pymoo 0.6.1.5.dev0__cp313-cp313-manylinux_2_17_x86_64.manylinux2014_x86_64.whl

pymoo/util/vectors.py

@@ -0,0 +1,40 @@
+import numpy as np
+
+
+def max_alpha(point, direction, xl, xu, mode="one_hits_bound"):
+    bounds = []
+
+    if xl is not None:
+        bounds.append(xl)
+
+    if xu is not None:
+        bounds.append(xu)
+
+    if len(bounds) == 0:
+        return np.inf
+
+    # the bounds in one array
+    bounds = np.column_stack(bounds)
+
+    # if the direction is too small we can not divide by 0 - nan will make it being ignored
+    dir = direction.copy()
+    dir[dir == 0] = np.nan
+
+    # calculate the max factor to be not out of bounds
+    val = (bounds - point[:, None]) / dir[:, None]
+
+    # remove nan values by setting them to a negative number
+    val[np.isnan(val)] = - np.inf
+
+    # if no value there - no bound exist
+    if len(val) == 0:
+        return np.inf
+    # otherwise return the minimum of values considered
+    else:
+        if mode == "one_hits_bound":
+            if not np.any(val >= 0):
+                return 0.0
+            else:
+                return val[val >= 0].min()
+        else:
+            return val.max()

pymoo/util/vf_dominator.py

@@ -0,0 +1,99 @@
+import numpy as np
+
+def get_relation(ind_a, ind_b):
+    return Dominator.get_relation(ind_a.F, ind_b.F, ind_a.CV[0], ind_b.CV[0])
+
+
+class VFDominator:
+
+    def __init__(self, algorithm):
+
+        self.algorithm = algorithm
+
+    @staticmethod
+    def get_relation(a, b, cva=None, cvb=None):
+
+        if cva is not None and cvb is not None:
+            if cva < cvb:
+                return 1
+            elif cvb < cva:
+                return -1
+
+        val = 0
+        for i in range(len(a)):
+            if a[i] < b[i]:
+                # indifferent because once better and once worse
+                if val == -1:
+                    return 0
+                val = 1
+            elif b[i] < a[i]:
+                # indifferent because once better and once worse
+                if val == 1:
+                    return 0
+                val = -1
+        return val
+
+    @staticmethod
+    def calc_domination_matrix_loop(F, G):
+
+        n = F.shape[0]
+        CV = np.sum(G * (G > 0).astype(float), axis=1)
+        M = np.zeros((n, n))
+        for i in range(n):
+            for j in range(i + 1, n):
+                M[i, j] = Dominator.get_relation(F[i, :], F[j, :], CV[i], CV[j])
+                M[j, i] = -M[i, j]
+
+        return M
+
+    def calc_domination_matrix(self, F, _F=None, epsilon=0.0):
+
+        if _F is None:
+            _F = F
+
+        # look at the obj for dom
+        n = F.shape[0]
+        m = _F.shape[0]
+
+        L = np.repeat(F, m, axis=0)
+        R = np.tile(_F, (n, 1))
+
+        smaller = np.reshape(np.any(L + epsilon < R, axis=1), (n, m))
+        larger = np.reshape(np.any(L > R + epsilon, axis=1), (n, m))
+    
+        non_dom = np.logical_and(smaller, np.logical_not(larger))
+        dom = np.logical_and(larger, np.logical_not(smaller))
+
+        if self.algorithm.vf_res is not None: 
+
+            # Figure out what the v2 value is 
+            v2 = self.algorithm.v2
+
+            # Get the value function
+            vf = self.algorithm.vf_res.vf
+
+            # How much does the DM value each solution?
+            F_vf = vf(F * -1)[:,np.newaxis]
+            _F_vf = vf(_F * -1)[:,np.newaxis]
+
+            # We want to compare each solution to the others 
+            Lv = np.repeat(F_vf, m, axis=0)
+            Rv = np.tile(_F_vf, (n, 1))
+            
+            # Which values are greater than (better) V2? 
+            gtv2 = np.reshape(Lv < v2, (n, m))
+            # Which values are less than (worst) V2? 
