def Egalitarian_Sexequal_3(mpref, wpref, algo):
mrank, wrank = changeStructure(mpref, wpref, n1, n2)
#Initialize M
M, f = LTIU2(mpref, wpref)
Mbest = M
bestCost = 2 * n1 * n2
iters = 500
clist = []
start = time.time()
while iters and time.time() - start < 2000:
f = F(M, mpref, wpref, mrank, wrank)
if f != 0:
M, f = LTIU2(mpref, wpref)
else:
cost = eg_cost(M, mrank, wrank) if algo == "eg" else se_cost(
M, mrank, wrank)
clist.append(cost)
if cost < bestCost:
bestCost = cost
Mbest = M
if cost == 0:
return M, cost
else:
neighbors = BreakMarriageSMTI(M, mrank, wrank, mpref, wpref)
print("neigbors")
for ne in neighbors:
print(ne)
randwalk = True if random.random() <= 0.2 else False
if randwalk:
M = neighbors[random.randint(0, len(neighbors) - 1)]
else:
costs = [eg_cost(nei, mrank, wrank)
for nei in neighbors] if algo == "eg" else [
se_cost(nei, mrank, wrank)
for nei in neighbors
]
min_cost = min(costs)
if min_cost < cost:
indices = [
i for i, v in enumerate(costs) if v == min_cost
]
M = neighbors[random.choice(indices)]
else:
M = neighbors[random.randint(0, len(neighbors) - 1)]
iters -= 1
plt.plot(clist)
plt.show()
return_cost = eg_cost(Mbest, mrank, wrank) if algo == "eg" else se_cost(
Mbest, mrank, wrank)
return Mbest, return_cost
def LCV_deney(n1, n2, mpref, wpref, iters):
n = n1
mrank, wrank = changeStructure(mpref, wpref, n, n)
arraysize = 2 * n
M = [-1] * arraysize
M = random_match2(M, mrank, n)
Mbest = M
iterstart = iters
y = []
while iters:
if ns2(Mbest) == 0:
break
X = findbps2(M, mrank, wrank, n)
y.append(len(X) + ns2(Mbest))
if len(X) == 0:
#M is stable
#random restart
if ns2(M) < ns2(Mbest):
Mbest = M
M = [-1] * arraysize
M = random_match2(M, mrank, n)
else:
m = man_that_generates_max_bp(M, mrank, wrank, n)
randWalk = random.random()
if randWalk <= 0.2:
pIndex = random.randint(0, len(mpref[m]) - 1)
wIndex = random.randint(0, len(mpref[m][pIndex]) - 1)
w = mpref[m][pIndex][wIndex]
else:
w = random_best_woman_for_m(mpref, m)
M = RemoveBP2(M, m, w, n)
iters -= 1
ns_ = ns2(Mbest)
nbp_ = nbp2(Mbest, mrank, wrank, n)
steptotal = iterstart - iters
return ns_ + nbp_, nbp_, ns_, steptotal, y
def LTIU2(mpref, wpref):
n1 = len(mpref)
n2 = len(wpref)
mrank, wrank = changeStructure(mpref, wpref, n1, n2)
arraysize = (n1 + n2)
M = [-1] * arraysize
M = random_match(M, mrank, n1, n2)
Fprev = F(M, mpref, wpref, mrank, wrank)
Fbest = Fprev
stableFound = False
NSbest = 2 * n1 * n2
Mbest = M
steps = 5000
print("Random Restart (LTIU)")
while steps:
nbp, ns = NBP_NS(M, mpref, wpref, mrank, wrank)
#nbp = NBP(M, mrank, wrank)
#ns = NS(M)
f = nbp + ns
if stableFound == False and f < Fbest:
Fbest = f
Mbest = M
#If a perfect matching found
if f == 0:
Mbest = M
Fbest = 0 #f
break
#If a stable (but not perfect) matching found
if nbp == 0:
stableFound = True
#Is it the best Stable matching among stable matchings found so far?
