its dual representation `u*`."""
+        coeff = self.to_coefficient(u)
+        cp = self._coefficient_to_component(self._metric @ coeff)
+        return self.dual.from_components(cp)
 
+    def from_dual(self, up: "LinearForm") -> np.ndarray:
+        """Maps a dual vector `u*` back to its primal representation `u`."""
+        cp = self.dual.to_components(up)
+        dual_coeff = self._component_to_coefficient(cp)
+        primal_coeff = self._inverse_metric @ dual_coeff
+        return self.from_coefficient(primal_coeff)
+
+    def vector_multiply(self, x1: np.ndarray, x2: np.ndarray) -> np.ndarray:
+        """
+        Computes the pointwise product of two vectors.
+        """
+        return x1 * x2
+
+    def eigenfunction_norms(self) -> np.ndarray:
+        """Returns a list of the norms of the eigenfunctions."""
+        return np.fromiter(
+            [np.sqrt(2) if i > 0 else 1 for i in range(self.dim)],
+            dtype=float,
+        )
 
     def __eq__(self, other: object) -> bool:
         """
-        Checks for mathematical equality with another Sobolev space on a circle.
+        Checks for mathematical equality with another Lebesgue space on a circle.
 
         Two spaces are considered equal if they are of the same type and have
         the same defining parameters (kmax, order, scale, and radius).
         """
-        if not isinstance(other, Sobolev):
+        if not isinstance(other, Lebesgue):
             return NotImplemented
-        
-        return (self.kmax == other.kmax and
-                self.order == other.order and
-                self.scale == other.scale and
-                self.radius == other.radius)
-
-    # ================================================================#
-    #                         Private methods                         #
-    # ================================================================#
-
-    def _sobolev_function(self, k: int) -> float:
-        """Computes the diagonal entries of the Sobolev metric in Fourier space."""
-        return (1 + (self.scale * k / self.radius) ** 2) ** self.order
-
-    def _coefficient_to_component(self, coeff: np.ndarray) -> np.ndarray:
-        """Packs complex Fourier coefficients into a real component vector."""
-        return np.concatenate((coeff.real, coeff.imag[1 : self.kmax]))
 
-    def _component_to_coefficient(self, c: np.ndarray) -> np.ndarray:
-        """Unpacks a real component vector into complex Fourier coefficients."""
-        coeff_real = c[: self.kmax + 1]
-        coeff_imag = np.concatenate([[0], c[self.kmax + 1 :], [0]])
-        return coeff_real + 1j * coeff_imag
+        return self.kmax == other.kmax and self.radius == other.radius
 
-    def _to_components(self, u: np.ndarray) -> np.ndarray:
-        """Converts a function vector to its real component representation."""
-        coeff = self.to_coefficient(u)
-        return self._coefficient_to_component(coeff)
+    def invariant_automorphism(self, f: Callable[[float], float]):
+        """
+        Implements an invariant automorphism of the form f(Δ) using Fourier
+        expansions on a circle.
 
-    def _from_components(self, c: np.ndarray) -> np.ndarray:
-        """Converts a real component vector back to a function vector."""
-        coeff = self._component_to_coefficient(c)
-        return self.from_coefficient(coeff)
+        Args:
+            f: A real-valued function that is well-defined on the spectrum
+               of the Laplacian, Δ.
+        """
 
-    def _inner_product(self, u1: np.ndarray, u2: np.ndarray) -> float:
-        """Computes the H^s inner product in the Fourier domain."""
-        coeff1 = self.to_coefficient(u1)
-        coeff2 = self.to_coefficient(u2)
-        return np.real(np.vdot(self._metric @ coeff1, coeff2))
+        values = np.fromiter(
+            (f(self.laplacian_eigenvalue(k)) for k in range(self.kmax + 1)),
+            dtype=float,
+        )
+        matrix = diags([values], [0])
 
-    def _to_dual(self, u: np.ndarray) -> "LinearForm":
-        """Maps a vector `u` to its dual representation `u*`."""
-        coeff = self.to_coefficient(u)
-        cp = self._coefficient_to_component(self._metric @ coeff)
-        return self.dual.from_components(cp)
+        def mapping(u):
+            coeff = self.to_coefficient(u)
+            coeff = matrix @ coeff
+            return self.from_coefficient(coeff)
 
-    def _from_dual(self, up: "LinearForm") -> np.ndarray:
-        """Maps a dual vector `u*` back to its primal representation `u`."""
-        cp = self.dual.to_components(up)
-        coeff = self._component_to_coefficient(cp)
-        c = self._coefficient_to_component(self._inverse_metric @ coeff)
-        return self.from_components(c)
+        return LinearOperator.self_adjoint(self, mapping)
 
 
-class Lebesgue(Sobolev):
+class Sobolev(
+    CircleHelper,
+    MassWeightedHilbertModule,
+    AbstractInvariantSobolevSpace,
+):
     """
-    Implementation of the Lebesgue space L^2 on a circle.
+    Implementation of the Sobolev space Hˢ on a circle.
 
