def process(idnum, t, chromosome, bpp, items):
# Process solutions (decode, check constraints, calculate
# fitness values, make solution object)
x, y = bp.ed(idnum, chromosome, bpp, items)
concheck(idnum, x, bpp)
fit = mop.calcfits(x, y, items)
a = MultiSol(idnum, chromosome, x, y, t, fit, 0, 0.0)
return a
def subfit(self, k, chrom, bpp, method):
# This function calculates the fitness value for subprob. k
# Options: 'ed', 'll', 'dp', 'combo'
if method == 'ed':
x, y = bp.ed(0, chrom, bpp, self.items)
else:
x, y = bp.ll(chrom, bpp.getwbin(), bpp.getub(), bpp.getlb(), self.items)
subfit = np.dot(self.lambdas[k], self.calcfits(x, y))
return x, y, subfit
def makeq(self):
if self.t == 0:
new = range(self.pop)
else:
new, self.q = sols.oldnew(self.archive, self.q, self.newgenes)
# Make new solutions in q
for m in new:
x, y = bp.ed(self.idnum, self.newgenes[m], self.bpp)
self.addnewsol(x, y, self.newgenes[m])
def rungen(self):
if self.t == 0:
newgenes = self.initialize()
for m in range(self.pop):
x, y = bp.ed(m, newgenes[m], self.bpp, self.items)
self.swarm.append(Particle(self.n, x, y, newgenes[m]))
self.t += 1
print('t = ', self.t)
self.fitevalarch()
self.psooperator()
self.mutation()
self.heuristic()
print(self.funkeval, 'function evaluations have been performed.\n')
def perturb(self, xk):
# This function performs the pertubation in Step 3.1 of MOMAD.
alpha = random.random()
if alpha == 0:
alpha = random.random()
blocklen = int(alpha * self.n)
splitj = self.n - blocklen
cut1 = xk.getgenes()[:splitj]
cut2 = xk.getgenes()[splitj:]
cut3 = random.sample(cut2, len(cut2))
newgene = cut1 + cut3
x, y = bp.ed(self.idnum, newgene, self.bpp, self.items)
fit = self.moop.calcfits(x, y)
xksquiggly = sol.MultiSol(self.idnum, newgene, x, y, self.t, fit, 0, 0.0)
self.updateid()
self.updatefe()
return xksquiggly
def neighborhood(self, m, hoodsize):
# This function takes solution m and finds all of its neighbors.
# input: m
# output: the set of m's neighbors
# Limiting neighborhood to 100 neighbors to reduce computationload
neighbors = []
while len(neighbors) < hoodsize:
newgene = m.getgenes()
a = random.randrange(0, len(newgene))
b = random.randrange(0, len(newgene))
newgene[a], newgene[b] = newgene[b], newgene[a]
x, y = bp.ed(self.idnum, newgene, self.bpp, self.items)
fit = self.moop.calcfits(x, y)
a = sol.MultiSol(self.idnum, newgene, x, y, self.t, fit, 0, 0.0)
neighbors.append(a)
self.updateid()
self.updatefe()
return neighbors
def makeqbybins(self):
# This function is almost the same as makeq() except modified for
# the bin packing operators.
# Sort out what solutions were modified by the crossover operation
new, self.q = sols.oldnew(self.archive, self.q, self.newgenes)
# First go through what's left in q
for m in range(len(self.q)):
qx = self.q[m].getx()
qy = self.q[m].gety()
x, y = self.binmutation(qx, qy)
if np.all(np.equal(x, qx)) is False: # Indicates new solution
del self.q[m]
newchrom = self.updatechrom(x)
self.addnewsol(x, y, newchrom)
# Then operate on new solutions
for m in new:
newx, newy = bp.ed(self.idnum, self.newgenes[m], self.bpp)
x, y = self.binmutation(newx, newy)
if np.all(np.equal(x, newx)) is False:
self.newgenes[m] = self.updatechrom(x)
self.addnewsol(x, y, self.newgenes[m])