if ns < NSbest:
Mbest = M
NSbest = ns
Fbest = f # = ns
M = random_match(M, mrank, n1, n2)
#If unstable
else:
#Find neighbors
undominated_bps = UND(M, mrank, wrank, steps % 2)
neis = []
for un in undominated_bps:
copyM = M[:]
copyM = RemoveBP(copyM, un, n1)
neis.append(copyM)
random_walk = True if random.random() <= 0.2 else False
if random_walk:
M = neis[random.randint(0,
len(neis) -
1)] #Choose a random neighbor
else:
#Choose a neighbor with the minimum evaluation value
evaluations = [
F(nei, mpref, wpref, mrank, wrank) for nei in neis
]
min_eval = min(evaluations)
if min_eval <= f: #if there is an improvement
indices = [
i for i, v in enumerate(evaluations) if v == min_eval
]
M = neis[random.choice(indices)]
else: #no improvement, choose a random neighbor
M = neis[random.randint(0, len(neis) - 1)]
steps -= 1
nbp, ns = NBP_NS(Mbest, mpref, wpref, mrank, wrank)
f = ns + nbp
return Mbest, f
evaluations = [F(nei, mpref, wpref, mrank, wrank) for nei in neis]
min_eval = min(evaluations)
if min_eval <= f: #if there is an improvement
indices = [i for i, v in enumerate(evaluations) if v == min_eval]
M = neis[random.choice(indices)]
else: #no improvement, choose a random neighbor
M = neis[random.randint(0, len(neis)-1)]
steps -= 1
nbp, ns = NBP_NS(Mbest, mpref, wpref, mrank, wrank)
f = ns + nbp
return Mbest, f
n1 = 20
n2 = n1
mpref, wpref = GentProsser(n1, n2, 0.2, 0.5)
mrank, wrank = changeStructure(mpref, wpref, n1, n2)
#Initialize M
M, f = LTIU2(mpref, wpref)
iters = 1000
Mbest = M
best_cost = 2*n1*n2
while iters:
f = F(M, mpref, wpref, mrank, wrank)
if f != 0:
M, f = LTIU2(mpref, wpref)
else:
cost = eg_cost(M, mrank, wrank)
if cost == 0:
print(M) #return M
def LTIU(mpref, wpref):
n1 = len(mpref)
n2 = len(wpref)
mrank, wrank = changeStructure(mpref, wpref, n1, n2)
arraysize = (n1 + n2)
M = [-1] * arraysize
M = random_match(M, mrank, n1, n2)
Fprev = F(M, mpref, wpref, mrank, wrank)
Fbest = Fprev
stableFound = False
NSbest = 2 * n1 * n2
Mbest = M
steps = 10000
begin = steps
y = []
while steps:
print("| LTIU")
print("| step:", begin - steps)
nbp, ns = NBP_NS(M, mpref, wpref, mrank, wrank)
#nbp = NBP(M, mrank, wrank)
#ns = NS(M)
f = nbp + ns
print("| f:", f)
print("-------------")
y.append(f)
if stableFound == False and f < Fbest:
Fbest = f
Mbest = M
#If a perfect matching found
if f == 0:
Mbest = M
Fbest = 0 #f
break
#If a stable (but not perfect) matching found
if nbp == 0:
stableFound = True
#Is it the best Stable matching among stable matchings found so far?
if ns < NSbest:
Mbest = M
NSbest = ns
Fbest = f # = ns
M = random_match(M, mrank, n1, n2)
#If unstable
else:
#Find neighbors
undominated_bps = UND(M, mrank, wrank, steps % 2)
neis = []
for un in undominated_bps:
copyM = M[:]
copyM = RemoveBP(copyM, un, n1)
neis.append(copyM)
random_walk = True if random.random() <= 0.2 else False
if random_walk:
M = neis[random.randint(0,
len(neis) -
1)] #Choose a random neighbor
else:
#Choose a neighbor with the minimum evaluation value
evaluations = [
F(nei, mpref, wpref, mrank, wrank) for nei in neis
]
min_eval = min(evaluations)
if min_eval <= f: #if there is an improvement
indices = [
i for i, v in enumerate(evaluations) if v == min_eval
]
M = neis[random.choice(indices)]
else: #no improvement, choose a random neighbor
M = neis[random.randint(0, len(neis) - 1)]
steps -= 1
"""
plt.plot(y)
plt.title("LTIU n:" + str(n1) + ", p1:" + str(p1) + ", p2:"+ str(p2))
plt.xlabel("steps")
plt.ylabel("f = nbp + ns")
plt.show()
"""
if Fbest == 0:
print("Perfect matching found")
elif stableFound == True:
print("Stable matching found")
else:
print("Nonstable matching found")
#print(Mbest)
nbp, ns = NBP_NS(Mbest, mpref, wpref, mrank, wrank)
f = ns + nbp
print("ns:", ns, " nbp:", nbp, " f:", f, " steps:", begin - steps)
print("------------------------------")
return f, nbp, ns, begin - steps, y
def Egalitarian_SexEqual(mpref, wpref, steps, algo):
n1 = len(mpref)
n2 = len(wpref)
mrank, wrank = changeStructure(mpref, wpref, n1, n2)
arraysize = (n1 + n2)
M = [-1] * arraysize
M = random_match(M, mrank, n1, n2)
stableFound = False
Mbest = M
begin = steps
fs_return = []
costs_return = []
perfectFound = False
