learn

pyrimidine/learn/__init__.py

@@ -2,30 +2,50 @@
 
 from sklearn.base import BaseEstimator as BE
 
+from ..chromosome import BinaryChromosome
+
 import warnings, os
 warnings.filterwarnings("ignore")
 os.environ['TF_CPP_MIN_LOG_LEVEL'] = '3'
 
 
 class BaseEstimator(BE):
 
+    """Base class for machine learning by GA
+    
+    Attributes:
+        estimated_params (tuple): estimtated/learnable parameters by GA
+        pop (Population): the population for GA
+    """
+
+    pop = None
     estimated_params = ()
 
     @classmethod
-    def config(cls, X, Y, *args, **kwargs):
+    def config(cls, X, Y=None, *args, **kwargs):
         # configure a population for GA
         raise NotImplementedError
 
-    def fit(self, X, Y, pop=None, warm_start=False):
+    def fit(self, X, Y=None, pop=None, warm_start=False):
         if warm_start:
             self.pop = pop or self.pop or self.config(X, Y)
         else:
             self.pop = pop or self.config(X, Y)
-        self._fit(X, Y)
+        self._fit()
         return self
 
-    def _fit(self, X, Y):
+    def _fit(self):
         self.pop.ezolve(n_iter=self.max_iter)
         model_ = self.pop.solution
         for k in self.estimated_params:
-            setattr(self, k, getattr(model_, k))
+            setattr(self, k, getattr(model_, k))
+
+
+# def ga_opt(model_class):
+#     class cls(BaseEstimator, model_class):
+#         @classmethod
+#         def create_model(cls, *args, **kwargs):
+#             return model_class(*args, **kwargs)
+
+#     return cls
+

pyrimidine/learn/cluster.py

@@ -0,0 +1,43 @@
+#!/usr/bin/env python3
+
+"""GA for linear regression
+"""
+
+import numpy as np
+from sklearn.cluster import KMeans
+from ..learn import BaseEstimator
+
+from .population import StandardPopulation
+
+
+class GALinearRegression(BaseEstimator, KMeans):
+    '''Linear Regression by GA
+    '''
+
+    estimated_params = ('cluster_centers_',)
+
+    @classmethod
+    def create_model(cls, *args, **kwargs):
+        return KMeans(*args, **kwargs)
+
+    @classmethod
+    def config(cls, X, Y=None, n_clusters=2, n_individuals=10, *args, **kwargs):
+
+        n_features = X.shape[1]
+
+        class MyIndividual(FloatChromosome // (n_clusters, n_features)):
+
+            def decode(self):
+                model = cls.create_model(*args, **kwargs)
+                model.cluster_centers_ = np.asarray(self)
+                return model
+
+            def _fitness(self):
+                model = self.decode()
+                return model.score(X)
+
+        MyPopulation = StandardPopulation[MyIndividual]
+
+        pop = MyPopulation.random(n_individuals=n_individuals)
+
+        return pop

pyrimidine/learn/linear_regression.py

@@ -1,7 +1,14 @@
 #!/usr/bin/env python3
 
+"""GA for linear regression
+"""
 
-from pyrimidine import StandardPopulation
+import numpy as np
+import numpy.linalg as LA
+from sklearn.linear_model import LinearRegression
+from ..learn import BaseEstimator
+
+from .population import StandardPopulation
 from digit_converter import IntervalConverter
 
 c = IntervalConverter(lb=-60, ub=60)
@@ -10,12 +17,6 @@ def decode(self):
         return c(self)
 
 
-import numpy as np
-import numpy.linalg as LA
-from sklearn.linear_model import LinearRegression
-from ..learn import BaseEstimator
-
-
 class GALinearRegression(BaseEstimator, LinearRegression):
     '''Linear Regression by GA
     '''

pyrimidine/learn/neural_network.py

@@ -1,9 +1,9 @@
 #!/usr/bin/env python3
 
+"""GA for optimization of neural networks
+"""
+
 import numpy as np
-import numpy.linalg as LA
-from scipy.special import softmax
-from scipy.stats import entropy
 
 from sklearn.neural_network import MLPRegressor
 
@@ -12,12 +12,10 @@
 from ..learn import BaseEstimator
 
 
-class GAANN(BaseEstimator, MLPRegressor):
-    """GA for ANN
+class GAMLPRegression(BaseEstimator, MLPRegressor):
+    """GA for MLP Regression
     """
 
-    pop = None
-
     hidden_dim = 4
     max_iter = 100
     n_layers = 3
@@ -33,7 +31,7 @@ def create_model(cls, *args, **kwargs):
         return model
 
     def __init__(self, *args, **kwargs):
-        # create GAANN object
+        # create MLPRegressor
         super().__init__(hidden_layer_sizes=(self.hidden_dim,), max_iter=1, *args, **kwargs)
         self.out_activation_ = 'identity'
         self.n_layers_ = 3