Merge pull request guofei9987#164 from samueljsluo/master

Add early stop feature for GA

sko/GA.py

@@ -14,13 +14,14 @@
 class GeneticAlgorithmBase(SkoBase, metaclass=ABCMeta):
     def __init__(self, func, n_dim,
                  size_pop=50, max_iter=200, prob_mut=0.001,
-                 constraint_eq=tuple(), constraint_ueq=tuple()):
+                 constraint_eq=tuple(), constraint_ueq=tuple(), early_stop=None):
         self.func = func_transformer(func)
         assert size_pop % 2 == 0, 'size_pop must be even integer'
         self.size_pop = size_pop  # size of population
         self.max_iter = max_iter
         self.prob_mut = prob_mut  # probability of mutation
         self.n_dim = n_dim
+        self.early_stop = early_stop
 
         # constraint:
         self.has_constraint = len(constraint_eq) > 0 or len(constraint_ueq) > 0
@@ -75,6 +76,7 @@ def mutation(self):
 
     def run(self, max_iter=None):
         self.max_iter = max_iter or self.max_iter
+        best = []
         for i in range(self.max_iter):
             self.X = self.chrom2x(self.Chrom)
             self.Y = self.x2y()
@@ -90,6 +92,14 @@ def run(self, max_iter=None):
             self.all_history_Y.append(self.Y)
             self.all_history_FitV.append(self.FitV)
 
+            if self.early_stop:
+                best.append(min(self.generation_best_Y))
+                if len(best) >= self.early_stop:
+                    if best.count(min(best)) == len(best):
+                        break
+                    else:
+                        best.pop(0)
+
         global_best_index = np.array(self.generation_best_Y).argmin()
         self.best_x = self.generation_best_X[global_best_index]
         self.best_y = self.func(np.array([self.best_x]))
@@ -141,8 +151,8 @@ def __init__(self, func, n_dim,
                  prob_mut=0.001,
                  lb=-1, ub=1,
                  constraint_eq=tuple(), constraint_ueq=tuple(),
-                 precision=1e-7):
-        super().__init__(func, n_dim, size_pop, max_iter, prob_mut, constraint_eq, constraint_ueq)
+                 precision=1e-7, early_stop=None):
+        super().__init__(func, n_dim, size_pop, max_iter, prob_mut, constraint_eq, constraint_ueq, early_stop)
 
         self.lb, self.ub = np.array(lb) * np.ones(self.n_dim), np.array(ub) * np.ones(self.n_dim)
         self.precision = np.array(precision) * np.ones(self.n_dim)  # works when precision is int, float, list or array
@@ -244,7 +254,7 @@ def chrom2x(self, Chrom):
             register('chrom2x', chrom2x)
 
         return self
-    
+
 class RCGA(GeneticAlgorithmBase):
     """real-coding genetic algorithm