From 6eebbcdac09f6bd6356e46823d6115cc00e1efdf Mon Sep 17 00:00:00 2001
From: Khurram Ghani <113982802+khurram-ghani@users.noreply.github.com>
Date: Tue, 28 May 2024 11:08:18 +0100
Subject: [PATCH] Don't take TR ymin from init dataset (#843)

---
 tests/unit/acquisition/test_rule.py | 15 ++++++++-------
 trieste/acquisition/rule.py         |  6 ++++--
 2 files changed, 12 insertions(+), 9 deletions(-)

diff --git a/tests/unit/acquisition/test_rule.py b/tests/unit/acquisition/test_rule.py
index b44a56780b..aed49595e9 100644
--- a/tests/unit/acquisition/test_rule.py
+++ b/tests/unit/acquisition/test_rule.py
@@ -620,7 +620,7 @@ def test_trego_for_default_state(
     npt.assert_array_almost_equal(ret_subspace.lower, lower_bound)
     npt.assert_array_almost_equal(ret_subspace.upper, upper_bound)
     npt.assert_array_almost_equal(query_point, [expected_query_point], 5)
-    npt.assert_array_almost_equal(ret_subspace._y_min, [0.012])
+    npt.assert_array_almost_equal(ret_subspace._y_min, [np.inf])
     assert ret_subspace._is_global
 
 
@@ -1684,9 +1684,9 @@ def test_multi_trust_region_box_inits_regions_that_need_it() -> None:
 
     # Change all eps values, with the second region have a lower eps than the min. This region
     # should be re-initialized.
-    subspaces[0].eps = 0.35
+    subspaces[0].eps = 0.45
     subspaces[1].eps = 0.25
-    subspaces[2].eps = 0.32
+    subspaces[2].eps = 0.42
 
     # Check the property values.
     assert bool(subspaces[0].requires_initialization) is False
@@ -1698,9 +1698,9 @@ def test_multi_trust_region_box_inits_regions_that_need_it() -> None:
 
     # Check that the second region was re-initialized.
     assert state is not None
-    assert cast(TestTrustRegionBox, state.subspaces[0]).eps < 0.35  # Expect reduction.
+    assert cast(TestTrustRegionBox, state.subspaces[0]).eps > 0.45  # Expect increase, step success.
     assert cast(TestTrustRegionBox, state.subspaces[1]).eps == 0.4  # Expect re-initialized value.
-    assert cast(TestTrustRegionBox, state.subspaces[2]).eps < 0.32  # Expect reduction.
+    assert cast(TestTrustRegionBox, state.subspaces[0]).eps > 0.42  # Expect increase, step success.
 
 
 def test_multi_trust_region_box_acquire_with_state() -> None:
@@ -1742,14 +1742,15 @@ def test_multi_trust_region_box_acquire_with_state() -> None:
 
     assert next_state is not None
     assert points.shape == [1, 3, 2]
-    # The regions correspond to first, third and first points in the dataset.
+    # The regions correspond to first, third and first points in the dataset. However, for the
+    #   region that is initialized, the point is not used and value is set to infinity.
     # First two regions should be updated.
     # The third region should be initialized and not updated, as it is too close to the first
     # subspace.
     for point, subspace, exp_obs, exp_eps in zip(
         points[0],
         cast(Sequence[TestTrustRegionBox], next_state.subspaces),
-        [dataset.observations[0], dataset.observations[2], dataset.observations[0]],
+        [dataset.observations[0], dataset.observations[2], np.inf],
         [0.1, 0.1, 0.07],  # First two regions updated, third region initialized.
     ):
         assert point in subspace
diff --git a/trieste/acquisition/rule.py b/trieste/acquisition/rule.py
index 1335a8013b..d9a6c68e48 100644
--- a/trieste/acquisition/rule.py
+++ b/trieste/acquisition/rule.py
@@ -1621,7 +1621,7 @@ def __init__(
         self._update_domain()
         # Initial value of the region minimum is set to infinity as we have not yet observed any
         # data.
-        self._y_min = np.inf
+        self._y_min = tf.constant(np.inf, dtype=self.location.dtype)
 
     def _init_eps(self) -> None:
         self.eps = self._zeta * (self.global_search_space.upper - self.global_search_space.lower)
@@ -1665,7 +1665,9 @@ def initialize(
         self._step_is_success = False
         self._init_eps()
         self._update_domain()
-        _, self._y_min = self.get_dataset_min(datasets)
+        # We haven't necessarily observed any data yet for this region; force first step to always
+        # be successful by setting the minimum to infinity.
+        self._y_min = tf.constant(np.inf, dtype=self.location.dtype)
         self._initialized = True
 
     def update(