class DE():
def __init__(self, model="DTLZ7"):
self.top_bound = [3.6, 0.8, 28.0, 8.3, 8.3, 3.9, 5.5]
self.bottom_bound = [2.6, 0.7, 17.0, 7.3, 7.3, 2.9, 5]
self.f2_high = -10**6
self.f2_low = 10**6
self.f1_high = -10**6
self.f1_low = 10**6
self.evals = 0
self.median = []
if model == "Osyczka":
self.model = Osyczka()
elif model == "Golinski":
self.model = Golinski()
elif model == "Kursawe":
self.model = Kursawe()
elif model == "Schaffer":
self.model = Schaffer()
elif model == "DTLZ7":
self.model = DTLZ7(10, 2)
self.best_solution = Thing()
self.best_solution.score = 0
self.best_solution.energy = 1
self.best_solution.have = []
self.previous_era = []
#Returns three different things that are not 'avoid'.
def threeOthers(self, lst, avoid):
def oneOther():
x = self.a(lst)
while x in seen:
x = self.a(lst)
seen.append(x)
return x
# -----------------------
seen = list(avoid.have)
this = oneOther()
that = oneOther()
theOtherThing = oneOther()
return this, that, theOtherThing
def a(self, lst):
return lst[self.n(len(lst))]
def n(self, number):
return int(uniform(0, number))
def candidate(self):
# something = [uniform(self.model.low(d), self.model.high(d))
# for d in self.model.decisions()]
something = self.model.get_random_state()
new = Thing()
new.have = something
new.score = self.model.energy(new.have)
new.energy = self.model.energy(new.have)
return new
def run(
self,
maximum=100, # number of repeats
np=100, # number of candidates
f=0.75, # extrapolate amount
cf=0.3, # prob of cross-over
epsilon=0.01,
s=0.1):
print "asd" + str(np)
frontier = [self.candidate() for _ in range(np)]
print "test"
median = []
print maximum
for k in range(0, 100):
total, n, output = self.update(f, cf, frontier)
self.evals += n
frontier.sort(key=lambda x: x.energy)
#print output + "Frontier energy:" + str(total) + " Count:" + str(n) + " Max Energy:" + str(frontier[0].energy) + " Min Energy:" + str(frontier[len(frontier)-1].energy)
min = frontier[0].energy
max = frontier[len(frontier) - 1].energy
median.append(frontier[int(len(frontier) / 2)].score)
big = max - (max - min) * s / 100
new_frontier = [x for x in frontier if x.energy <= big]
frontier = new_frontier
if (n > 0 and total / n >
(1 - epsilon)) or n <= 0 or len(frontier) < 3:
break
print "Median values:"
print median
return self.previous_era
def update(self, f, cf, frontier, total=0.0, n=0):
output = ""
MAX_LIVES = 10
ERA_LENGTH = 10
era_List = []
current_era = []
lives = MAX_LIVES
for x in frontier:
s = x.energy
new = self.extrapolate(frontier, x, f, cf)
# if self.model.type1(x.have, new.have):
# output += "."
if s < new.energy:
output += "."
# elif self.model.type1(new.have, x.have):
# x.energy = new.energy
# x.score = new.score
# x.have = new.have
# output += "+"
elif new.energy < s:
x.energy = new.energy
x.score = new.score
x.have = new.have
output += "+"
# if self.model.type1(new.have, self.best_solution.have):
# self.best_solution.score = new.score
# self.best_solution.have = new.have
# self.best_solution.energy = new.energy
# self.best_solution.evals = self.evals + n
# output += "!"
if new.energy < self.best_solution.energy:
self.best_solution.score = new.score
self.best_solution.have = new.have
self.best_solution.energy = new.energy
self.best_solution.evals = self.evals + n
output += "!"
