Merge pull request #394 from smattis/v3-steve

Updates from review

.travis.yml

@@ -16,6 +16,7 @@ install:
   - pip install .
   - pip install codecov pytest-cov Sphinx sphinx_rtd_theme
   - pip install git+https://github.com/CU-Denver-UQ/LUQ
+  - pip install mpi4py
 
 script:
   - pytest --cov=./bet/ ./test/

README.md

@@ -15,7 +15,10 @@ Another option is to clone the repository and install BET using
 
 
 ## Dependencies
-BET is tested on Python 3.6 and 3.7 (but should work on most recent Python 3 versions) and depends on [NumPy](http://www.numpy.org/), [SciPy](http://www.scipy.org/), [matplotlib](http://matplotlib.org/), [pyDOE](https://pythonhosted.org/pyDOE/), [pytest](https://docs.pytest.org/), and [mpi4py](https://mpi4py.readthedocs.io/en/stable/) (optional) (see [requirements.txt](requirements.txt) for version information). For some optional features [LUQ](https://github.com/CU-Denver-UQ/LUQ) is also required.
+BET is tested on Python 3.6 and 3.7 (but should work on most recent Python 3 versions) and depends on [NumPy](http://www.numpy.org/), [SciPy](http://www.scipy.org/), [matplotlib](http://matplotlib.org/), [pyDOE](https://pythonhosted.org/pyDOE/), [pytest](https://docs.pytest.org/), and [mpi4py](https://mpi4py.readthedocs.io/en/stable/) (optional) (see [requirements.txt](requirements.txt) for version information). For some optional features [LUQ](https://github.com/CU-Denver-UQ/LUQ) is also required. mpi4py is required to take advantage of parallel features and requires an mpi implementation. It can be installed by:
+
+    pip install mpi4py
+
 
 ## License
 [GNU Lesser General Public License (LGPL)](LICENSE.txt)

bet/sampling/useLUQ.py

@@ -39,7 +39,7 @@ class useLUQ:
     from LUQ output.
     """
 
-    def __init__(self, predict_set, obs_set, lb_model, times):
+    def __init__(self, predict_set, lb_model, times, obs_set=None):
         """
         Initialize the object.
         :param predict_set: Sample set defining input prediction samples.
@@ -80,7 +80,16 @@ def get_obs(self):
         """
         self.obs_time_series = self.lb_model(self.obs_set.get_values(), self.times)
 
-    def initialize(self, predicted_time_series, obs_time_series, times):
+    def set_observed_time_series(self, obs_time_series):
+        """
+        Set observed time series data manually.
+        :param obs_time_series: time series data
+        :type obs_time_series:
+        :return: :class:`numpy.ndarray` with shape (num_obs, num_times)
+        """
+        self.obs_time_series = obs_time_series
+
+    def initialize(self, predicted_time_series=None, obs_time_series=None, times=None):
         """
         Initialize the LUQ object. This can be used manually if time series are pre-computed.
 
@@ -96,6 +105,13 @@ def initialize(self, predicted_time_series, obs_time_series, times):
         except ImportError:
             raise missing_module("luq cannot be imported")
 
+        if predicted_time_series is None:
+            predicted_time_series = self.predicted_time_series
+        if obs_time_series is None:
+            obs_time_series = self.obs_time_series
+        if times is None:
+            times = self.times
+
         self.learn = LUQ(predicted_time_series, obs_time_series, times)
 
     def setup(self):
@@ -128,7 +144,7 @@ def make_disc(self):
         """
         Construct `bet.sample.discretization` objects for predict and obs sets.
         :return: predict_disc, obs_disc
-        :rtype: `bet.sample.discretization`, `bet.sample.discretization`
+        :rtype: `bet.sample.discretization`, `bet.sample.discretization` or None if no observation set.
         """
         out_dim = self.learn.num_pcs[0]
 
@@ -144,12 +160,15 @@ def make_disc(self):
         disc1 = sample.discretization(input_sample_set=self.predict_set,
                                       output_sample_set=predict_output,
                                       output_observed_set=obs_output)
+        disc1.local_to_global()
 
         # Observation discretization
-        disc2 = sample.discretization(input_sample_set=self.obs_set,
-                                      output_sample_set=obs_output)
-        disc1.local_to_global()
-        disc2.local_to_global()
+        if self.obs_set is None:
+            disc2 = None
+        else:
+            disc2 = sample.discretization(input_sample_set=self.obs_set,
+                                          output_sample_set=obs_output)
+            disc2.local_to_global()
 
         return disc1, disc2
 

doc/examples_overview.rst

@@ -7,3 +7,4 @@ Examples
 Examples
 ------------
 Documented examples can be found `here <https://github.com/UT-CHG/BET/tree/master/examples>`_.
+Jupyter notebooks of examples can be found `here <https://github.com/CU-Denver-UQ/BET-notebooks>`_.

requirements.txt

@@ -3,4 +3,3 @@ scipy>=1.3.1
 matplotlib>=3.0
 pyDOE
 pytest
-mpi4py

test/test_postProcess/test_plotP.py

@@ -202,6 +202,10 @@ def test_plot_marginals_1D(self):
             plotP.plot_1D_marginal_probs(marginals, bins, self.samples,
                                          filename="file", interactive=False)
             go = True
+            if os.path.exists("file_1D_0.png") and comm.rank == 0:
+                os.remove("file_1D_0.png")
+            if os.path.exists("file_1D_1.png") and comm.rank == 0:
+                os.remove("file_1D_1.png")
         except (RuntimeError, TypeError, NameError):
             go = False
         nptest.assert_equal(go, True)
@@ -222,6 +226,8 @@ def test_plot_marginals_2D(self):
             go = True
             if os.path.exists("file_2D_0_1.png") and comm.rank == 0:
                 os.remove("file_2D_0_1.png")
+            if os.path.exists("file_surf_0_1.png") and comm.rank == 0:
+                os.remove("file_surf_0_1.png")
         except (RuntimeError, TypeError, NameError):
             go = False
         nptest.assert_equal(go, True)