Revert "Add functionality to Region class (#444)"

This reverts commit 0dcbc52.

src/estimagic/optimization/tranquilo/estimate_variance.py

@@ -5,7 +5,7 @@
 
 from estimagic.optimization.tranquilo.get_component import get_component
 from estimagic.optimization.tranquilo.new_history import History
-from estimagic.optimization.tranquilo.region import Region
+from estimagic.optimization.tranquilo.options import Region
 
 
 def get_variance_estimator(fitter, user_options):

src/estimagic/optimization/tranquilo/geometry.py

@@ -2,8 +2,12 @@
 
 import numpy as np
 
-from estimagic.optimization.tranquilo.region import Region
-from estimagic.optimization.tranquilo.sample_points import get_sampler
+from estimagic.optimization.tranquilo.options import Region
+from estimagic.optimization.tranquilo.sample_points import (
+    _get_effective_bounds,
+    _map_from_feasible_trustregion,
+    get_sampler,
+)
 
 
 def get_geometry_checker_pair(
@@ -22,7 +26,7 @@ def get_geometry_checker_pair(
             samples drawn inside a box or a ball, respectively.
         n_params (int): Number of parameters.
         n_simulations (int): Number of simulations for the mean calculation.
-        bounds (Bounds): The parameter bounds. See module bounds.py.
+        bounds (Bounds): The parameter bounds.
 
     Returns:
         callable: The sample quality calculator.
@@ -41,7 +45,7 @@ def get_geometry_checker_pair(
 
     _checker = built_in_checker[checker]
 
-    quality_calculator = _checker["quality_calculator"]
+    quality_calculator = partial(_checker["quality_calculator"], bounds=bounds)
     cutoff_simulator = partial(
         _checker["cutoff_simulator"],
         reference_sampler=reference_sampler,
@@ -62,26 +66,26 @@ def log_d_cutoff_simulator(
         rng (np.random.Generator): The random number generator.
         reference_sampler (str): Either "box" or "ball", corresponding to comparison
             samples drawn inside a box or a ball, respectively.
-        bounds (Bounds): The parameter bounds. See module bounds.py.
+        bounds (Bounds): The parameter bounds.
         n_params (int): Dimensionality of the sample.
         n_simulations (int): Number of simulations for the mean calculation.
 
     Returns:
         float: The simulated mean logarithm of the d-optimality criterion.
 
     """
-    _sampler = get_sampler(reference_sampler)
-    trustregion = Region(center=np.zeros(n_params), radius=1.0, bounds=bounds)
+    _sampler = get_sampler(reference_sampler, bounds)
+    trustregion = Region(center=np.zeros(n_params), radius=1, shape=None)
     sampler = partial(_sampler, trustregion=trustregion)
     raw = []
     for _ in range(n_simulations):
         x = sampler(n_points=n_samples, rng=rng)
-        raw.append(log_d_quality_calculator(x, trustregion))
+        raw.append(log_d_quality_calculator(x, trustregion, bounds))
     out = np.nanmean(raw)
     return out
 
 
-def log_d_quality_calculator(sample, trustregion):
+def log_d_quality_calculator(sample, trustregion, bounds):
     """Logarithm of the d-optimality criterion.
 
     For a data sample x the log_d_criterion is defined as log(det(x.T @ x)). If the
@@ -90,13 +94,15 @@ def log_d_quality_calculator(sample, trustregion):
 
     Args:
         sample (np.ndarray): The data sample, shape = (n, p).
-        trustregion (Region): Trustregion. See module region.py.
+        trustregion (TrustRegion): NamedTuple with attributes center and radius.
+        bounds (Bounds): The parameter bounds.
 
     Returns:
         np.ndarray: The criterion values, shape = (n, ).
 
     """
-    points = trustregion.map_to_unit(sample)
+    effective_bounds = _get_effective_bounds(trustregion, bounds)
+    points = _map_from_feasible_trustregion(sample, effective_bounds)
     n_samples, n_params = points.shape
     xtx = points.T @ points
     det = np.linalg.det(xtx / n_samples)

src/estimagic/optimization/tranquilo/models.py

@@ -4,7 +4,7 @@
 import numpy as np
 from numba import njit
 
-from estimagic.optimization.tranquilo.region import Region
+from estimagic.optimization.tranquilo.options import Region
 
 
 @dataclass

src/estimagic/optimization/tranquilo/options.py

@@ -1,3 +1,4 @@
+from dataclasses import dataclass
 from typing import NamedTuple
 
 import numpy as np
@@ -8,6 +9,17 @@
 )
 
 
+@dataclass
+class Bounds:
+    """Stopping criteria."""
+
+    lower: np.ndarray
+    upper: np.ndarray
+
+    def __post_init__(self):
+        self.has_any = _check_if_there_are_bounds(self.lower, self.upper)
+
+
 class StopOptions(NamedTuple):
     """Criteria for stopping without successful convergence."""
 
@@ -42,6 +54,12 @@ class RadiusOptions(NamedTuple):
     max_radius_to_step_ratio: float = np.inf
 
 
+class Region(NamedTuple):
+    center: np.ndarray
+    radius: float
+    shape: str
+
+
 class AcceptanceOptions(NamedTuple):
     confidence_level: float = 0.8
     power_level: float = 0.8
@@ -58,3 +76,12 @@ class StagnationOptions(NamedTuple):
     sample_increment: int = 1
     max_trials: int = 1
     drop: bool = True
+
+
+def _check_if_there_are_bounds(lb, ub):
+    out = False
+    if lb is not None and np.isfinite(lb).any():
+        out = True
+    if ub is not None and np.isfinite(ub).any():
+        out = True
+    return out