Cost bo loop

emukit/core/loop/loop_state.py

@@ -59,6 +59,7 @@ def create_loop_state(x_init: np.ndarray, y_init: np.ndarray, cost: np.ndarray =
 
     :param x_init: x values for initial function evaluations.
     :param y_init: y values for initial function evaluations
+    :param cost: observed costs for initial function evaluations
     """
     if x_init.shape[0] != y_init.shape[0]:
         error_message = "X and Y should have the same length. Actual length x_init {}, y_init {}".format(

emukit/examples/gp_bayesian_optimization/optimization_loops.py

@@ -14,13 +14,15 @@
 
 def create_bayesian_optimization_loop(x_init: np.ndarray, y_init: np.ndarray, parameter_space: ParameterSpace,
                                       acquisition_type: AcquisitionType, model_type: ModelType,
+                                      cost_init: np.ndarray = None,
                                       model_kwargs: dict=None,
                                       acquisition_optimizer: AcquisitionOptimizerBase = None) -> OuterLoop:
     """
     Creates Bayesian optimization loop for Bayesian neural network or random forest models.
 
     :param x_init: 2d numpy array of shape (no. points x no. input features) of initial X data
     :param y_init: 2d numpy array of shape (no. points x no. targets) of initial Y data
+    :param cost_init: 2d numpy array of shape (no. points x no. targets) of initial cost of each function evaluation
     :param parameter_space: parameter space describing input domain
     :param acquisition_type: an AcquisitionType enumeration object describing which acquisition function to use
     :param model_type: A ModelType enumeration object describing which model to use.
@@ -54,6 +56,6 @@ def create_bayesian_optimization_loop(x_init: np.ndarray, y_init: np.ndarray, pa
     if acquisition_optimizer is None:
         acquisition_optimizer = GradientAcquisitionOptimizer(parameter_space)
     candidate_point_calculator = SequentialPointCalculator(acquisition, acquisition_optimizer)
-    loop_state = create_loop_state(x_init, y_init)
+    loop_state = create_loop_state(x_init, y_init, cost_init)
     model_updater = FixedIntervalUpdater(model, 1)
     return OuterLoop(candidate_point_calculator, model_updater, loop_state)

integration_tests/emukit/bayesian_optimization/test_create_bayesian_optimization_loop.py

@@ -0,0 +1,22 @@
+import numpy as np
+
+from emukit.test_functions.forrester import forrester_function
+from emukit.examples.gp_bayesian_optimization.optimization_loops import create_bayesian_optimization_loop
+from emukit.examples.gp_bayesian_optimization.enums import AcquisitionType, ModelType
+
+
+def test_loop_state():
+
+    fcn, cs = forrester_function()
+    n_init = 5
+
+    x_init = np.random.rand(n_init, 1)
+    y_init = np.random.rand(n_init, 1)
+    c_init = np.random.rand(n_init, 1)
+    bo = create_bayesian_optimization_loop(x_init=x_init, y_init=y_init,
+                                           parameter_space=cs, acquisition_type=AcquisitionType.EI,
+                                           model_type=ModelType.RandomForest, cost_init=c_init)
+
+    assert bo.loop_state.X.shape[0] == n_init
+    assert bo.loop_state.Y.shape[0] == n_init
+    assert bo.loop_state.cost.shape[0] == n_init