if len(output) == 50:
print output + " Best Solution: [",
for a in self.best_solution.have:
print("%.2f " % a),
print "] Energy: " + str(
self.best_solution.score) + " Evals: " + str(self.evals +
n)
current_era.append(x.have)
if len(current_era) == ERA_LENGTH and len(era_List) > 0:
increment = self.model.type2(current_era, era_List[-1])
lives += increment
if lives <= 0:
self.previous_era = era_List[-1]
print "Early termination"
break
era_List.append(current_era)
current_era = []
n += 1
total += x.energy
return total, n, output
def extrapolate(self, frontier, one, f, cf):
out = Thing(id=one.id,
have=list(one.have),
score=self.model.energy(one.have),
energy=self.model.energy(one.have))
two, three, four = self.threeOthers(frontier, one)
changed = False
for d in self.model.decisions():
x, y, z = two.have[d], three.have[d], four.have[d]
if random() < cf:
changed = True
new = x + f * (y - z)
out.have[d] = max(0.0, min(new, 1.0)) # keep in range
if not changed:
d = self.a(self.model.decisions())
out.have[d] = two.have[d]
out.score = self.model.energy(out.have) # remember to score it
out.energy = self.model.energy(out.have)
return out
class DE():
def __init__(self, model = "DTLZ7"):
self.top_bound = [3.6, 0.8, 28.0, 8.3, 8.3, 3.9, 5.5]
self.bottom_bound = [2.6, 0.7, 17.0, 7.3, 7.3, 2.9, 5]
self.f2_high = -10**6
self.f2_low = 10**6
self.f1_high = -10**6
self.f1_low = 10**6
self.evals = 0
self.median = []
if model == "Osyczka":
self.model = Osyczka()
elif model == "Golinski":
self.model = Golinski()
elif model == "Kursawe":
self.model = Kursawe()
elif model == "Schaffer":
self.model = Schaffer()
elif model == "DTLZ7":
self.model = DTLZ7(10, 2)
self.best_solution = Thing()
self.best_solution.score = 0;
self.best_solution.energy = 1;
self.best_solution.have = []
self.previous_era = []
#Returns three different things that are not 'avoid'.
def threeOthers(self, lst, avoid):
def oneOther():
x = self.a(lst)
while x in seen:
x = self.a(lst)
seen.append( x )
return x
# -----------------------
seen = list(avoid.have)
this = oneOther()
that = oneOther()
theOtherThing = oneOther()
return this, that, theOtherThing
def a(self, lst):
return lst[self.n(len(lst))]
def n(self, number):
return int(uniform(0, number))
def candidate(self):
# something = [uniform(self.model.low(d), self.model.high(d))
# for d in self.model.decisions()]
something = self.model.get_random_state()
new = Thing()
new.have = something
new.score = self.model.energy(new.have)
new.energy = self.model.energy(new.have)
return new
def run(self, maximum = 100, # number of repeats
np = 100, # number of candidates
f = 0.75, # extrapolate amount
cf = 0.3, # prob of cross-over
epsilon = 0.01,
s = 0.1
):
print "asd" + str(np)
frontier = [self.candidate() for _ in range(np)]
print "test"
median = []
print maximum
for k in range(0, 100):
total, n, output = self.update(f, cf, frontier)
self.evals += n
frontier.sort(key=lambda x: x.energy)
#print output + "Frontier energy:" + str(total) + " Count:" + str(n) + " Max Energy:" + str(frontier[0].energy) + " Min Energy:" + str(frontier[len(frontier)-1].energy)
min = frontier[0].energy
max = frontier[len(frontier)-1].energy
median.append(frontier[int(len(frontier)/2)].score)
big = max - (max-min)*s/100
new_frontier = [x for x in frontier if x.energy <= big]
frontier = new_frontier
if (n > 0 and total/n > (1 - epsilon)) or n <= 0 or len(frontier) < 3:
break
print "Median values:"
print median
return self.previous_era
def update(self, f, cf, frontier, total = 0.0, n = 0):
output = ""
MAX_LIVES = 10
ERA_LENGTH = 10
era_List = []
current_era = []
lives = MAX_LIVES
for x in frontier:
s = x.energy
new = self.extrapolate(frontier, x, f, cf)
# if self.model.type1(x.have, new.have):
# output += "."
if s < new.energy:
output += "."
# elif self.model.type1(new.have, x.have):
# x.energy = new.energy
# x.score = new.score
# x.have = new.have
# output += "+"
elif new.energy < s:
x.energy = new.energy
x.score = new.score
x.have = new.have
output += "+"
# if self.model.type1(new.have, self.best_solution.have):
# self.best_solution.score = new.score
# self.best_solution.have = new.have
# self.best_solution.energy = new.energy
# self.best_solution.evals = self.evals + n
# output += "!"
if new.energy < self.best_solution.energy:
self.best_solution.score = new.score
self.best_solution.have = new.have
self.best_solution.energy = new.energy
self.best_solution.evals = self.evals + n
output += "!"
