Euclidean updated

prefsampling/__init__.py

@@ -6,15 +6,14 @@
 from itertools import chain
 
 from prefsampling.approval import NoiseType
-from prefsampling.core.euclidean import EuclideanSpace
 from prefsampling.ordinal import TreeSampler
 
 
 class CONSTANTS(Enum):
     """ All constants of the package """
     _ignore_ = 'member cls'
     cls = vars()
-    for member in chain(list(EuclideanSpace), list(TreeSampler), list(NoiseType)):
+    for member in chain(list(TreeSampler), list(NoiseType)):
         if member.name in cls:
             raise ValueError(f"The name {member.name} is used in more than one enumeration. The"
                              f"CONSTANTS class needs unique names to be well-defined.")

prefsampling/approval/euclidean.py

@@ -1,18 +1,22 @@
 from __future__ import annotations
 
+from collections.abc import Callable
+
 import numpy as np
 
-from prefsampling.core.euclidean import election_positions, EuclideanSpace
+from prefsampling.core.euclidean import sample_election_positions
 from prefsampling.inputvalidators import validate_num_voters_candidates
 
 
 @validate_num_voters_candidates
 def euclidean(
     num_voters: int,
     num_candidates: int,
+    point_sampler: Callable,
+    point_sampler_args: dict,
+    candidate_point_sampler: Callable = None,
+    candidate_point_sampler_args: dict = None,
     radius: float = 0.5,
-    space: EuclideanSpace = EuclideanSpace.UNIFORM,
-    dimension: int = 2,
     seed: int = None,
 ) -> list[set[int]]:
     """
@@ -35,29 +39,44 @@ def euclidean(
             Number of Voters.
         num_candidates : int
             Number of Candidates.
+        point_sampler : Callable
+            The sampler used to sample point in the space. Used for both voters and candidates
+            unless a `candidate_space` is provided.
+        point_sampler_args : dict
+            The arguments passed to the `point_sampler`. The argument `num_points` is ignored
+            and replaced by the number of voters or candidates.
+        candidate_point_sampler : Callable, default: :code:`None`
+            The sampler used to sample the points of the candidates. If a value is provided,
+            then the `space` argument is only used for voters.
+        candidate_point_sampler_args : dict
+            The arguments passed to the `candidate_point_sampler`. The argument `num_points`
+            is ignored and replaced by the number of candidates.
         radius : float, default: 0.5
             Radius of approval.
-        space : EuclideanSpace, default: :py:const:`~prefsampling.core.euclidean.EuclideanSpace.UNIFORM`
-            Type of space considered. Should be a constant defined in the
-            :py:class:`~prefsampling.core.euclidean.EuclideanSpace`.
-        dimension : int, default: 2
-            Number of Dimensions.
-        seed : int
-            Seed for numpy random number generator.
+        seed : int, default: :code:`None`
+            Seed for numpy random number generator. Also passed to the point samplers if
+            a value is provided.
 
     Returns
     -------
         list[set[int]]
             Approval votes.
     """
-    rng = np.random.default_rng(seed)
-    votes = [set() for _ in range(num_voters)]
-    voters, candidates = election_positions(
-        num_voters, num_candidates, space, dimension, rng
+
+    voters_pos, candidates_pos = sample_election_positions(
+        num_voters,
+        num_candidates,
+        point_sampler,
+        point_sampler_args,
+        candidate_point_sampler,
+        candidate_point_sampler_args,
+        seed,
     )
+
+    votes = [set() for _ in range(num_voters)]
     for v in range(num_voters):
         for c in range(num_candidates):
-            if radius >= np.linalg.norm(voters[v] - candidates[c]):
+            if radius >= np.linalg.norm(voters_pos[v] - candidates_pos[c]):
                 votes[v].add(c)
 
     return votes

prefsampling/core/euclidean.py

@@ -1,112 +1,73 @@
 from __future__ import annotations
 
-from enum import Enum
+from collections.abc import Callable
 
 import numpy as np
 
 from prefsampling.inputvalidators import validate_num_voters_candidates
 
 
-class EuclideanSpace(Enum):
-    """
-    Constants used to represent Euclidean spaces
-    """
-
-    UNIFORM = "Uniform Space"
-    """
-    Uniform space
-    """
-    GAUSSIAN = "Gaussian Space"
-    """
-    Gaussian space
-    """
-    SPHERE = "Spherical Space"
-    """
-    Spherical space
-    """
-    BALL = "Ball Space"
-    """
-    Ball-shaped space
-    """
-
-
 @validate_num_voters_candidates
-def election_positions(
+def sample_election_positions(
     num_voters: int,
     num_candidates: int,
-    space: EuclideanSpace,
-    dimension: int,
-    rng: np.random.Generator,
-) -> (np.ndarray, np.ndarray):
+    point_sampler: Callable,
+    point_sampler_args: dict,
+    candidate_point_sampler: Callable = None,
+    candidate_point_sampler_args: dict = None,
+    seed: int = None,
+) -> tuple[np.ndarray, np.ndarray]:
     """
-    Returns the position of the voters and the candidates in a Euclidean space.
 
