Add mc func (#12)

* Add MCMcCourt08

* Add classifier

* bump version

Author: bolongcheng

evalset/multicriteria_test_funcs.py

@@ -478,6 +478,67 @@ def do_component_function(self, d, x):
             return -1 / (1 / numpy.linalg.norm(x - sol1) + 1 / numpy.linalg.norm(x - sol2))
 
 
+class MCMcCourt08(MulticriteriaTestFunction):
+    def __init__(self, dim=12, verify=True):
+        assert dim == 12
+        super(MCMcCourt08, self).__init__(dim, verify)
+        self.output_dim = 2
+        self.bounds = (
+            [700, 1277], [200, 1277], [30, 175], [50, 140], [18, 80], [5, 58],
+            [51, 140], [30, 60], [0, 58], [175, 210], [0, 12], [5, 35],
+        )
+        self.frontier = numpy.array([
+            [-0.973, -1.559], [-0.928, -1.601], [-0.951, -1.591], [-0.892, -1.612], [-0.935, -1.598],
+            [-1.496, -0.785], [-0.840, -1.613], [-0.947, -1.593], [-0.921, -1.603], [-0.965, -1.577],
+            [-0.933, -1.600], [-0.900, -1.609], [-0.954, -1.588], [-0.898, -1.610], [-0.941, -1.596],
+            [-0.972, -1.564], [-0.940, -1.597], [-0.949, -1.592], [-0.967, -1.574], [-0.961, -1.581],
+            [-0.904, -1.609], [-1.502, -0.779], [-1.503, -0.777], [-1.507, -0.769], [-0.931, -1.600],
+            [-0.895, -1.611], [-0.970, -1.570], [-0.903, -1.609], [-1.496, -0.787], [-1.502, -0.777],
+            [-0.912, -1.606], [-1.499, -0.784], [-1.501, -0.781], [-1.505, -0.774], [-1.505, -0.772],
+            [-1.507, -0.770], [-1.508, -0.761], [-1.510, -0.749], [-1.513, -0.735], [-1.514, -0.730],
+            [-1.515, -0.727], [-1.513, -0.737], [-0.815, -1.615], [-0.815, -1.615], [-0.820, -1.615],
+            [-0.816, -1.615], [-0.829, -1.614], [-0.814, -1.615], [-1.435, -0.810], [-0.996, -1.499],
+            [-1.385, -0.819], [-1.002, -1.480], [-1.009, -1.437], [-1.375, -0.826], [-1.006, -1.452],
+            [-1.320, -0.833], [-1.326, -0.833], [-1.014, -1.401], [-1.016, -1.364], [-1.332, -0.832],
+            [-1.018, -1.329], [-1.281, -0.841], [-1.020, -1.305], [-1.022, -1.265], [-1.021, -1.297],
+            [-1.255, -0.846], [-1.256, -0.845], [-1.160, -0.858], [-1.222, -0.850], [-1.183, -0.853],
+            [-1.179, -0.854], [-1.174, -0.855], [-1.024, -1.211], [-1.101, -0.865], [-1.091, -0.866],
+            [-1.091, -0.866], [-1.095, -0.865], [-1.025, -1.204], [-1.028, -1.131], [-1.056, -0.870],
+            [-1.028, -1.098], [-1.029, -1.065], [-1.029, -1.027], [-1.031, -0.893], [-1.030, -0.990],
+            [-1.030, -0.970], [-0.916, -1.604], [-1.031, -0.923], [-0.896, -1.610], [-0.886, -1.612],
+            [-0.886, -1.612], [-0.832, -1.614], [-1.081, -0.867], [-1.030, -1.015], [-1.030, -1.001],
+            [-1.240, -0.848], [-1.018, -1.360], [-1.023, -1.255], [-1.125, -0.861], [-1.031, -0.941],
+            [-1.030, -0.961], [-1.024, -1.250], [-1.026, -1.191], [-1.021, -1.296], [-0.977, -1.530],
+            [-1.444, -0.797], [-1.467, -0.795], [-0.980, -1.501], [-1.420, -0.819], [-1.445, -0.795],
+        ])
+        self.classifiers = ['nonconvex', 'rescaled']
+
+    def do_component_function(self, d, x):
+        sx = numpy.empty(self.dim)
+        for k, b in enumerate(self.bounds):
+            b_min, b_max = b
+            sx[k] = (x[k] - b_min) / (b_max - b_min)
+        nonconvex_factor = 3
+        if d == 0:
+            scaling1 = numpy.linspace(1 / self.dim, 1.0, self.dim)
+            scaling2 = numpy.linspace(1.3, 1 / self.dim, self.dim)
+            distance1_sq = numpy.dot(sx - 1 / 3, scaling1 * (sx - 1 / 3))
+            distance2_sq = numpy.dot(sx - 2 / 3, scaling2 * (sx - 2 / 3))
+            return (
+                -numpy.exp(-2 * distance1_sq * nonconvex_factor) -
+                1.5 * numpy.exp(-1.6 * distance2_sq * nonconvex_factor)
+            )
+        else:
+            scaling1 = numpy.linspace(1.3, 1 / self.dim, self.dim)
+            scaling2 = numpy.linspace(1 / self.dim, 1.0, self.dim)
+            distance1_sq = numpy.dot(sx - 0.25, scaling1 * (sx - 0.25))
+            distance2_sq = numpy.dot(sx - 0.6, scaling2 * (sx - 0.6))
+            return (
+                -1.6 * numpy.exp(-1.2 * distance1_sq * nonconvex_factor) -
+                .8 * numpy.exp(-1.7 * distance2_sq * nonconvex_factor)
+            )
+
+
 class FonsecaFleming(MulticriteriaTestFunction):
     def __init__(self, dim=2, verify=True):
         assert dim == 2

setup.py

@@ -4,7 +4,7 @@
 
 setup(
   name='evalset',
-  version='1.2.2',
+  version='1.2.3',
   description='Benchmark suite of test functions suitable for evaluating black-box optimization strategies',
   author='SigOpt',
   author_email='[email protected]',