if len(output) == 50:
print output + " Best Solution: [",
for a in self.best_solution.have:
print("%.2f " % a),
print "] Energy: " + str(self.best_solution.score) + " Evals: " + str(self.evals + n)
current_era.append(x.have)
if len(current_era) == ERA_LENGTH and len(era_List) > 0:
increment = self.model.type2(current_era, era_List[-1])
lives += increment
if lives <= 0:
self.previous_era = era_List[-1]
print "Early termination"
break
era_List.append(current_era)
current_era = []
n += 1
total += x.energy
return total, n, output
def extrapolate(self, frontier, one, f, cf):
out = Thing(id = one.id, have = list(one.have), score = self.model.energy(one.have), energy = self.model.energy(one.have))
two, three, four = self.threeOthers(frontier, one)
changed = False
for d in self.model.decisions():
x, y, z = two.have[d], three.have[d], four.have[d]
if random() < cf:
changed = True
new = x + f*(y - z)
out.have[d] = max(0.0, min(new, 1.0)) # keep in range
if not changed:
d = self.a(self.model.decisions())
out.have[d] = two.have[d]
out.score = self.model.energy(out.have) # remember to score it
out.energy = self.model.energy(out.have)
return out
class DE():
def __init__(self, model = "Osyczka"):
self.top_bound = [3.6, 0.8, 28.0, 8.3, 8.3, 3.9, 5.5]
self.bottom_bound = [2.6, 0.7, 17.0, 7.3, 7.3, 2.9, 5]
self.f2_high = -10**6
self.f2_low = 10**6
self.f1_high = -10**6
self.f1_low = 10**6
self.evals = 0
self.median = []
if model == "Osyczka":
self.model = Osyczka()
elif model == "Golinski":
self.model = Golinski()
elif model == "Kursawe":
self.model = Kursawe()
elif model == "Schaffer":
self.model = Schaffer()
self.best_solution = Thing()
self.best_solution.score = 0;
self.best_solution.energy = 1;
self.best_solution.have = []
#Returns three different things that are not 'avoid'.
def threeOthers(self, lst, avoid):
def oneOther():
x = self.a(lst)
while x in seen:
x = self.a(lst)
seen.append( x )
return x
# -----------------------
seen = list(avoid.have)
this = oneOther()
that = oneOther()
theOtherThing = oneOther()
return this, that, theOtherThing
def a(self, lst):
return lst[self.n(len(lst))]
def n(self, number):
return int(uniform(0, number))
def candidate(self):
something = [uniform(self.model.low(d), self.model.high(d))
for d in self.model.decisions()]
new = Thing()
new.have = something
new.score = self.model.energy(new.have)
new.energy = self.model.aggregate_energy(new.have)
return new
def run(self, max = 100, # number of repeats
np = 100, # number of candidates
f = 0.75, # extrapolate amount
cf = 0.3, # prob of cross-over
epsilon = 0.01,
s = 0.1
):
frontier = [self.candidate() for _ in range(np)]
median = []
for k in range(max):
total, n, output = self.update(f, cf, frontier)
self.evals += n
frontier.sort(key=lambda x: x.energy)
#print output + "Frontier energy:" + str(total) + " Count:" + str(n) + " Max Energy:" + str(frontier[0].energy) + " Min Energy:" + str(frontier[len(frontier)-1].energy)
min = frontier[0].energy
max = frontier[len(frontier)-1].energy
median.append(frontier[int(len(frontier)/2)].energy)
big = max - (max-min)*s/100
new_frontier = [x for x in frontier if x.energy <= big]
frontier = new_frontier
if (n > 0 and total/n > (1 - epsilon)) or n <= 0 or len(frontier) < 3:
break
# print "Median values:"
# print median
return self.best_solution.have
def update(self, f, cf, frontier, total = 0.0, n = 0):
output = ""
for x in frontier:
s = x.energy
new = self.extrapolate(frontier, x, f, cf)
if s < new.energy:
output += "."
elif new.energy < s:
x.energy = new.energy
x.score = new.score
x.have = new.have
output += "+"
if new.energy < self.best_solution.energy:
self.best_solution.score = new.score
self.best_solution.have = new.have
self.best_solution.energy = new.energy
self.best_solution.evals = self.evals + n
output += "!"