stableFound = False
while steps:
print("| step:", begin - steps)
nbp, ns = NBP_NS(M, mpref, wpref, mrank, wrank)
cost_val = se_cost(M, mrank, wrank) if algo == "se" else eg_cost(
M, mrank, wrank)
f = nbp + ns
#print("| f:", f)
fs_return.append(f)
costs_return.append(cost_val)
if f == 0 and cost_val == 0:
Mbest = M
break
if nbp == 0:
if ns == 0:
if perfectFound == True:
compare_cost = se_cost(Mbest, mrank,
wrank) if algo == "se" else eg_cost(
Mbest, mrank, wrank)
if compare_cost < cost_val:
Mbest = M
else:
Mbest = M
perfectFound = True
else:
if perfectFound == False:
if stableFound == True:
if ns < F(Mbest, mpref, wpref, mrank, wrank):
Mbest = M
else:
Mbest = M
stableFound = True
M = random_match(M, mrank, n1, n2)
else:
undominated_bps = UND(M, mrank, wrank, steps % 2)
#print(undominated_bps)
neis = []
for un in undominated_bps:
copyM = M[:]
copyM = RemoveBP(copyM, un, n1)
neis.append(copyM)
random_walk = True if random.random() <= 0.2 else False
if random_walk:
M = neis[random.randint(0,
len(neis) -
1)] #Choose a random neighbor
else:
#Choose a neighbor with the minimum evaluation value
f_evals = [F(nei, mpref, wpref, mrank, wrank) for nei in neis]
min_eval = min(f_evals)
indices = [i for i, v in enumerate(f_evals) if v == min_eval]
cost_evals = [se_cost(nei, mrank, wrank)
for nei in neis] if algo == "se" else [
eg_cost(nei, mrank, wrank) for nei in neis
]
min_cost = min(cost_evals)
compare_cost2 = se_cost(M, mrank,
wrank) if algo == "se" else eg_cost(
M, mrank, wrank)
indices2 = [
i for i, v in enumerate(cost_evals) if v == min_cost
]
if min_eval < f: #if there is an improvement
if min_cost <= compare_cost2:
indices3 = []
for i in indices:
for j in indices2:
if i == j:
indices3.append(i)
if len(indices3) != 0:
M = neis[random.choice(indices3)]
else:
M = neis[random.choice(indices)]
else:
#M = neis[random.randint(0, len(neis)-1)]
M = neis[random.choice(indices2)]
else:
M = neis[random.choice(indices2)]
#M = neis[random.randint(0, len(neis)-1)]
steps -= 1
nbp_val, ns_val = NBP_NS(Mbest, mpref, wpref, mrank, wrank)
f_val = nbp_val + ns_val
cost_val = se_cost(M, mrank, wrank) if algo == "se" else eg_cost(
M, mrank, wrank)
return f_val, cost_val, nbp_val, ns_val, fs_return, costs_return, begin - steps
def LTIU_always_min_neighbor(n1, n2, mpref, wpref, steps):
mrank, wrank = changeStructure(mpref, wpref, n1, n2)
arraysize = (n1 + n2)
M = [-1] * arraysize
M = random_match(M, mrank, n1, n2)
Fprev = F(M, mpref, wpref, mrank, wrank)
Fbest = Fprev
stableFound = False
NSbest = 2 * n1 * n2
Mbest = M
begin = steps
y = []
while steps:
nbp, ns = NBP_NS(M, mpref, wpref, mrank, wrank)
#nbp = NBP(M, mrank, wrank)
#ns = NS(M)
f = nbp + ns
y.append(f)
if stableFound == False and f < Fbest:
Mbest = M
#If a perfect matching found
if f == 0:
Mbest = M
Fbest = 0 #f
break
#If a stable (but not perfect) matching found
if nbp == 0:
stableFound = True
#Is it the best Stable matching among stable matchings found so far?
if ns < NSbest:
Mbest = M
NSbest = ns
Fbest = f # = ns
M = random_match(M, mrank, n1, n2)
#If unstable
else:
if stableFound == False and f < Fbest:
Fbest = f
Mbest = M
#Find neighbors
undominated_bps = UND(M, mrank, wrank, steps % 2)
#undominated_bps = UND_BP_MFWS(M, mpref, wpref, mrank, wrank)
#undominated_bps = UND_BP_WFMS(M, mpref, wpref, mrank, wrank)
#print(undominated_bps)
neis = []
for un in undominated_bps:
copyM = M[:]
copyM = RemoveBP(copyM, un, n1)
neis.append(copyM)
random_walk = 1 if random.random() <= 0.2 else 0
if random_walk:
M = neis[random.randint(0,
len(neis) -
1)] #Choose a random neighbor
else:
#Choose a neighbor with the minimum evaluation value
evaluations = [
F(nei, mpref, wpref, mrank, wrank) for nei in neis
]
min_eval = min(evaluations)
if min_eval < f: #if there is an improvement
indices = [
i for i, v in enumerate(evaluations) if v == min_eval
]
M = neis[random.choice(indices)]
else: #no improvement, random restart
#M = random_match(M, mrank, n1, n2)
indices = [
i for i, v in enumerate(evaluations) if v == min_eval
]
M = neis[random.choice(indices)]
steps -= 1
#print(Mbest)
nbp, ns = NBP_NS(Mbest, mpref, wpref, mrank, wrank)
f = ns + nbp
return f, nbp, ns, begin - steps, y
