def varClusterFromMPS(path=os.path.join('MIPLIB', 'binkar10_1.mps'),
numClu=10,
varFilter=lambda x: x[0] == 'x'):
m = read(path)
if not varFilter:
klist = list(map(lambda var: var.VarName, m.getVars()))
else:
klist = list(
filter(varFilter, list(map(lambda var: var.VarName, m.getVars()))))
outerNbhs = {i: (0, ) for i in range(1, numClu + 1)}
labels = genClusterNeighborhoods(path=path,
nClusters=numClu,
verbose=True,
fNbhs=True,
varFilter=varFilter)
def funNbhs(varName, depth, param):
return depth != 1 + labels[varName]
return Neighborhoods(lowest=1,
highest=numClu,
keysList=klist,
outerNeighborhoods=outerNbhs,
useFunction=True,
randomSet=False,
funNeighborhoods=funNbhs)
def genNeighborhoods(self, verbose=False, classic=False):
klist = [
'x_{}_{}'.format(i, j) for i in range(self.m)
for j in range(self.n)
]
if not classic:
numClu = min(int(self.m * self.n / 10), 20)
print(numClu)
labelsDict = genClusterNeighborhoods(
self.pathMPS,
numClu,
fNbhs=True,
varFilter=lambda x: x[0] == 'x')
def fClusterNbhs(varName, depth, param):
return labelsDict[varName] != depth - 1
outerNbhs = {i: (0, ) for i in range(1, numClu + 1)}
return Neighborhoods(lowest=1,
highest=numClu,
keysList=klist,
randomSet=False,
outerNeighborhoods=outerNbhs,
useFunction=True,
funNeighborhoods=fClusterNbhs)
else:
nSecond = 4
outerNbhs = {
1: tuple([i for i in range(1, self.m + 1)]),
2: tuple([i for i in range(1, nSecond + 1)])
}
def fClassicNbhs(varName, depth, param):
elements = varName.split('_')
machine = int(elements[1])
job = int(elements[2])
if depth == 1:
return machine != param
elif depth == 2:
return (param - 1) * int(
self.n / nSecond) > job or job > param * int(
self.n / nSecond)
else:
return True
return Neighborhoods(lowest=1,
highest=2,
keysList=klist,
randomSet=False,
outerNeighborhoods=outerNbhs,
useFunction=True,
funNeighborhoods=fClassicNbhs)
def genNeighborhoods(self, varCluster=False, funNbhs=False):
if varCluster:
numClu = int(self.trucks * self.retailers / 7)
outerNbhs = {i: (0, ) for i in range(1, numClu + 1)}
labelsDict = genClusterNeighborhoods(
self.pathMPS,
numClu,
fNbhs=True,
varFilter=lambda x: x[0] == 'y')
def fClusterNbhs(varName, depth, param):
return labelsDict[varName] != depth - 1
klist = [
'y_{}_{}_{}'.format(i, j, k) for i in range(self.retailers + 1)
for j in range(self.retailers + 1)
for k in range(1, self.trucks + 1) if i < j
]
return Neighborhoods(lowest=1,
highest=numClu,
keysList=klist,
randomSet=False,
outerNeighborhoods=outerNbhs,
useFunction=True,
funNeighborhoods=fClusterNbhs)
if funNbhs:
def fNbhs(varName, depth, param):
elements = varName.split('_')
if len(elements) < 4:
return False
else:
kl = int(elements[3])
il = int(elements[1])
jl = int(elements[2])
if depth == 1:
return kl != param
elif depth == 2:
return kl != param[0] and kl != param[1]
else:
print('Error 23 Nbhds Function!! ')
return 0
return False
outer = {
1:
tuple([tf for tf in range(1, self.trucks + 1)]),
2:
tuple([(tr1, tr2) for tr1 in range(1, self.trucks + 1)
for tr2 in range(1, self.trucks + 1) if tr1 < tr2])
}
klist = [
'y_{}_{}_{}'.format(i, j, k)
for k in range(1, self.trucks + 1)
for i in range(self.retailers + 1)
for j in range(self.retailers + 1) if i < j
]
return Neighborhoods(lowest=1,
highest=2,
keysList=klist,
randomSet=False,
outerNeighborhoods=outer,
funNeighborhoods=fNbhs,
useFunction=True)
outerN = {
1: {
truck: [
'y_{}_{}_{}'.format(i, j, k)
for k in range(1, self.trucks + 1)
for i in range(self.totalNodes)
for j in range(self.totalNodes) if i < j and k != truck
]
for truck in range(1, self.trucks + 1)
},
2: {(tr1, tr2): [
'y_{}_{}_{}'.format(i, j, k)
for k in range(1, self.trucks + 1)
for i in range(self.totalNodes) for j in range(self.totalNodes)
if i < j and k != tr1 and k != tr2
]
for tr1 in range(1, self.trucks + 1)
for tr2 in range(1, self.trucks + 1) if tr1 < tr2}
}
return Neighborhoods(lowest=1,
highest=2,
keysList=None,
randomSet=False,
outerNeighborhoods=outerN)
def genNeighborhoods(self,
k=25,
Kvicinities=False,
funNbhs=False,
varCluster=False):
if varCluster:
numClu = int(self.H * self.V / 7)
outerNbhs = {i: (0, ) for i in range(1, numClu + 1)}
labelsDict = genClusterNeighborhoods(
self.pathMPS,
numClu,
fNbhs=True,
varFilter=lambda x: x[0] == 'x')
def fClusterNbhs(varName, depth, param):
return labelsDict[varName] != depth - 1
klist = [
'x_{}_{}_{}'.format(i, j, t) for i in range(self.V + 1)
for j in range(self.V + 1) for t in range(1, self.H + 1)
if i != j
]
return Neighborhoods(lowest=1,
highest=numClu,
keysList=klist,
randomSet=False,
outerNeighborhoods=outerNbhs,
useFunction=True,
funNeighborhoods=fClusterNbhs)
if funNbhs:
X = self.positions
# In larrain et al 2019 the neighbors parameter is set to 20.
nbrs = NearestNeighbors(n_neighbors=20,
algorithm='ball_tree').fit(X)
indices = nbrs.kneighbors(X)[1]
def fNbhs(varName, depth, param):
elements = varName.split('_')
if len(elements) < 4:
return False
else:
tl = int(elements[3])
il = int(elements[1])
jl = int(elements[2])
if depth == 1:
return tl != param
elif depth == 2:
return il not in indices[param] and jl not in indices[
param]
else:
print('Error 23 Nbhds Function!! ')
return 0
return True
outer = {
1: tuple([tf for tf in range(1, self.H + 1)]),
2: tuple([i for i in range(1, self.V + 1)])
}
klist = [
'x_{}_{}_{}'.format(i, j, t) for t in range(1, self.H + 1)
for i in range(self.V + 1) for j in range(self.V + 1) if i != j
]
return Neighborhoods(lowest=1,
highest=2,
keysList=klist,
randomSet=False,
outerNeighborhoods=outer,
funNeighborhoods=fNbhs,
useFunction=True)
if Kvicinities:
X = self.positions
nbrs = NearestNeighbors(n_neighbors=k,
algorithm='ball_tree').fit(X)
indices = nbrs.kneighbors(X)[1]
outerMVRPD = {
1: { # Period Neighborhood
tAct: tuple([
'x_{}_{}_{}'.format(i, j, t) for i in range(self.V + 1)
for j in range(self.V + 1)
for t in range(1, self.H + 1) if i != j and t != tAct
])
for tAct in range(1, self.H + 1)
},
2: { # k Vecinity Neighborhood
ip: tuple([
'x_{}_{}_{}'.format(i, j, t) for i in range(self.V + 1)
for j in range(self.V + 1)
for t in range(1, self.H + 1)
if i != j and (i in indices[ip] or j in indices[ip])
])
for ip in range(self.V + 1)
}
}
return Neighborhoods(lowest=1,
highest=2,
keysList=None,
randomSet=False,
outerNeighborhoods=outerMVRPD,
useFunction=False)
else:
X = self.positions
kmp = 3 # K means parameter
kmeans = KMeans(n_clusters=kmp, random_state=0).fit(X)
labels = kmeans.labels_
outerMVRPD = {
1: { # Period Neighborhood
tAct: tuple([
'x_{}_{}_{}'.format(i, j, t) for i in range(self.V + 1)
for j in range(self.V + 1)
for t in range(1, self.H + 1) if i != j and t != tAct
])
for tAct in range(1, self.H + 1)
},
2: { # K means Neighborhood
(selK, tAct): tuple([
'x_{}_{}_{}'.format(i, j, t) for i in range(self.V + 1)
for j in range(self.V + 1)
for t in range(1, self.H + 1) if i != j and t != tAct
and labels[i] != selK and labels[j] != selK
])
for selK in range(kmp) for tAct in range(1, self.H + 1)
}
}
return Neighborhoods(lowest=1,
highest=2,
keysList=None,
randomSet=False,
outerNeighborhoods=outerMVRPD)
def genNeighborhoods(self, setRandom=False, varCluster=False):
if varCluster:
numClu = int(self.n / 20) + 1
outerNbhs = {i: (0, ) for i in range(1, numClu + 1)}
labelsDict = genClusterNeighborhoods(
self.pathMPS,
numClu,
fNbhs=True,
varFilter=lambda x: x[0] == 'x')
def fClusterNbhs(varName, depth, param):
return labelsDict[varName] != depth - 1
klist = [
'x_{}_{}'.format(i, j) for i in range(self.n)
for j in range(self.n) if i < j
]
return Neighborhoods(lowest=1,
highest=numClu,
keysList=klist,
randomSet=False,
outerNeighborhoods=outerNbhs,
useFunction=True,
funNeighborhoods=fClusterNbhs)
if not setRandom:
outer = {}
groups = [5]
for k in range(1, 2):
kmeans = KMeans(n_clusters=groups[k - 1],
random_state=0).fit(self.positions)
labels = kmeans.labels_
params = {}
for i in range(groups[k - 1]):
indexes = [