     Parameters
     ----------
-    num_voters: int
-        The number of voters.
-    num_candidates: int
-        The number of candidates.
-    space : :py:class:`~prefsampling.core.euclidean.EuclideanSpace`
-        Type of space considered. Should be a constant defined in the
-        :py:class:`~prefsampling.core.euclidean.EuclideanSpace`.
-    dimension : int
-        Number of dimensions for the space considered.
-    rng : np.random.Generator
-        The numpy generator to use for randomness.
+        num_voters : int
+            Number of Voters.
+        num_candidates : int
+            Number of Candidates.
+        point_sampler : Callable
+            The sampler used to sample point in the space. Used for both voters and candidates
+            unless a `candidate_space` is provided.
+        point_sampler_args : dict
+            The arguments passed to the `point_sampler`. The argument `num_points` is ignored
+            and replaced by the number of voters or candidates.
+        candidate_point_sampler : Callable, default: :code:`None`
+            The sampler used to sample the points of the candidates. If a value is provided,
+            then the `space` argument is only used for voters.
+        candidate_point_sampler_args : dict
+            The arguments passed to the `candidate_point_sampler`. The argument `num_points`
+            is ignored and replaced by the number of candidates.
+        seed : int, default: :code:`None`
+            Seed for numpy random number generator. Also passed to the point samplers if
+            a value is provided.
 
     Returns
     -------
-    (np.ndarray, np.ndarray)
-        The position of the voters and of the candidates respectively.
-    """
-    if isinstance(space, Enum):
-        space = EuclideanSpace(space.value)
-    else:
-        space = EuclideanSpace(space)
-    if space == EuclideanSpace.UNIFORM:
-        voters = rng.random((num_voters, dimension))
-        candidates = rng.random((num_candidates, dimension))
-    elif space == EuclideanSpace.GAUSSIAN:
-        voters = rng.normal(loc=0.5, scale=0.15, size=(num_voters, dimension))
-        candidates = rng.normal(loc=0.5, scale=0.15, size=(num_candidates, dimension))
-    elif space == EuclideanSpace.SPHERE:
-        voters = np.array(
-            [list(random_sphere(dimension, rng)[0]) for _ in range(num_voters)]
-        )
-        candidates = np.array(
-            [list(random_sphere(dimension, rng)[0]) for _ in range(num_candidates)]
-        )
-    elif space == EuclideanSpace.BALL:
-        voters = np.array(
-            [list(random_ball(dimension, rng)[0]) for _ in range(num_voters)]
-        )
-        candidates = np.array(
-            [list(random_ball(dimension, rng)[0]) for _ in range(num_candidates)]
-        )
-    else:
-        raise ValueError(
-            "The `space` argument needs to be one of the constant defined in the "
-            "core.euclidean.EuclideanSpace enumeration. Choices are: "
-            + ", ".join(str(s) for s in EuclideanSpace)
-        )
-    return voters, candidates
-
+        tuple[np.ndarray, np.ndarray]
+            The positions of the voters and of the candidates.
 
-def random_ball(
-    dimension: int, rng: np.random.Generator, num_points: int = 1, radius: float = 1
-) -> np.ndarray:
-    random_directions = rng.normal(size=(dimension, num_points))
-    random_directions /= np.linalg.norm(random_directions, axis=0)
-    random_radii = rng.random(num_points) ** (1 / dimension)
-    x = radius * (random_directions * random_radii).T
-    return x
+    """
+    if candidate_point_sampler is not None and candidate_point_sampler_args is None:
+        raise ValueError("If candidate_point_sampler is not None, a value needs to be "
+                         "passed to candidate_point_sampler_args (even if it's just "
+                         "an empty dictionary).")
 
+    if seed is not None:
+        point_sampler_args["seed"] = seed
+        if candidate_point_sampler is not None:
+            candidate_point_sampler_args["seed"] = seed
 
-def random_sphere(
-    dimension: int, rng: np.random.Generator, num_points: int = 1, radius: float = 1
-) -> np.ndarray:
-    random_directions = rng.normal(size=(dimension, num_points))
-    random_directions /= np.linalg.norm(random_directions, axis=0)
-    random_radii = 1.0
-    return radius * (random_directions * random_radii).T
+    point_sampler_args['num_points'] = num_voters
+    voters_pos = point_sampler(**point_sampler_args)
+    dimension = len(voters_pos[0])
+    if candidate_point_sampler is None:
+        point_sampler_args['num_points'] = num_candidates
+        candidates_pos = point_sampler(**point_sampler_args)
+    else:
+        candidate_point_sampler_args['num_points'] = num_candidates
+        candidates_pos = candidate_point_sampler(**candidate_point_sampler_args)
+        if len(candidates_pos[0]) != dimension:
+            raise ValueError("The position of the voters and of the candidates do not have the "
+                             "same dimension. Use different point samplers to solve this "
+                             "problem.")
+    return voters_pos, candidates_pos

prefsampling/ordinal/euclidean.py

@@ -1,18 +1,22 @@
 from __future__ import annotations
 
+from collections.abc import Callable
+
 import numpy as np
 from numpy import linalg
 
