Merge pull request #30 from ELC/MinorChanges

Minor changes - First 3 commits of #29 (@ELC contribution)

yabox/algorithms/base.py

@@ -10,23 +10,21 @@ def binomial_crossover(target, mutant, cr):
     return np.where(p, mutant, target)
 
 
-def random_sample(population, exclude, size):
+def random_sample(population, exclude, size=3):
     # Optimized version using numpy
-    idxs = [i for i in range(population.shape[0]) if i != exclude]
-    np.random.shuffle(idxs)
-    # Select the number of elements specified by size
-    sample = idxs[:size]
+    idxs = list(range(population.shape[0]))
+    idxs.remove(exclude)
+    sample = np.random.choice(idxs, size=size, replace=False)
     return population[sample]
 
 
 def rand1(target_idx, population, f):
-    sample = random_sample(population, target_idx, 3)
-    a, b, c = sample
+    a, b, c = random_sample(population, target_idx)
     return a + f * (b - c)
 
 
-def denormalize(min, diff, matrix):
-    return min + matrix * diff
+def denormalize(min_, diff, matrix):
+    return min_ + matrix * diff
 
 
 def random_repair(x):

yabox/algorithms/de.py

@@ -96,23 +96,23 @@ def __init__(self, fobj, bounds, mutation=(0.5, 1.0), crossover=0.7, maxiters=10
         # Convert crossover param to an interval, as in mutation. If min/max values in the interval are
         # different, a dither mechanism is used for crossover (although this is not recommended, but still supported)
         # TODO: Clean duplicate code
-        if hasattr(crossover, '__len__'):
-            self.crossover_bounds = crossover
-        else:
-            self.crossover_bounds = (crossover, crossover)
-        if hasattr(mutation, '__len__'):
-            self.mutation_bounds = mutation
-        else:
-            self.mutation_bounds = (mutation, mutation)
+
+        self.crossover_bounds = crossover
+        self.mutation_bounds = mutation
+
+        if getattr(crossover, '__len__', None) is None:
+            self.crossover_bounds = [crossover, crossover]
+
+        if getattr(mutation, '__len__', None) is None:
+            self.mutation_bounds = [mutation, mutation]
+
         # If self-adaptive, include mutation and crossover as two new variables
         bnd = list(bounds)
         if self_adaptive:
             bnd.append(self.mutation_bounds)
             bnd.append(self.crossover_bounds)
             self.extra_params = 2
-        B = np.asarray(bnd, dtype='f8').T
-        self._MIN = B[0]
-        self._MAX = B[1]
+        self._MIN, self._MAX = np.asarray(bnd, dtype='f8').T
         self._DIFF = np.fabs(self._MAX - self._MIN)
         self.dims = len(bnd)
         self.fobj = fobj