+            ltv2 = np.reshape(Rv > v2, (n, m))
+         
+            # If you are greater than V2, you dominate all those who are smaller than V2
+            split_by_v2 = np.logical_and(gtv2, ltv2)
+           
+            dom = np.logical_or(dom, split_by_v2)
+
+        M = non_dom * 1 \
+            + dom * -1
+       
+        return M
+
+
+

pymoo/vendor/cec2018.py

@@ -0,0 +1,398 @@
+# The code is translated from Matlab by ahcheriet@gmail.com
+# ========================================================|#
+# The 14 test functions are for cec2018 competition on    |
+# dynamic multiobjective optimisation. This document is   |
+# free to disseminate for academic use.                   |
+# --------------------------------------------------------|#
+# The "time" term in the test suite is defined as:        |
+#          t=1/nt*floor(tau/taut)                         |
+# where - nt:    severity of change                       |
+#       - taut:  frequency of change                      |
+#       - tau:   current generation counter               |
+# --------------------------------------------------------|#
+# Any questions can be directed to                        |
+#    Dr. Shouyong Jiang at math4neu@gmail.com.            |
+#                                                         |
+# ========================================================|#
+
+# cec2018_DF(probID, x, tau, taut, nt)
+# INPUT:
+#       probID: test problem identifier (i.e. 'DF1')
+#       x:      variable vector
+#       tau:    current generation counter
+#       taut:   frequency of change
+#       nt:     severity of change
+#
+# OUTPUT:
+#       f:      objective vector
+#
+from numpy import power, setdiff1d, exp, prod
+
+
+def get_bounds(problem_id='DF1', n_vars=10):
+    lx = np.zeros(n_vars)
+    ux = np.ones(n_vars)
+    if problem_id == 'DF3':
+        lx = -1 * np.ones(n_vars)
+        ux = 2 * np.ones(n_vars)
+        lx[0] = 0.0
+        ux[0] = 1.0
+    if problem_id == 'DF4':
+        lx = -2 * np.ones(n_vars)
+        ux = -2 * np.ones(n_vars)
+    if problem_id == 'DF5':
+        lx = -1 * np.ones(n_vars)
+        ux = -1 * np.ones(n_vars)
+        lx[0] = 0.0
+        ux[0] = 1.0
+    if problem_id == 'DF6':
+        lx = -1 * np.ones(n_vars)
+        ux = -1 * np.ones(n_vars)
+        lx[0] = 0.0
+        ux[0] = 1.0
+    if problem_id == 'DF7':
+        lx = 0 * np.ones(n_vars)
+        ux = 1 * np.ones(n_vars)
+        lx[0] = 1.0
+        ux[0] = 4.0
+    if problem_id == 'DF8':
+        lx = -1 * np.ones(n_vars)
+        ux = -1 * np.ones(n_vars)
+        lx[0] = 0.0
+        ux[0] = 1.0
+    if problem_id == 'DF9':
+        lx = -1 * np.ones(n_vars)
+        ux = -1 * np.ones(n_vars)
+        lx[0] = 0.0
+        ux[0] = 1.0
+    if problem_id == 'DF10':
+        lx = -1 * np.ones(n_vars)
+        ux = -1 * np.ones(n_vars)
+        lx[0] = 0.0
+        lx[1] = 0.0
+        ux[0] = 1.0
+        ux[1] = 1.0
+    if problem_id == 'DF12':
+        lx = -1 * np.ones(n_vars)
+        ux = -1 * np.ones(n_vars)
+        lx[0] = 0.0
+        lx[1] = 0.0
+        ux[0] = 1.0
+        ux[1] = 1.0
+    if problem_id == 'DF13':
+        lx = -1 * np.ones(n_vars)
+        ux = -1 * np.ones(n_vars)
+        lx[0] = 0.0
+        lx[1] = 0.0
+        ux[0] = 1.0
+        ux[1] = 1.0
+    if problem_id == 'DF14':
+        lx = -1 * np.ones(n_vars)
+        ux = -1 * np.ones(n_vars)
+        lx[0] = 0.0
+        lx[1] = 0.0
+        ux[0] = 1.0
+        ux[1] = 1.0
+    return lx, ux
+
+
+def cec2018_DF(problemID='DF1', x=None, t=None):
+    # INPUT:
+    #       probID: test problem identifier (i.e. 'DF1')
+    #       x:      variable vector
+    #       tau:    current generation counter
+    #       taut:   frequency of change
+    #       nt:     severity of change
+
+    # OUTPUT:
+    #       f:      objective vector
+
+    # the first change occurs after T0 generations, that is, the
+    # generation at which a change occurs is (T0+1), (T0+taut+1), etc.