-    This is a special case of the Sobolev space with order s=0.
+    This class represents functions with a specified degree of smoothness. It is
+    constructed as a `MassWeightedHilbertModule` over the `Lebesgue` space, where
+    the mass operator weights the Fourier coefficients to enforce smoothness. This
+    is the primary class for defining smooth, random function fields on the circle.
     """
 
     def __init__(
         self,
         kmax: int,
+        order: float,
+        scale: float,
         /,
         *,
         radius: float = 1.0,
     ):
         """
         Args:
-            kmax: The maximum Fourier degree to be represented.
-            radius: Radius of the circle. Defaults to 1.0.
+        kmax: The maximum Fourier degree to be represented.
+        order: The Sobolev order, controlling the smoothness of functions.
+        scale: The Sobolev length-scale.
+        radius: Radius of the circle. Defaults to 1.0.
+        """
+
+        CircleHelper.__init__(self, kmax, radius)
+        AbstractInvariantSobolevSpace.__init__(self, order, scale)
+
+        lebesgue = Lebesgue(kmax, radius=radius)
+
+        mass_operator = lebesgue.invariant_automorphism(self.sobolev_function)
+        inverse_mass_operator = lebesgue.invariant_automorphism(
+            lambda k: 1.0 / self.sobolev_function(k)
+        )
+
+        MassWeightedHilbertModule.__init__(
+            self, lebesgue, mass_operator, inverse_mass_operator
+        )
+
+    @staticmethod
+    def from_sobolev_parameters(
+        order: float,
+        scale: float,
+        /,
+        *,
+        radius: float = 1.0,
+        rtol: float = 1e-6,
+        power_of_two: bool = False,
+    ) -> "Sobolev":
         """
-        super().__init__(kmax, 0.0, 1.0, radius=radius)
+        Creates an instance with `kmax` chosen based on Sobolev parameters.
+
+        The method estimates the truncation error for the Dirac measure and is
+        only applicable for spaces with order > 0.5.
+
+        Args:
+            order: The Sobolev order. Must be > 0.5.
+            scale: The Sobolev length-scale.
+            radius: The radius of the circle. Defaults to 1.0.
+            rtol: Relative tolerance used in assessing truncation error.
+                Defaults to 1e-8.
+            power_of_two: If True, `kmax` is set to the next power of two.
+
+        Returns:
+            An instance of the Sobolev class with an appropriate `kmax`.
+
+        Raises:
+            ValueError: If order is <= 0.5.
+        """
+        if order <= 0.5:
+            raise ValueError("This method is only applicable for orders > 0.5")
+
+        summation = 1.0
+        k = 0
+        err = 1.0
+        while err > rtol:
+            k += 1
+            term = (1 + (scale * k / radius) ** 2) ** -order
+            summation += 2 * term
+            err = 2 * term / summation
+            if k > 100000:
+                raise RuntimeError("Failed to converge on a stable kmax.")
+
+        if power_of_two:
+            n = int(np.log2(k))
+            k = 2 ** (n + 1)
+
+        return Sobolev(k, order, scale, radius=radius)
+
+    def __eq__(self, other: object) -> bool:
+        """
+        Checks for mathematical equality with another Sobolev space on a circle.
+
+        Two spaces are considered equal if they are of the same type and have
+        the same defining parameters (kmax, order, scale, and radius).
+        """
+        if not isinstance(other, Sobolev):
+            return NotImplemented
+
+        return (
+            self.kmax == other.kmax
+            and self.radius == other.radius
+            and self.order == other.order
+            and self.scale == other.scale
+        )
+
+    def eigenfunction_norms(self) -> np.ndarray:
+        """Returns a list of the norms of the eigenfunctions."""
+        values = self.underlying_space.eigenfunction_norms()
+
+        i = 0
+        for k in range(self.kmax + 1):
+            values[i] *= np.sqrt(self.sobolev_function(self.laplacian_eigenvalue(k)))
+            i += 1
+
+        for k in range(1, self.kmax):
+            values[i] *= np.sqrt(self.sobolev_function(self.laplacian_eigenvalue(k)))
+            i += 1
+
+        return values
+
+    def dirac(self, point: float) -> LinearForm:
+        """
+        Returns the linear functional corresponding to a point evaluation.
+
+        This represents the action of the Dirac delta measure based at the given
+        point.
+
+        Args:
+            point: The angle for the point at which the measure is based.
+
+        Raises:
+            ValueError: If the Sobolev order is less than 1/2.
+        """
+        if self.order <= 1 / 2:
+            raise NotImplementedError(
+                "This method is only applicable for orders >= 1/2"
+            )
+
+        coeff = np.zeros(self.kmax + 1, dtype=complex)
+        fac = np.exp(-1j * point)
+        coeff[0] = 1.0
+        for k in range(1, coeff.size):
+            coeff[k] = coeff[k - 1] * fac
+        coeff *= 1.0 / np.sqrt(2 * np.pi * self.radius)
+        coeff[1:] *= 2.0
+        cp = self._coefficient_to_component(coeff)
+        return LinearForm(self, components=cp)