VERSION 0.2.0

Author: blankjul

MANIFEST.in

@@ -146,14 +146,14 @@ Kriging
     import numpy as np
 
     from ezmodel.models.kriging import Kriging
-    from ezmodel.util.sample_from_func import sine_function
+    from ezmodel.util.sample_from_func import square_function
 
     model = Kriging(regr="linear",
                     corr="gauss",
                     ARD=False)
 
     # create some data to test this model on
-    X, y, _X, _y = sine_function(100, 20)
+    X, y, _X, _y = square_function(100, 20)
 
     # let the model fit the data
     model.fit(X, y)

README.rst

@@ -1,4 +1,4 @@
-ezmodel - A common interface for models and model selection
+ezmodel - A Common Interface for Models and Model Selection
 ====================================================================
 
 For more information about our toolbox, users are encouraged to read our documentation.

docs/github/README.template

@@ -239,7 +239,7 @@ def results(self,
 
         return ret
 
-    def statistics(self, metric="mae", vals=['mean', 'std', 'min', 'max', 'median'], sort=None):
+    def statistics(self, metric="mae", vals=['mean', 'std', 'min', 'max', 'median'], sort_by="mean", ascending=True):
         try:
             import pandas as pd
         except:
@@ -249,8 +249,8 @@ def statistics(self, metric="mae", vals=['mean', 'std', 'min', 'max', 'median'],
         tbl = df.groupby(["label"]).agg({metric: vals})
         df.groupby(["label"]).agg({metric: vals})
 
-        if sort is not None:
-            tbl = tbl.sort_values((metric, vals[0]), ascending=sort)
+        if sort_by is not None:
+            tbl = tbl.sort_values((metric, sort_by), ascending=ascending)
 
         return tbl
 

ezmodel/core/benchmark.py

@@ -2,14 +2,14 @@
 import numpy as np
 
 from ezmodel.models.kriging import Kriging
-from ezmodel.util.sample_from_func import sine_function
+from ezmodel.util.sample_from_func import square_function
 
 model = Kriging(regr="linear",
                 corr="gauss",
                 ARD=False)
 
 # create some data to test this model on
-X, y, _X, _y = sine_function(100, 20)
+X, y, _X, _y = square_function(100, 20)
 
 # let the model fit the data
 model.fit(X, y)

ezmodel/usage/models/usage_kriging.py

@@ -17,7 +17,7 @@
     description="Machine Learning, Model, Surrogate, Metamodels, Response Surface",
     license='Apache License 2.0',
     keywords="model",
-    install_requires=['numpy>=1.15', "pandas", "scipy", "sklearn", "pydacefit", "pySOT==0.2.3"],
+    install_requires=['numpy', "pandas", "scipy", "sklearn", "pydacefit", "matplotlib"],
     include_package_data=True,
     platforms='any',
     classifiers=[

ezmodel/usage/scratch.py

@@ -0,0 +1,29 @@
+import numpy as np
+
+import pandas as pd
+pd.set_option('display.expand_frame_repr', False)
+pd.set_option('max_colwidth', 1000)
+
+from ezmodel.core.benchmark import Benchmark
+from ezmodel.core.factory import models_from_clazzes
+from ezmodel.models.kriging import Kriging
+from ezmodel.models.rbf import RBF
+from ezmodel.util.partitioning.crossvalidation import CrossvalidationPartitioning
+
+X = np.random.random((100, 3)) * 2 * np.pi
+y = np.sin(X).sum(axis=1)
+
+models = models_from_clazzes(RBF, Kriging)
+
+# set up the benchmark and add the models to be used
+benchmark = Benchmark(models, n_threads=4, verbose=True, raise_exception=True)
+
+# create partitions to validate the performance of each model
+partitions = CrossvalidationPartitioning(k_folds=5, seed=1).do(X)
+
+# runs the experiment with the specified partitioning
+benchmark.do(X, y, partitions=partitions)
+
+# print out the benchmark results
+print(benchmark.statistics("mae"))
+