+
+    T0 = 50
+    # calculate time instant
+    n = len(x)
+    f = {}
+    if problemID == 'DF1':
+        G = abs(sin(0.5 * pi * t))
+        H = 0.75 * sin(0.5 * pi * t) + 1.25
+        g = 1 + sum((x[1:] - G) ** 2)
+        f[0] = x[0]
+        f[1] = g * power(1 - (x[0] / g), H)
+    if problemID == 'DF2':
+        G = abs(sin(0.5 * pi * t))
+        r = 1 + floor((n - 1) * G)
+        tmp = setdiff1d(range(0, n), [int(r)])
+        g = 1 + sum([(x[int(index)] - G) ** 2 for index in tmp])
+        f[0] = x[int(r - 1)]
+        f[1] = g * (power(1 - (x[int(r - 1)] / g), 0.5))
+    if problemID == 'DF3':
+        G = sin(0.5 * pi * t)
+        H = G + 1.5
+        g = 1 + sum(power(x[1:] - G - x[0], H) ** 2)
+        f[0] = x[0]
+        f[1] = g * power(1 - (x[0] / g), H)
+    if problemID == 'DF4':
+        a = sin(0.5 * pi * t)
+        b = 1 + abs(cos(0.5 * pi * t))
+        H = 1.5 + a
+        g = 1 + sum((x[1:] - a * x[0] ** 2 / x[1:]) ** 2)
+        f[0] = g * power(abs(x[0] - a), H)
+        f[1] = g * power(abs(x[0] - a - b), H)
+    if problemID == 'DF5':
+        G = sin(0.5 * pi * t)
+        w = floor(10 * G)
+        g = 1 + sum((x[1:] - G) ** 2)
+        f[0] = g * (x[0] + 0.02 * sin(w * pi * x[0]))
+        f[1] = g * (1 - x[0] + 0.02 * sin(w * pi * x[0]))
+    if problemID == 'DF6':
+        G = sin(0.5 * pi * t)
+        a = 0.2 + 2.8 * abs(G)
+        y = x[1:] - G
+        g = 1 + sum((abs(G) * y ** 2 - 10 * cos(2 * pi * y) + 10))
+        f[0] = g * power(x[0] + 0.1 * sin(3 * pi * x[0]), a)
+        f[1] = g * power(1 - x[0] + 0.1 * sin(3 * pi * x[0]), a)
+    if problemID == 'DF7':
+        a = 5 * cos(0.5 * pi * t)
+        tmp = 1 / (1 + exp(a * (x[0] - 2.5)))
+        g = 1 + sum(power(x[1:] - tmp, 2))
+        f[0] = g * (1 + t) / x[0]
+        f[1] = g * x[0] / (1 + t)
+    if problemID == 'DF8':
+        G = sin(0.5 * pi * t)
+        a = 2.25 + 2 * cos(2 * pi * t)
+        b = 100 * G ** 2
+        tmp = G * sin(power(4 * pi * x[0], b)) / (1 + abs(G))
+        g = 1 + sum((x[1:] - tmp) ** 2)
+        f[0] = g * (x[0] + 0.1 * sin(3 * pi * x[0]))
+        f[1] = g * power(1 - x[1] + 0.1 * sin(3 * pi * x[1]), a)
+    if problemID == 'DF9':
+        N = 1 + floor(10 * abs(sin(0.5 * pi * t)))
+        g = 1
+        for i in range(1, n):
+            tmp = x[i] - cos(4 * t + x[0] + x[i - 1])
+            g = g + tmp ** 2
+        f[0] = g * (x[0] + max(0, (0.1 + 0.5 / N) * sin(2 * N * pi * x[0])))
+        f[1] = g * (1 - x[0] + max(0, (0.1 + 0.5 / N) * sin(2 * N * pi * x[0])))
+    if problemID == 'DF10':
+        G = sin(0.5 * pi * t)
+        H = 2.25 + 2 * cos(0.5 * pi * t)
+        tmp = sin(2 * pi * (x[0] + x[1])) / (1 + abs(G))
+        g = 1 + sum((x[2:] - tmp) ** 2)
+        f[0] = g * power(sin(0.5 * pi * x[0]), H)
+        f[1] = g * power(sin(0.5 * pi * x[1]), H) * power(cos(0.5 * pi * x[0]), H)
+        f[2] = g * power(cos(0.5 * pi * x[1]), H) * power(cos(0.5 * pi * x[0]), H)
+    if problemID == 'DF11':
+        G = abs(sin(0.5 * pi * t))