if len(output) == 50:
print output + " Best Solution: [",
for a in self.best_solution.have:
print("%.2f " % a),
print "] Energy: " + str(self.best_solution.score) + " Evals: " + str(self.evals + n)
n += 1
total += x.energy
return total, n, output
def extrapolate(self, frontier, one, f, cf):
out = Thing(id = one.id, have = list(one.have), score = self.model.energy(one.have), energy = self.model.aggregate_energy(one.have))
two, three, four = self.threeOthers(frontier, one)
changed = False
for d in self.model.decisions():
x, y, z = two.have[d], three.have[d], four.have[d]
if random() < cf:
changed = True
new = x + f*(y - z)
out.have[d] = self.model.trim(new, d) # keep in range
if not changed:
d = self.a(self.model.decisions())
out.have[d] = two.have[d]
out.score = self.model.energy(out.have) # remember to score it
out.energy = self.model.aggregate_energy(out.have)
return out
class DE():
def __init__(self, model="Osyczka"):
self.top_bound = [3.6, 0.8, 28.0, 8.3, 8.3, 3.9, 5.5]
self.bottom_bound = [2.6, 0.7, 17.0, 7.3, 7.3, 2.9, 5]
self.f2_high = -10**6
self.f2_low = 10**6
self.f1_high = -10**6
self.f1_low = 10**6
self.evals = 0
self.median = []
if model == "Osyczka":
self.model = Osyczka()
elif model == "Golinski":
self.model = Golinski()
elif model == "Kursawe":
self.model = Kursawe()
elif model == "Schaffer":
self.model = Schaffer()
self.best_solution = Thing()
self.best_solution.score = 0
self.best_solution.energy = 1
self.best_solution.have = []
#Returns three different things that are not 'avoid'.
def threeOthers(self, lst, avoid):
def oneOther():
x = self.a(lst)
while x in seen:
x = self.a(lst)
seen.append(x)
return x
# -----------------------
seen = list(avoid.have)
this = oneOther()
that = oneOther()
theOtherThing = oneOther()
return this, that, theOtherThing
def a(self, lst):
return lst[self.n(len(lst))]
def n(self, number):
return int(uniform(0, number))
def candidate(self):
something = [
uniform(self.model.low(d), self.model.high(d))
for d in self.model.decisions()
]
new = Thing()
new.have = something
new.score = self.model.energy(new.have)
new.energy = self.model.aggregate_energy(new.have)
return new
def run(
self,
max=100, # number of repeats
np=100, # number of candidates
f=0.75, # extrapolate amount
cf=0.3, # prob of cross-over
epsilon=0.01,
s=0.1):
frontier = [self.candidate() for _ in range(np)]
median = []
for k in range(max):
total, n, output = self.update(f, cf, frontier)
self.evals += n
frontier.sort(key=lambda x: x.energy)
#print output + "Frontier energy:" + str(total) + " Count:" + str(n) + " Max Energy:" + str(frontier[0].energy) + " Min Energy:" + str(frontier[len(frontier)-1].energy)
min = frontier[0].energy
max = frontier[len(frontier) - 1].energy
median.append(frontier[int(len(frontier) / 2)].energy)
big = max - (max - min) * s / 100
new_frontier = [x for x in frontier if x.energy <= big]
frontier = new_frontier
if (n > 0 and total / n >
(1 - epsilon)) or n <= 0 or len(frontier) < 3:
break
# print "Median values:"
# print median
return self.best_solution.have
def update(self, f, cf, frontier, total=0.0, n=0):
output = ""
for x in frontier:
s = x.energy
new = self.extrapolate(frontier, x, f, cf)
if s < new.energy:
output += "."
elif new.energy < s:
x.energy = new.energy
x.score = new.score
x.have = new.have
output += "+"
if new.energy < self.best_solution.energy:
self.best_solution.score = new.score
self.best_solution.have = new.have
self.best_solution.energy = new.energy
self.best_solution.evals = self.evals + n
output += "!"
if len(output) == 50:
print output + " Best Solution: [",
for a in self.best_solution.have:
print("%.2f " % a),
print "] Energy: " + str(
self.best_solution.score) + " Evals: " + str(self.evals +
n)
n += 1
total += x.energy
return total, n, output
def extrapolate(self, frontier, one, f, cf):
out = Thing(id=one.id,
have=list(one.have),
score=self.model.energy(one.have),
energy=self.model.aggregate_energy(one.have))
two, three, four = self.threeOthers(frontier, one)
changed = False
for d in self.model.decisions():
x, y, z = two.have[d], three.have[d], four.have[d]
if random() < cf:
changed = True
new = x + f * (y - z)
out.have[d] = self.model.trim(new, d) # keep in range
if not changed:
d = self.a(self.model.decisions())
out.have[d] = two.have[d]
out.score = self.model.energy(out.have) # remember to score it
out.energy = self.model.aggregate_energy(out.have)
return out
