Fix bugs in discrete optimization

1. Fix bugs in the initialization step during the discrete optimization
2. Fix the bugs that index_set size equals zero during the discrete
optimization
3. Now users can initialize sample set using “init_samples =
[Solution(), …]“

docs/Explanation-of-Parameters-in-ZOOpt.rst

@@ -234,7 +234,7 @@ be set to ``True``. ``num_sre``, ``low_dimension`` and
 ``withdraw_alpha`` are parameters used in ``SRE`` and should be provided
 by users. ``num_sre`` means the number of sequential random embedding.
 ``low_dimension`` stands for the low dimension ``SRE`` projects to.
-``withdraw_alpha`` and ``variance_A`` are optimal parameters.
+``withdraw_alpha`` and ``variance_A`` are optional parameters.
 ``withdraw_alpha``, a withdraw variable to the previous solution, is a
 ``Dimension`` object with only one dimension. ``variance_A`` specifies
 the variance of the projection matrix A. By default, ``withdraw_alpha``

example/sequential_random_embedding/continuous_sre_opt.py

@@ -26,7 +26,7 @@ def minimize_sphere_sre():
     budget = 2000  # number of calls to the objective function
     parameter = Parameter(budget=budget, high_dim_handling=True, reducedim=True, num_sre=5,
                           low_dimension=Dimension(10, [[-1, 1]] * 10, [True] * 10))
-    solution_list = ExpOpt.min(objective, parameter, repeat=5, plot=True, plot_file="img/minimize_sphere_sre.png")
+    solution_list = ExpOpt.min(objective, parameter, repeat=5, plot=False, plot_file="img/minimize_sphere_sre.png")
 
 
 if __name__ == "__main__":

example/simple_functions/quick_start.py

@@ -5,13 +5,13 @@
     Yu-Ren Liu, Xiong-Hui Chen
 """
 
-from zoopt import Dimension, Objective, Parameter, ExpOpt
+from zoopt import Dimension, Objective, Parameter, ExpOpt, Solution
 from simple_function import ackley
 
 if __name__ == '__main__':
     dim = 100  # dimension
     objective = Objective(ackley, Dimension(dim, [[-1, 1]] * dim, [True] * dim))  # setup objective
-    parameter = Parameter(budget=100 * dim, sequential=False)
+    parameter = Parameter(budget=100 * dim, init_samples=[Solution([0] * 100)])  # init with init_samples
     solution_list = ExpOpt.min(objective, parameter, repeat=5, plot=False, plot_file="img/quick_start.png")
     for solution in solution_list:
         x = solution.get_x()

example/test.py

@@ -0,0 +1,18 @@
+import numpy as np
+from zoopt import Dimension, Objective, Parameter, ExpOpt, Solution, Opt
+from zoopt.utils.zoo_global import gl
+
+def fn(solution):
+    x = solution.get_x()
+    x1 = [0,3,1,0,0]
+    # print(x)
+    val = sum([abs(v-x1[i]) for i,v in enumerate(x)])+ ((np.sum([i for i in x[0:3]])-1.0)**2)
+    return val
+
+dim = 5
+gl.set_seed(0)
+dimobj = Dimension(dim, regs=[[0,5]]*dim, tys=[False]*dim)
+obj = Objective(fn, dimobj)
+param = Parameter(budget=1000)
+solution = Opt.min(obj, param)
+print(solution.get_x(), solution.get_value())

zoopt/algos/opt_algorithms/racos/racos_classification.py

@@ -149,6 +149,8 @@ def mixed_classification(self):
                             i += 1
                     if delete != 0:
                         index_set.remove(k)
+                    if len(index_set) == 0:
+                        index_set.append(k)
         self.set_uncertain_bit(index_set)
         return
 

zoopt/algos/opt_algorithms/racos/racos_common.py

@@ -5,8 +5,9 @@
     Yu-Ren Liu
 """
 
-import copy
+import copy, math
 from zoopt.utils.tool_function import ToolFunction
+from zoopt.solution import Solution
 
 
 class RacosCommon:
@@ -62,8 +63,12 @@ def init_attribute(self):
             if iteration_num < size:
                 size = iteration_num
             for j in range(size):
-                x = self._objective.construct_solution(data_temp[j])
-                self._objective.eval(x)
+                if isinstance(data_temp[j], Solution) is False:
+                    x = self._objective.construct_solution(data_temp[j])
+                else:
+                    x = data_temp[j]
+                if math.isnan(x.get_value()):
+                    self._objective.eval(x)
                 self._data.append(x)
                 ToolFunction.log("init solution %s, value: %s" % (i, x.get_value()))
                 i += 1

zoopt/dimension.py

@@ -113,11 +113,8 @@ def rand_sample(self):
             if self._types[i] is True:
                 value = gl.rand.uniform(
                     self._regions[i][0], self._regions[i][1])
-            elif self._order is True:
-                value = gl.rand.randint(self._regions[i][0], self._regions[i][1])
             else:
-                rand_index = gl.rand.randint(0, len(self._regions[i]) - 1)
-                value = self._regions[i][rand_index]
+                value = gl.rand.randint(self._regions[i][0], self._regions[i][1])
             x.append(value)
         return x
 

zoopt/solution.py

@@ -4,11 +4,11 @@
 Author:
     Yu-Ren Liu, Xiong-Hui Chen
 """
-
 from zoopt.utils.zoo_global import pos_inf, neg_inf, nan, gl
 from zoopt.utils.tool_function import ToolFunction
 import copy
 
+
 class Solution:
     """
     A solution encapsulates a solution vector with attached properties, including dimension information, objective value,