+        g = 1 + G + sum((x[2:] - 0.5 * G * x[0]) ** 2)
+        y = [pi * G / 6.0 + (pi / 2 - pi * G / 3.0) * x[i] for i in [0, 1]]
+        f[0] = g * sin(y[0])
+        f[1] = g * sin(y[1]) * cos(y[0])
+        f[2] = g * cos(y[1]) * cos(y[0])
+    if problemID == 'DF12':
+        k = 10 * sin(pi * t)
+        tmp1 = x[2:] - sin(t * x[0])
+        tmp2 = [sin(floor(k * (2 * x[0] - 1)) * pi / 2)]
+        g = 1 + sum(tmp1 ** 2) + prod(tmp2)
+        f[0] = g * cos(0.5 * pi * x[1]) * cos(0.5 * pi * x[0])
+        f[1] = g * sin(0.5 * pi * x[1]) * cos(0.5 * pi * x[0])
+        f[2] = g * sin(0.5 * pi * x[1])
+    if problemID == 'DF13':
+        G = sin(0.5 * pi * t);
+        p = floor(6 * G);
+        g = 1 + sum((x[2:] - G) ** 2)
+        f[0] = g * cos(0.5 * pi * x[0]) ** 2
+        f[1] = g * cos(0.5 * pi * x[1]) ** 2
+        f[2] = g * sin(0.5 * pi * x[0]) ** 2 + sin(0.5 * pi * x[0]) * cos(p * pi * x[0]) ** 2 + sin(
+            0.5 * pi * x[1]) ** 2 + sin(0.5 * pi * x[1]) * cos(p * pi * x[1]) ** 2
+    if problemID == 'DF14':
+        G = sin(0.5 * pi * t)
+        g = 1 + sum((x[2:] - G) ** 2)
+        y = 0.5 + G * (x[0] - 0.5)
+        f[0] = g * (1 - y + 0.05 * sin(6 * pi * y))
+        f[1] = g * (1 - x[1] + 0.05 * sin(6 * pi * x[1])) * (y + 0.05 * sin(6 * pi * y))
+        f[2] = g * (x[1] + 0.05 * sin(6 * pi * x[1])) * (y + 0.05 * sin(6 * pi * y))
+    return f
+
+
+import numpy as np
+from numpy import pi, dot, floor, sin, cos, multiply, arange
+from copy import copy
+
+false = False
+true = True
+
+
+# cec2018_pf.m
+
+# ========================================================|#
+# PF calculation for 14 cec2018 test functions on         |
+# dynamic multiobjective optimisation. This document is   |
+# free to disseminate for academic use.                   |
+# --------------------------------------------------------|#
+# The "time" term in the test suite is defined as:        |
+#          t=1/nt*floor(tau/taut)                         |
+# where - nt:    severity of change                       |
+#       - taut:  frequency of change                      |
+#       - tau:   current generation counter               |
+# --------------------------------------------------------|#
+# Any questions can be directed to                        |
+#    Dr. Shouyong Jiang at math4neu@gmail.com.            |
+#                                                         |
+# ========================================================|#
+
+
+def cec2018_DF_PF(probID=None, t=1, n_points=100, *args, **kwargs):
+    # INPUT:
+    #       probID: test problem identifier (i.e. 'DF1')
+    #       tau:    current generation counter
+    #       taut:   frequency of change
+    #       nt:     severity of change
+
+    # OUTPUT:
+    #       h:      nondominated solutions
+
+    T0 = 50
+    g = 1
+    H = 50
+
+    if 'DF1' == (probID):
+        x = np.linspace(0, 1, n_points)
+        H = dot(0.75, sin(dot(dot(0.5, pi), t))) + 1.25
+        f1 = copy(x)