index for index in range(self.n) if labels[index] == i
]
params[i] = [
'x_{}_{}'.format(i, j) for i in indexes
for j in indexes if i < j
]
outer[k] = params
nNbh = Neighborhoods(1,
2,
keysList=None,
randomSet=False,
outerNeighborhoods=outer)
return nNbh
else:
nNbh = Neighborhoods(1,
4,
keysList=[
'x_{}_{}'.format(i, j)
for i in range(self.n)
for j in range(self.n) if i < j
],
randomSet=True,
outerNeighborhoods=None)
return nNbh
def genNeighborhoods(self, varCluster=False, k=20, funNbhs=False):
if funNbhs:
def fNbhs(varName, depth, param):
elements = varName.split('_')
if len(elements) < 5:
return False
else:
kl = int(elements[3])
tl = int(elements[4])
if depth == 2:
return (kl, tl) != param
elif depth == 3:
return tl != param
elif depth == 4:
return kl != param
elif depth == 5:
return tl != param[0] and tl != param[1]
else:
print('Error 23 Nbhds Function!! ')
return 0
return False
outer = {
2:
tuple([(kf, tf) for kf in range(1, self.K + 1)
for tf in range(1, self.H + 1)]),
3:
tuple([tf for tf in range(1, self.H + 1)]),
4:
tuple([kf for kf in range(1, self.K + 1)]),
5:
tuple([(tf1, tf2) for tf1 in range(1, self.H + 1)
for tf2 in range(1, self.H + 1) if tf1 < tf2])
}
klist = [
'x_{}_{}_{}_{}'.format(i, j, k, t)
for k in range(1, self.K + 1) for t in range(1, self.H + 1)
for i in range(self.V + 1) for j in range(self.V + 1) if i < j
]
return Neighborhoods(lowest=2,
highest=5,
keysList=klist,
randomSet=False,
outerNeighborhoods=outer,
funNeighborhoods=fNbhs,
useFunction=True)
if varCluster:
amountClusters = self.H * self.K
outerNbhs = {i: (0, ) for i in range(1, amountClusters + 1)}
labelsDict = genClusterNeighborhoods(
self.pathMPS,
amountClusters,
fNbhs=True,
varFilter=lambda x: x[0] == 'x')
def fClusterNbhs(varName, depth, param):
return labelsDict[varName] == depth
klist = [
'x_{}_{}_{}_{}'.format(i, j, k, t)
for k in range(1, self.K + 1) for t in range(1, self.H + 1)
for i in range(self.V + 1) for j in range(self.V + 1) if i < j
]
return Neighborhoods(lowest=1,
highest=amountClusters,
keysList=klist,
randomSet=False,
outerNeighborhoods=outerNbhs,
useFunction=True,
funNeighborhoods=fClusterNbhs)
outerNhs = {
2: {(kf, tf): [
'{}_{}_{}_{}_{}'.format('x', i, j, k, t)
for k in range(1, self.K + 1) for t in range(1, self.H + 1)
for i in range(self.V + 1) for j in range(self.V + 1)
if i < j and (k, t) != (kf, tf)
]
for kf in range(1, self.K + 1) for tf in range(1, self.H + 1)},
3: {
tf: [
'{}_{}_{}_{}_{}'.format('x', i, j, k, t)
for k in range(1, self.K + 1)
for t in range(1, self.H + 1) for i in range(self.V + 1)
for j in range(self.V + 1) if i < j and t != tf
]
for tf in range(1, self.H + 1)
},
4: {
kf: [
'{}_{}_{}_{}_{}'.format('x', i, j, k, t)
for k in range(1, self.K + 1)
for t in range(1, self.H + 1) for i in range(self.V + 1)
for j in range(self.V + 1) if i < j and k != kf
]
for kf in range(1, self.K + 1)
},
5: {(tf1, tf2): [
'{}_{}_{}_{}_{}'.format('x', i, j, k, t)
for k in range(1, self.K + 1) for t in range(1, self.H + 1)
for i in range(self.V + 1) for j in range(self.V + 1)
if i < j and tf1 != t and tf2 != t
]
for tf1 in range(1, self.H + 1)
for tf2 in range(1, self.H + 1) if tf1 < tf2}
}
return Neighborhoods(lowest=2,
highest=5,
keysList=None,
randomSet=False,
outerNeighborhoods=outerNhs,
useFunction=False)