Use scipy for model free designs

emukit/core/initial_designs/latin_design.py

@@ -6,11 +6,7 @@
 
 
 import numpy as np
-
-try:
-    import pyDOE
-except ImportError:
-    raise ImportError("pyDOE needs to be installed in order to use latin design")
+from scipy.stats import qmc
 
 from .. import ParameterSpace
 from .base import InitialDesignBase
@@ -20,7 +16,8 @@ class LatinDesign(InitialDesignBase):
     """
     Latin hypercube experiment design.
 
-    Based on pyDOE implementation. For further reference see https://pythonhosted.org/pyDOE/randomized.html#latin-hypercube
+    Based on scipy implementation. For further reference see
+    https://docs.scipy.org/doc/scipy/reference/generated/scipy.stats.qmc.LatinHypercube.html
     """
 
     def __init__(self, parameter_space: ParameterSpace) -> None:
@@ -37,15 +34,14 @@ def get_samples(self, point_count: int) -> np.ndarray:
         :return: A numpy array of generated samples, shape (point_count x space_dim)
         """
         bounds = self.parameter_space.get_bounds()
-        X_design_aux = pyDOE.lhs(len(bounds), point_count, criterion="center")
-        ones = np.ones((X_design_aux.shape[0], 1))
-
-        lower_bound = np.asarray(bounds)[:, 0].reshape(1, len(bounds))
-        upper_bound = np.asarray(bounds)[:, 1].reshape(1, len(bounds))
-        diff = upper_bound - lower_bound
+        d = len(bounds)
+        lower_bounds = [x[0] for x in bounds]
+        upper_bounds = [x[1] for x in bounds]
 
-        X_design = np.dot(ones, lower_bound) + X_design_aux * np.dot(ones, diff)
+        sampler = qmc.LatinHypercube(d)
+        samples = sampler.random(n=point_count)
+        samples = qmc.scale(samples, lower_bounds, upper_bounds)
 
-        samples = self.parameter_space.round(X_design)
+        X_design = self.parameter_space.round(samples)
 
-        return samples
+        return X_design

emukit/core/initial_designs/sobol_design.py

@@ -1,20 +1,12 @@
-# Copyright 2020-2024 The Emukit Authors. All Rights Reserved.
+# Copyright 2020-2026 The Emukit Authors. All Rights Reserved.
 # SPDX-License-Identifier: Apache-2.0
 
 # Copyright 2018-2020 Amazon.com, Inc. or its affiliates. All Rights Reserved.
 # SPDX-License-Identifier: Apache-2.0
 
 
-# Copyright 2020 Amazon.com, Inc. or its affiliates. All Rights Reserved.
-# SPDX-License-Identifier: Apache-2.0
-
-
 import numpy as np
-
-try:
-    from sobol_seq import i4_sobol_generate
-except ImportError:
-    raise ImportError("sobol_seq needs to be installed in order to use sobol design")
+from scipy.stats import qmc
 
 from .. import ParameterSpace
 from .base import InitialDesignBase
@@ -23,7 +15,8 @@
 class SobolDesign(InitialDesignBase):
     """
     Sobol experiment design.
-    Based on sobol_seq implementation. For further reference see https://github.com/naught101/sobol_seq
+    Based on scipy implementation. For further reference see
+    https://docs.scipy.org/doc/scipy/reference/generated/scipy.stats.qmc.Sobol.html
     """
 
     def __init__(self, parameter_space: ParameterSpace) -> None:
@@ -40,12 +33,14 @@ def get_samples(self, point_count: int) -> np.ndarray:
         :return: A numpy array of generated samples, shape (point_count x space_dim)
         """
         bounds = self.parameter_space.get_bounds()
-        lower_bound = np.asarray(bounds)[:, 0].reshape(1, len(bounds))
-        upper_bound = np.asarray(bounds)[:, 1].reshape(1, len(bounds))
-        diff = upper_bound - lower_bound
+        d = len(bounds)
+        lower_bounds = [x[0] for x in bounds]
+        upper_bounds = [x[1] for x in bounds]
 
-        X_design = np.dot(i4_sobol_generate(len(bounds), point_count), np.diag(diff[0, :])) + lower_bound
+        sampler = qmc.Sobol(d)
+        samples = sampler.random(n=point_count)
+        samples = qmc.scale(samples, lower_bounds, upper_bounds)
 
-        samples = self.parameter_space.round(X_design)
+        X_design = self.parameter_space.round(samples)
 
-        return samples
+        return X_design

pyproject.toml

@@ -31,8 +31,6 @@ dependencies = [
   "scipy",
   "matplotlib>=3.9",
   "emcee>=2.2.1",
-  "pydoe<0.9.6",
-  "sobol_seq>=0.1.2",
 ]
 
 [project.optional-dependencies]

requirements/integration_test_requirements.txt

@@ -2,8 +2,6 @@
 scikit-learn
 # For BNN model
 pybnn==0.0.5
-# For Latin design
-PyDOE>=0.3.0
 # For notebook tests
 jupyter==1.0.0
 pandas>=1.0.5

requirements/requirements.txt

@@ -7,5 +7,3 @@ matplotlib>=3.9
 # GPy>=1.13.0
 emcee>=2.2.1
 scipy
-PyDOE>=0.3.0
-sobol_seq>=0.1.2

tests/emukit/core/test_model_free_designs.py

@@ -1,9 +1,10 @@
-# Copyright 2020-2024 The Emukit Authors. All Rights Reserved.
+# Copyright 2020-2026 The Emukit Authors. All Rights Reserved.
 # SPDX-License-Identifier: Apache-2.0
 
 # Copyright 2018-2020 Amazon.com, Inc. or its affiliates. All Rights Reserved.
 # SPDX-License-Identifier: Apache-2.0
 
+import numpy as np
 
 from emukit.core import CategoricalParameter, ContinuousParameter, DiscreteParameter, ParameterSpace
 from emukit.core.initial_designs import RandomDesign
@@ -25,8 +26,22 @@ def test_design_returns_correct_number_of_points():
         points = design.get_samples(points_count)
 
         assert points_count == len(points)
-        columns_count = 1
-        assert all([len(p) == columns_count for p in points])
+        assert all([len(p) == 1 for p in points])
+
+
+def test_design_returns_points_within_bounds():
+    p1 = ContinuousParameter("p1", 0.01, 0.05)
+    p2 = ContinuousParameter("p2", -100.0, -90.0)
+    space = ParameterSpace([p1, p2])
+    points_count = 5
+
+    designs = create_model_free_designs(space)
+    for design in designs:
+        points = design.get_samples(points_count)
+
+        for i, p in enumerate(space.parameters):
+            assert np.all(p.min <= points[:, i])
+            assert np.all(points[:, i] <= p.max)
 
 
 def test_design_with_mixed_domain(encoding):