+        f2 = dot(g, (1 - (x / g) ** H))
+        h = get_PF(np.array([f1, f2]), false)
+    if 'DF2' == (probID):
+        x = np.linspace(0, 1, n_points)
+        G = abs(sin(dot(dot(0.5, pi), t)))
+        f1 = copy(x)
+        f2 = dot(g, (1 - (x / g) ** 0.5))  # To be sure
+        h = get_PF(np.array([f1, f2]), false)
+    if 'DF3' == (probID):
+        x = np.linspace(0, 1, n_points)
+        G = sin(dot(dot(0.5, pi), t))
+        H = G + 1.5
+        f1 = copy(x)
+        f2 = dot(g, (1 - (x / g) ** H))
+        h = get_PF(np.array([f1, f2]), false)
+    if 'DF4' == (probID):
+        a = sin(dot(dot(0.5, pi), t))
+        b = 1 + abs(cos(dot(dot(0.5, pi), t)))
+        x = np.linspace(a, a + b)
+        H = 1.5 + a
+        f1 = dot(g, abs(x - a) ** H)
+        f2 = dot(g, abs(x - a - b) ** H)  # Maybe
+        h = get_PF(np.array([f1, f2]), false)
+    if 'DF5' == (probID):
+        x = np.linspace(0, 1, n_points)
+        G = sin(dot(dot(0.5, pi), t))
+        w = floor(dot(10, G))
+        f1 = dot(g, (x + dot(0.02, sin(dot(dot(w, pi), x)))))
+        f2 = dot(g, (1 - x + dot(0.02, sin(dot(dot(w, pi), x)))))
+        h = get_PF(np.array([f1, f2]), false)
+    if 'DF6' == (probID):
+        x = np.linspace(0, 1, n_points)
+        G = sin(dot(dot(0.5, pi), t))
+        a = 0.2 + dot(2.8, abs(G))
+        f1 = dot(g, (x + dot(0.1, sin(dot(dot(3, pi), x)))) ** a)
+        f2 = dot(g, (1 - x + dot(0.1, sin(dot(dot(3, pi), x)))) ** a)
+        h = get_PF(np.array([f1, f2]), false)
+    if 'DF7' == (probID):
+        x = np.linspace(1, 4, n_points)
+        f1 = dot(g, (1 + t)) / x
+        f2 = dot(g, x) / (1 + t)
+        h = get_PF(np.array([f1, f2]), false)
+    if 'DF8' == (probID):
+        x = np.linspace(0, 1, n_points)
+        a = 2.25 + dot(2, cos(dot(dot(2, pi), t)))
+        f1 = dot(g, (x + dot(0.1, sin(dot(dot(3, pi), x)))))
+        f2 = dot(g, (1 - x + dot(0.1, sin(dot(dot(3, pi), x)))) ** a)
+        h = get_PF(np.array([f1, f2]), false)
+    if 'DF9' == (probID):
+        x = np.linspace(0, 1, n_points)
+        #        N = 1 + floor(dot(10, abs(sin(dot(dot(0.5, pi), t)))))
+
+        N = 1 + floor(10 * abs(sin(0.5 * pi * t)))
+        print(max(0, max((0.1 + 0.5 / N) * sin(2 * N * pi * x))))
+        f1 = g * (x + max(0, max((0.1 + 0.5 / N) * sin(2 * N * pi * x))))
+        f2 = g * (1 - x + max(0, max((0.1 + 0.5 / N) * sin(2 * N * pi * x))))
+
+        #        f1 = dot(g, (x + max(0, dot((0.1 + 0.5 / N), sin(dot(dot(dot(2, N), pi), x))))))
+        #        f2 = dot(g, (1 - x + max(0, dot((0.1 + 0.5 / N), sin(dot(dot(dot(2, N), pi), x))))))
+        h = get_PF(np.array([f1, f2]), true)
+    if 'DF10' == (probID):
+        x1, x2 = np.meshgrid(np.linspace(0, 1, H), np.linspace(0, 1, H), indexing='xy')
+        H = 2.25 + dot(2, cos(dot(dot(0.5, pi), t)))
+        f1 = dot(g, sin(dot(dot(0.5, pi), x1)) ** H)
+        f2 = multiply(dot(g, sin(dot(dot(0.5, pi), x2)) ** H),
+                      cos(dot(dot(0.5, pi), x1)) ** H)
+        f3 = multiply(dot(g, cos(dot(dot(0.5, pi), x2)) ** H),
