learn

pyrimidine/learn/__init__.py

@@ -9,13 +9,23 @@
 
 class BaseEstimator(BE):
 
+    estimated_params = ()
+
     @classmethod
     def config(cls, X, Y, *args, **kwargs):
+        # configure a population for GA
         raise NotImplementedError
 
     def fit(self, X, Y, 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(X, Y)
+        return self
+
+    def _fit(self, X, Y):
+        self.pop.ezolve(n_iter=self.max_iter)
+        model_ = self.pop.solution
+        for k in self.estimated_params:
+            setattr(self, k, getattr(model_, k))

pyrimidine/learn/linear_regression.py

@@ -1,7 +1,7 @@
 #!/usr/bin/env python3
 
 
-from pyrimidine import *
+from pyrimidine import StandardPopulation
 from digit_converter import IntervalConverter
 
 c = IntervalConverter(lb=-60, ub=60)
@@ -20,24 +20,21 @@ class GALinearRegression(BaseEstimator, LinearRegression):
     '''Linear Regression by GA
     '''
 
+    estimated_params = ('coefs_', 'intercepts_')
+
     @classmethod
     def create_model(cls, *args, **kwargs):
         return LinearRegression(*args, **kwargs)
 
-    def _fit(self, X, Y):
-        self.pop.ezolve(n_iter=self.max_iter)
-        model_ = self.pop.solution
-        self.coef_ = model_.coef_
-        self.intercept_ = model_.intercept_
-
     @classmethod
     def config(cls, X, Y, n_individuals=10, *args, **kwargs):
 
         input_dim = X.shape[1]
+        assert np.ndim(Y) == 1, 'only support 1D array for `Y`'
         # output_dim = Y.shape[1]
 
         class MyIndividual(MixedIndividual):
-            params={'sigma':0.02}
+
             element_class = FloatChromosome, _BinaryChromosome
 
             def decode(self):
@@ -50,8 +47,7 @@ def _fitness(self):
                 model = self.decode()
                 return model.score(X, Y)
 
-        class MyPopulation(HOFPopulation):
-            element_class = MyIndividual
+        MyPopulation = StandardPopulation[MyIndividual]
 
         pop = MyPopulation.random(n_individuals=n_individuals, size=(input_dim, 8))
 

pyrimidine/learn/neural_network.py

@@ -8,7 +8,7 @@
 from sklearn.neural_network import MLPRegressor
 
 from .. import MixedIndividual, FloatChromosome, FloatMatrixChromosome
-from ..population import HOFPopulation
+from ..population import StandardPopulation
 from ..learn import BaseEstimator
 
 
@@ -22,6 +22,8 @@ class GAANN(BaseEstimator, MLPRegressor):
     max_iter = 100
     n_layers = 3
 
+    estimated_params = ('coefs_', 'intercepts_')
+
     @classmethod
     def create_model(cls, *args, **kwargs):
         # create MLPRegressor object
@@ -58,13 +60,7 @@ def decode(self):
                 model.n_layers_ = 3
                 return model
 
-        MyPopulation = HOFPopulation[MyIndividual]
+        MyPopulation = StandardPopulation[MyIndividual]
 
         return MyPopulation.random(n_individuals=n_individuals, size=((input_dim, cls.hidden_dim), cls.hidden_dim, (cls.hidden_dim, output_dim), output_dim))
 
-    def _fit(self, X, Y):
-        self.pop.ezolve(n_iter=self.max_iter)
-        model_ = self.pop.solution
-        self.coefs_ = model_.coefs_
-        self.intercepts_ = model_.intercepts_
-