-from prefsampling.core.euclidean import election_positions, EuclideanSpace
+from prefsampling.core.euclidean import sample_election_positions
 from prefsampling.inputvalidators import validate_num_voters_candidates
 
 
 @validate_num_voters_candidates
 def euclidean(
     num_voters: int,
     num_candidates: int,
-    space: EuclideanSpace = EuclideanSpace.UNIFORM,
-    dimension: int = 2,
+    point_sampler: Callable,
+    point_sampler_args: dict,
+    candidate_point_sampler: Callable = None,
+    candidate_point_sampler_args: dict = None,
     seed: int = None,
 ) -> np.ndarray:
     """
@@ -24,7 +28,8 @@ def euclidean(
 
     Several Euclidean spaces can be considered. The possibilities are defined in the
     :py:class:`~prefsampling.core.euclidean.EuclideanSpace` enumeration. You can also change the
-    dimension with the parameter :code:`dimension`.
+    dimension with the parameter :code:`dimension`. Note that you can specify different spaces for
+    the voters and for the candidates.
 
     A collection of `num_voters` vote is generated independently and identically following the
     process described above.
@@ -35,31 +40,45 @@ def euclidean(
             Number of Voters.
         num_candidates : int
             Number of Candidates.
-        space : EuclideanSpace, default: :py:class:`~prefsampling.core.euclidean.EuclideanSpace.UNIFORM`
-            Type of space considered. Should be a constant defined in the
-            :py:class:`~prefsampling.core.euclidean.EuclideanSpace` enumeration.
-        dimension : int, default: `2`
-            Number of dimensions for the space considered
+        point_sampler : Callable
+            The sampler used to sample point in the space. Used for both voters and candidates
+            unless a `candidate_space` is provided.
+        point_sampler_args : dict
+            The arguments passed to the `point_sampler`. The argument `num_points` is ignored
+            and replaced by the number of voters or candidates.
+        candidate_point_sampler : Callable, default: :code:`None`
+            The sampler used to sample the points of the candidates. If a value is provided,
+            then the `space` argument is only used for voters.
+        candidate_point_sampler_args : dict
+            The arguments passed to the `candidate_point_sampler`. The argument `num_points`
+            is ignored and replaced by the number of candidates.
         seed : int, default: :code:`None`
-            Seed for numpy random number generator.
+            Seed for numpy random number generator. Also passed to the point samplers if
+            a value is provided.
 
     Returns
     -------
         np.ndarray
             Ordinal votes.
 
     """
-    rng = np.random.default_rng(seed)
-    votes = np.zeros([num_voters, num_candidates], dtype=int)
 
-    voters, candidates = election_positions(
-        num_voters, num_candidates, space, dimension, rng
+    voters_pos, candidates_pos = sample_election_positions(
+        num_voters,
+        num_candidates,
+        point_sampler,
+        point_sampler_args,
+        candidate_point_sampler,
+        candidate_point_sampler_args,
+        seed,
     )
 
+    dimension = len(voters_pos[0])
+    votes = np.zeros([num_voters, num_candidates], dtype=int)
     distances = np.zeros([num_voters, num_candidates], dtype=float)
     for i in range(num_voters):
         for j in range(num_candidates):
-            distances[i][j] = np.linalg.norm(voters[i] - candidates[j], ord=dimension)
+            distances[i][j] = np.linalg.norm(voters_pos[i] - candidates_pos[j], ord=dimension)
         votes[i] = np.argsort(distances[i])
 
     return votes

prefsampling/point/__init__.py

@@ -0,0 +1,11 @@
+from prefsampling.point.ball import ball
+from prefsampling.point.sphere import sphere
+from prefsampling.point.uniform import uniform
+from prefsampling.point.gaussian import gaussian
+
+__all__ = [
+    "uniform",
+    "gaussian",
+    "ball",
+    "sphere"
+]

prefsampling/point/ball.py

@@ -0,0 +1,18 @@
+import numpy as np
+
+from prefsampling.inputvalidators import validate_int
+
+
+def ball(num_points: int, dimension: int, center_point: float = 0.5, width: float = 1, seed: int = None):
+    validate_int(num_points, "num_points", 0)
+    validate_int(dimension, "dimension", 1)
+    rng = np.random.default_rng(seed)
+
+    points = []
+    for _ in range(num_points):
+        random_directions = rng.normal(size=(dimension, num_points))
+        random_directions /= np.linalg.norm(random_directions, axis=0)
+        random_radii = rng.random(num_points) ** (1 / dimension)
+        point = width * (random_directions * random_radii).T
+        points.append(point[0])
+    return np.array(points)

prefsampling/point/gaussian.py

@@ -0,0 +1,10 @@
+import numpy as np
+
+from prefsampling.inputvalidators import validate_int
+
+
+def gaussian(num_points: int, dimension: int, center_point: float = 0.5, width: float = 0.15, seed: int = None):
+    validate_int(num_points, "num_points", 0)
+    validate_int(dimension, "dimension", 1)
+    rng = np.random.default_rng(seed)
+    return rng.normal(loc=center_point, scale=width, size=(num_points, dimension))