Version bump to 1.2.1

Added LIN + C -> LIN kernel simplification provided the offset-including LIN version is used, which is the case by default (i.e. the __USE_LIN_KERNEL_HORIZONTAL_OFFSET flag in config.py is True).
Added a plotly argument to model_printout in modelUtil.py and used it in the readme example.

GPy_ABCD/KernelExpressions/commutatives.py

@@ -5,6 +5,7 @@
 from GPy_ABCD.KernelExpressions.commutative_base import KernelExpression, SumOrProductKE
 from GPy_ABCD.KernelExpansion.kernelOperations import *
 from GPy_ABCD.KernelExpansion.kernelInterpretation import *
+import GPy_ABCD.config as config
 
 
 class SumKE(SumOrProductKE):
@@ -15,6 +16,10 @@ def simplify_base_terms(self):
         # WN and C are addition-idempotent
         if self.base_terms['WN'] > 1: self.base_terms['WN'] = 1
         if self.base_terms['C'] > 1: self.base_terms['C'] = 1
+
+        # If an offset-including LIN is used, remove any C in their presence (going through __dict__ because the class name gets prepended otherwise)
+        if config.__dict__['__USE_LIN_KERNEL_HORIZONTAL_OFFSET'] and self.base_terms['LIN'] > 0: self.base_terms['C'] = 0
+
         self.base_terms = + self.base_terms
         return self
 

GPy_ABCD/Kernels/baseKernels.py

@@ -31,8 +31,8 @@ def C(): return _Gk.Bias(1)
 
 # def LIN(): return _Gk.Linear(1) # Not the same as ABCD's; missing horizontal offset
 def LIN(): return _Lk.Linear(1) # Not the same as ABCD's; missing horizontal offset
-if __USE_LIN_KERNEL_HORIZONTAL_OFFSET: # The version in ABCD; not sure if a good idea; the horizontal offset is the same as a vertical one, which is just kC
-    def LIN(): return _LOk.LinearWithOffset(1)
+if __USE_LIN_KERNEL_HORIZONTAL_OFFSET: # This flag also enables LIN + C -> LIN simplification
+    def LIN(): return _LOk.LinearWithOffset(1) # The version in ABCD; the horizontal offset is the same as a vertical one, which is just kC
 
 def SE(): return _Gk.RBF(1)
 

GPy_ABCD/Util/modelUtil.py

@@ -22,12 +22,15 @@ def LA_LOO(m, ll, n, k): return np.mean(m.inference_method.LOO(m.kern, m.X, m.Y,
 GPy_optimisers = ['lbfgsb', 'org-bfgs', 'fmin_tnc', 'scg', 'simplex', 'adadelta', 'rprop', 'adam']
 
 
-def model_printout(m):
+def model_printout(m, plotly = False):
     print(m.kernel_expression)
     print(m.model.kern)
     print(f'Log-Lik: {m.model.log_likelihood()}')
     print(f'{m.cached_utility_function_type}: {m.cached_utility_function}')
-    m.plot()
+    if plotly:
+        m.change_plotting_library(library = 'plotly_offline')
+        m.plot()[0].show()
+    else: m.plot()
     print(m.interpret())
 
 

GPy_ABCD/__init__.py

@@ -1,6 +1,6 @@
 """GPy-ABCD - Basic implementation with GPy of an Automatic Bayesian Covariance Discovery (ABCD) system"""
 
-__version__ = '1.2' # Change it in setup.py too
+__version__ = '1.2.1' # Change it in setup.py too
 __author__ = 'Thomas Fletcher <[email protected]>'
 # __all__ = []
 

README.rst

@@ -75,11 +75,11 @@ A minimal example to showcase the various parameters follows:
 
         print('\n\nTop-3 models\' details:')
         for bm in best_mods[:3]:
-            model_printout(bm) # See the definition of this convenience function for examples of model details' extraction
+            model_printout(bm, plotly = False) # See the definition of this convenience function for examples of model details' extraction
             print('Prediction at X = 11:', bm.predict(np.array([11])[:, None]), '\n')
 
         from matplotlib import pyplot as plt
-        plt.show()
+        plt.show() # Not required for plotly = True above
 
 
 
@@ -158,8 +158,8 @@ Further Notes
 
 Generic:
 
-- Please let know me if you have successfully used this project in your own research
-- Please feel free to fork and expand this project (pull requests are welcome) since it is not the focus of my research; it was written just because I needed to use it in a broader adaptive statistical modelling context and therefore I have no need to expand its functionality in the near future
+- Please reach out if you have successfully used this project in your own research
+- Feel free to fork and expand this project (pull requests are welcome) since it is not the focus of my research; it was written just because I needed to use it in a broader adaptive statistical modelling context and therefore I have no need to expand its functionality in the near future
 
 Code-related:
 

Tests/checkModelSearchREADMEexample.py

@@ -19,10 +19,10 @@
 
     print('\n\nTop-3 models\' details:')
     for bm in best_mods[:3]:
-        model_printout(bm) # See the definition of this convenience function for examples of model details' extraction
+        model_printout(bm, plotly = False) # See the definition of this convenience function for examples of model details' extraction
         print('Prediction at X = 11:', bm.predict(np.array([11])[:, None]), '\n')
 
     from matplotlib import pyplot as plt
-    plt.show()
+    plt.show() # Not required for plotly = True above
 
 

Tests/test_grammar.py

@@ -86,3 +86,4 @@ def test_specific_expansions(self, es):
         for e in es: print(e)
         print(len(es))
 
+

setup.py

@@ -19,7 +19,7 @@ def read_requirements():
 
 setup(
     name = 'GPy-ABCD',
-    version = '1.2', # Change it in __init__.py too
+    version = '1.2.1', # Change it in __init__.py too
     url = 'https://github.com/T-Flet/GPy-ABCD',
     license = 'BSD 3-Clause',