feat: convert between trials and df

bbo/algorithms/bo.py

@@ -263,3 +263,10 @@ def suggest(self, count: Optional[int]=None) -> Sequence[Trial]:
 
     def update(self, completed: Sequence[Trial]) -> None:
         self._trials.extend(completed)
+
+    @property
+    def history(self) -> Sequence[Trial]:
+        return self._trials
+
+    def set_history(self, trials: Sequence[Trial]):
+        self._trials = trials

bbo/benchmarks/analyzers/utils.py

@@ -0,0 +1,33 @@
+from typing import Sequence
+
+import pandas as pd
+
+from bbo.utils.trial import Trial, ParameterDict, MetricDict
+from bbo.utils.problem_statement import ProblemStatement
+
+
+def trials2df(trials: Sequence[Trial]):
+    df = dict()
+    for name in trials[0].parameters:
+        values = [t.parameters[name].value for t in trials]
+        df[name] = values
+    for obj_name in trials[0].metrics:
+        obj_values = [t.metrics[obj_name].value for t in trials]
+        df[obj_name] = obj_values
+    df = pd.DataFrame(df)
+    return df
+
+def df2trials(df: pd.DataFrame, problem_statement: ProblemStatement):
+    sp = problem_statement.search_space
+    obj = problem_statement.objective
+    parameter_dict = [ParameterDict() for _ in range(len(df))]
+    metric_dict = [MetricDict() for _ in range(len(df))]
+    for name in sp.parameter_configs:
+        for i, v in enumerate(df[name]):
+            parameter_dict[i][name] = v
+    for name in obj.metric_informations:
+        for i, v in enumerate(df[name]):
+            metric_dict[i][name] = v
+
+    trials = [Trial(parameters=p, metrics=m) for p, m in zip(parameter_dict, metric_dict)]
+    return trials

tests/benchmarks/utils_test.py

@@ -0,0 +1,35 @@
+import unittest
+import random
+
+from bbo.utils.parameter_config import SearchSpace, ScaleType
+from bbo.utils.metric_config import Objective, ObjectiveMetricGoal
+from bbo.utils.problem_statement import ProblemStatement
+from bbo.utils.trial import Trial, MetricDict
+from bbo.benchmarks.analyzers.utils import trials2df, df2trials
+
+
+class UtilsTest(unittest.TestCase):
+    def setUp(self):
+        sp = SearchSpace()
+        sp.add_float_param('float', 0, 10)
+        sp.add_float_param('float_linear', 0, 10, scale_type=ScaleType.LINEAR)
+        sp.add_int_param('int', 1, 10)
+        sp.add_discrete_param('discrete', [0, 2, 4, 6])
+        sp.add_categorical_param('categorical', ['a', 'b', 'c'])
+        obj = Objective()
+        obj.add_metric('obj1', ObjectiveMetricGoal.MAXIMIZE)
+        obj.add_metric('obj2', ObjectiveMetricGoal.MAXIMIZE)
+        self.problem_statement = ProblemStatement(sp, obj)
+        n = 10
+        self.trials = [Trial(sp.sample()) for _ in range(n)]
+        metrics = [MetricDict({
+            'obj1': i + random.uniform(-0.5, 0.5),
+            'obj2': -i + random.uniform(-0.5, 0.5)
+        }) for i in range(n)]
+        for t, m in zip(self.trials, metrics):
+            t.complete(m)
+
+    def test_trials2df(self):
+        df = trials2df(self.trials)
+        trials = df2trials(df, self.problem_statement)
+        self.assertEqual(self.trials, trials)