Merge pull request #151 from neurolib-dev/fix/exploration/save_lists_…

…in_results

Support lists as exploration and evolution outputs

neurolib/optimize/evolution/evolution.py

@@ -147,6 +147,7 @@ def __init__(
 
         # -------- settings
         self.verbose = False
+        self.verbose_plotting = True
         self.plotColor = "C0"
 
         # -------- simulation
@@ -235,7 +236,7 @@ def __init__(
         self.evaluationCounter = 0
         self.last_id = 0
 
-    def run(self, verbose=False):
+    def run(self, verbose=False, verbose_plotting=True):
         """Run the evolution or continue previous evolution. If evolution was not initialized first
         using `runInitial()`, this will be done.
 
@@ -244,6 +245,7 @@ def run(self, verbose=False):
         """
 
         self.verbose = verbose
+        self.verbose_plotting = verbose_plotting
         if not self._initialPopulationSimulated:
             self.runInitial()
 
@@ -650,7 +652,7 @@ def runEvolution(self):
 
             # verbose output
             if self.verbose:
-                self.info(plot=True, info=True)
+                self.info(plot=self.verbose_plotting, info=True)
 
         logging.info("--- End of evolution ---")
         logging.info("Best individual is %s, %s" % (self.best_ind, self.best_ind.fitness.values))
@@ -710,7 +712,7 @@ def info(self, plot=True, bestN=5, info=True, reverse=False):
         print(f"Best {bestN} individuals:")
         eu.printIndividuals(self.toolbox.selBest(self.pop, bestN), self.paramInterval)
         print("--------------------")
-        # Plotting
+        # Plotting evolutionary progress
         if plot:
             # hack: during the evolution we need to use reverse=True
             # after the evolution (with evolution.info()), we need False
@@ -785,22 +787,27 @@ def _outputToDf(self, pop, df):
         :return: Dataframe with outputs
         :rtype: pandas.core.frame.DataFrame
         """
+        # defines which variable types will be saved in the results dataframe
+        SUPPORTED_TYPES = (float, int, np.ndarray, list)
+        SCALAR_TYPES = (float, int)
+        ARRAY_TYPES = (np.ndarray, list)
+
         assert len(pop) == len(df), "Dataframe and population do not have same length."
         nan_value = np.nan
         # load outputs into dataframe
         for i, p in enumerate(pop):
             if hasattr(p, "outputs"):
                 for key, value in p.outputs.items():
                     # only save floats, ints and arrays
-                    if isinstance(value, (float, int, np.ndarray)):
+                    if isinstance(value, SUPPORTED_TYPES):
                         # save 1-dim arrays
-                        if isinstance(value, np.ndarray):
+                        if isinstance(value, ARRAY_TYPES):
                             # to save a numpy array, convert column to object type
                             if key not in df:
                                 df[key] = None
                             df[key] = df[key].astype(object)
                             df.at[i, key] = value
-                        elif isinstance(value, (float, int)):
+                        elif isinstance(value, SCALAR_TYPES):
                             # save numbers
                             df.loc[i, key] = value
                     else:

neurolib/optimize/exploration/exploration.py

@@ -352,9 +352,12 @@ def aggregateResultsToDfResults(self, arrays=True, fillna=False):
         :type fillna: bool, optional
         """
         nan_value = np.nan
-        logging.info("Aggregating results to `dfResults` ...")
-        # for i, result in tqdm.tqdm(self.results.items()):
+        # defines which variable types will be saved in the results dataframe
+        SUPPORTED_TYPES = (float, int, np.ndarray, list)
+        SCALAR_TYPES = (float, int)
+        ARRAY_TYPES = (np.ndarray, list)
 
+        logging.info("Aggregating results to `dfResults` ...")
         for runId, parameters in tqdm.tqdm(self.dfResults.iterrows(), total=len(self.dfResults)):
             # if the results were previously loaded into memory, use them
             if hasattr(self, "results"):
@@ -370,16 +373,16 @@ def aggregateResultsToDfResults(self, arrays=True, fillna=False):
 
             for key, value in result.items():
                 # only save floats, ints and arrays
-                if isinstance(value, (float, int, np.ndarray)):
+                if isinstance(value, SUPPORTED_TYPES):
                     # save 1-dim arrays
-                    if isinstance(value, np.ndarray) and arrays:
+                    if isinstance(value, ARRAY_TYPES) and arrays:
                         # to save a numpy array, convert column to object type
                         if key not in self.dfResults:
                             self.dfResults[key] = None
                         self.dfResults[key] = self.dfResults[key].astype(object)
                         self.dfResults.at[runId, key] = value
-                    elif isinstance(value, (float, int)):
-                        # save numbers
+                    elif isinstance(value, SCALAR_TYPES):
+                        # save scalars
                         self.dfResults.loc[runId, key] = value
                 else:
                     self.dfResults.loc[runId, key] = nan_value

setup.py

@@ -11,7 +11,7 @@
 
 setuptools.setup(
     name="neurolib",
-    version="0.5.13",
+    version="0.5.14",
     description="Easy whole-brain neural mass modeling",
     long_description=long_description,
     long_description_content_type="text/markdown",

tests/test_exploration.py

@@ -9,6 +9,7 @@
 import neurolib.utils.pypetUtils as pu
 import numpy as np
 import xarray as xr
+
 from neurolib.models.aln import ALNModel
 from neurolib.models.fhn import FHNModel
 from neurolib.models.multimodel import MultiModel
@@ -151,19 +152,23 @@ def explore_me(traj):
             pars = search.getParametersFromTraj(traj)
             # let's calculate the distance to a circle
             computation_result = abs((pars["x"] ** 2 + pars["y"] ** 2) - 1)
-            result_dict = {"distance": computation_result}
+            result_dict = {"scalar_result": computation_result, "list_result": [1, 2, 3, 4], "array_result": np.ones(3)}
             search.saveToPypet(result_dict, traj)
 
         parameters = ParameterSpace({"x": np.linspace(-2, 2, 2), "y": np.linspace(-2, 2, 2)})
         search = BoxSearch(evalFunction=explore_me, parameterSpace=parameters, filename="test_circle_exploration.hdf")
         search.run()
         search.loadResults(pypetShortNames=False)
 
-        for i in search.dfResults.index:
-            search.dfResults.loc[i, "distance"] = search.results[i]["distance"]
-
+        # call the result dataframe
         search.dfResults
 
+        # test integrity of dataframe
+        for i in search.dfResults.index:
+            self.assertEqual(search.dfResults.loc[i, "scalar_result"], search.results[i]["scalar_result"])
+            self.assertListEqual(search.dfResults.loc[i, "list_result"], search.results[i]["list_result"])
+            np.testing.assert_array_equal(search.dfResults.loc[i, "array_result"], search.results[i]["array_result"])
+
 
 class TestExplorationMultiModel(unittest.TestCase):
     """