Allow loading of logs with constraint

Author: till-m

bayes_opt/bayesian_optimization.py

@@ -153,9 +153,9 @@ def max(self):
     def res(self):
         return self._space.res()
 
-    def register(self, params, target):
+    def register(self, params, target, constraint_value=None):
         """Expect observation with known target"""
-        self._space.register(params, target)
+        self._space.register(params, target, constraint_value)
         self.dispatch(Events.OPTIMIZATION_STEP)
 
     def probe(self, params, lazy=True):

bayes_opt/target_space.py

@@ -183,10 +183,14 @@ def register(self, params, target, constraint_value=None):
         self._params = np.concatenate([self._params, x.reshape(1, -1)])
         self._target = np.concatenate([self._target, [target]])
 
-        if constraint_value is None:
+        if self._constraint is None:
             # Insert data into unique dictionary
             self._cache[_hashable(x.ravel())] = target
         else:
+            if constraint_value is None:
+                msg = ("When registering a point to a constrained TargetSpace" +
+                    " a constraint value needs to be present.")
+                raise ValueError(msg)
             # Insert data into unique dictionary
             self._cache[_hashable(x.ravel())] = (target, constraint_value)
             self._constraint_values = np.concatenate([self._constraint_values,

bayes_opt/util.py

@@ -176,13 +176,11 @@ def load_logs(optimizer, logs):
                     break
 
                 iteration = json.loads(iteration)
-                try:
-                    optimizer.register(
-                        params=iteration["params"],
-                        target=iteration["target"],
-                    )
-                except KeyError:
-                    pass
+                optimizer.register(
+                    params=iteration["params"],
+                    target=iteration["target"],
+                    constraint_value=iteration["constraint"] if optimizer.is_constrained else None
+                )
 
     return optimizer
 

tests/test_logs_constrained.json

@@ -0,0 +1,7 @@
+{"target": -0.44299100812007164, "constraint": 1.9671294186934094, "params": {"x": -1.1774463661021484, "y": 0.7620692052866689}, "allowed": true, "datetime": {"datetime": "2022-10-13 11:04:33", "elapsed": 0.0, "delta": 0.0}}
+{"target": -3.80499987207446, "constraint": 1.7401939200536138, "params": {"x": -0.8211808663007414, "y": -1.032402976010423}, "allowed": true, "datetime": {"datetime": "2022-10-13 11:04:33", "elapsed": 0.000253, "delta": 0.000253}}
+{"target": 0.6173973697366524, "constraint": 1.4074565522579112, "params": {"x": -0.6184239653374695, "y": 1.0124269609972818}, "allowed": true, "datetime": {"datetime": "2022-10-13 11:04:33", "elapsed": 0.289497, "delta": 0.289244}}
+{"target": -0.0044260558252411375, "constraint": 4.004426055825241, "params": {"x": 0.06652860907339889, "y": 2.0}, "allowed": false, "datetime": {"datetime": "2022-10-13 11:04:33", "elapsed": 0.644552, "delta": 0.355055}}
+{"target": -0.48387082383442337, "constraint": 2.0734433009323867, "params": {"x": 1.2007323331806918, "y": 0.7947862385489817}, "allowed": true, "datetime": {"datetime": "2022-10-13 11:04:34", "elapsed": 1.040863, "delta": 0.396311}}
+{"target": -1.7071222601388523, "constraint": 5.694318845373349, "params": {"x": -1.3114437437598845, "y": 1.9935982926172482}, "allowed": false, "datetime": {"datetime": "2022-10-13 11:04:34", "elapsed": 1.360582, "delta": 0.319719}}
+{"target": 0.9685285807415294, "constraint": 0.9465681677055823, "params": {"x": 0.17224772488310958, "y": 0.9575483742235558}, "allowed": true, "datetime": {"datetime": "2022-10-13 11:04:35", "elapsed": 1.816862, "delta": 0.45628}}

tests/test_util.py

@@ -8,6 +8,7 @@
 from sklearn.gaussian_process.kernels import Matern
 from sklearn.gaussian_process import GaussianProcessRegressor
 
+from scipy.optimize import NonlinearConstraint
 
 def get_globals():
     X = np.array([
@@ -146,6 +147,30 @@ def f(x, y):
         load_logs(other_optimizer, ["./tests/test_logs.json"])
 
 
+def test_logs_constraint():
+    import pytest
+    def f(x, y):
+        return -x ** 2 - (y - 1) ** 2 + 1
+
+    def c(x, y):
+        return x ** 2 + y ** 2
+    
+    constraint = NonlinearConstraint(c, -np.inf, 3)
+
+    optimizer = BayesianOptimization(
+        f=f,
+        pbounds={"x": (-2, 2), "y": (-2, 2)},
+        constraint=constraint
+    )
+
+    with pytest.raises(KeyError):
+        load_logs(optimizer, ["./tests/test_logs.json"])
+    
+    load_logs(optimizer, ["./tests/test_logs_constrained.json"])
+
+    assert len(optimizer.space) == 7
+
+
 def test_colours():
     colour_wrappers = [
         (Colours.END, Colours.black),