+                      cos(dot(dot(0.5, pi), x1)) ** H)
+        h = get_PF(np.array([f1, f2, f3]), false)
+    if 'DF11' == (probID):
+        x1, x2 = np.meshgrid(np.linspace(0, 1, H), np.linspace(0, 1, H), indexing='xy')
+        G = abs(sin(dot(dot(0.5, pi), t)))
+        y1 = dot(pi, G) / 6 + dot((pi / 2 - dot(pi, G) / 3), x1)
+        y2 = dot(pi, G) / 6 + dot((pi / 2 - dot(pi, G) / 3), x2)
+        f1 = multiply(g, sin(y1))
+        f2 = dot(multiply(g, sin(y2)), cos(y1))
+        f3 = dot(multiply(g, cos(y2)), cos(y1))
+        h = get_PF(np.array([f1, f2, f3]), false)
+    if 'DF12' == (probID):
+        x1, x2 = np.meshgrid(np.linspace(0, 1, H), np.linspace(0, 1, H), indexing='xy')
+        k = dot(10, sin(dot(pi, t)))
+        tmp2 = abs(
+            multiply(sin(dot(floor(dot(k, (dot(2, x1) - 1))), pi) / 2),
+                     sin(dot(floor(dot(k, (dot(2, x2) - 1))), pi) / 2)))
+        g = 1 + tmp2
+        f1 = multiply(multiply(g, cos(dot(dot(0.5, pi), x2))),
+                      cos(dot(dot(0.5, pi), x1)))
+        f2 = multiply(multiply(g, sin(dot(dot(0.5, pi), x2))),
+                      cos(dot(dot(0.5, pi), x1)))
+        f3 = multiply(g, sin(dot(dot(0.5, pi), x1)))
+        h = get_PF(np.array([f1, f2, f3]), true)
+    if 'DF13' == (probID):
+        x1, x2 = np.meshgrid(np.linspace(0, 1, H), np.linspace(0, 1, H), indexing='xy')
+        G = sin(dot(dot(0.5, pi), t))
+        p = floor(dot(6, G))
+        f1 = multiply(g, cos(dot(dot(0.5, pi), x1)) ** 2)
+        f2 = multiply(g, cos(dot(dot(0.5, pi), x2)) ** 2)
+        f3 = multiply(g, sin(dot(dot(0.5, pi), x1)) ** 2) + multiply(
+            sin(dot(dot(0.5, pi), x1)),
+            cos(dot(dot(p, pi), x1)) ** 2) + sin(
+            dot(dot(0.5, pi), x2)) ** 2 + multiply(
+            sin(dot(dot(0.5, pi), x2)), cos(dot(dot(p, pi), x2)) ** 2)
+        h = get_PF(np.array([f1, f2, f3]), true)
+    if 'DF14' == (probID):
+        x1, x2 = np.meshgrid(np.linspace(0, 1, H), np.linspace(0, 1, H), indexing='xy')
+        G = sin(dot(dot(0.5, pi), t))
+        y = 0.5 + dot(G, (x1 - 0.5))
+        f1 = multiply(g, (1 - y + dot(0.05, sin(dot(dot(6, pi), y)))))
+        f2 = multiply(multiply(g, (
+                1 - x2 + dot(0.05, sin(dot(dot(6, pi), x2))))),
+                      (y + dot(0.05, sin(dot(dot(6, pi), y)))))
+        f3 = multiply(
+            multiply(g, (x2 + dot(0.05, sin(dot(dot(6, pi), x2))))),
+            (y + dot(0.05, sin(dot(dot(6, pi), y)))))
+        h = get_PF(np.array([f1, f2, f3]), false)
+    return h
+
+
+def get_PF(f=None, nondominate=None, *args, **kwargs):
+    ncell = len(f)
+    s = np.size(f[1])
+    h = []
+    for i in arange(ncell):
+        fi = np.reshape(f[i], s, order='F')
+        h.append(fi)
+    h = np.array(h).T
+    h = np.reshape(h, (s, ncell))
+
+    if nondominate:
+        print("Run Non dominating Sorting")
+        h = []
+        pass
+    #     in_ = get_skyline(h)
+    #     h = h(in_, arange())
+    return h