def exchange2(self, arg_timetable):
pena = penalty.Penalty()
excha_flag = True
for grade in range(1, GRADE + 1):
for cla in range(1, CLASS + 1):
for day in range(DAY):
flag, time = pena.judge_free(arg_timetable, grade, cla,
day)
if (flag is False):
if (excha_flag):
excha_grade = grade
excha_class = cla
old_day = day
old_time = time
excha_flag = False
while (True):
new_day = random.randint(0, DAY - 1)
new_time = random.randint(0, TIME - 2)
if (new_day == old_day and new_time == new_day):
continue
elif ((hasattr(
arg_timetable[excha_grade][excha_class][new_day][new_time],
'subname') is False) or arg_timetable[excha_grade]
[excha_class][new_day][new_time].fix_flag):
break
# 交換
arg_timetable_t = arg_timetable[excha_grade][excha_class][old_day][
old_time]
arg_timetable[excha_grade][excha_class][old_day][
old_time] = arg_timetable[excha_grade][excha_class][new_day][
new_time]
arg_timetable[excha_grade][excha_class][new_day][
new_time] = arg_timetable_t
def annealingoptimize1(self,
arg_timetable,
arg_teacher,
temp=10000,
cool=0.99):
# 初期化
cnt = 0
penal = 1000000
# 解の記録
penalty_history = []
# オブジェクト生成
obj1 = exchange.Exchange()
obj2 = penalty.Penalty()
obj3 = arrangement.Arrangement()
# 探索
while penal > MIN_PENALTY or cnt < MAX_CNT:
# 交換
old_timetable = copy.deepcopy(arg_timetable)
if (penal >= 1000000):
obj1.exchange1(arg_timetable)
elif (penal >= 10000):
obj1.exchange2(arg_timetable)
else:
obj1.exchange3(arg_timetable)
# 先生配置
old_teacher = copy.deepcopy(arg_teacher)
new_teacher = copy.deepcopy(arg_teacher)
obj3.arrange_teacher(old_timetable, old_teacher)
obj3.arrange_teacher(arg_timetable, new_teacher)
# ペナルティを計算
old_pena = obj2.calc_penalty_soft(old_timetable, old_teacher)
new_pena = obj2.calc_penalty_soft(arg_timetable, new_teacher)
# 温度から確立を定義
pro = pow(math.e, -abs(new_pena - old_pena) / temp)
# コストを比較し改善 or 確率
if (new_pena < old_pena or random.random() < pro):
penal = new_pena
teacher = copy.copy(new_teacher)
else:
arg_timetable = copy.deepcopy(old_timetable)
penal = old_pena
teacher = copy.copy(old_teacher)
# 温度を下げる
temp = temp * cool
# カウントインクリメント
cnt += 1
# 表示
if cnt % 100 == 0:
print('penalty:', penal)
print('count :', cnt)
print()
# 探索が停滞
if cnt > 300 and penal > 1000000:
break
# ペナルティ
penalty_history.append(penal)
# 表示
print('----------------------')
print('penalty: ', penal)
print('----------------------')
# 戻り値
return arg_timetable, teacher, penal, penalty_history
def greatdeluge2opt(self, arg_timetable, arg_teacher):
# 初期化 cnt_stag:停滞回数 penalty_history:解の記録
cnt_stag = 0
penalty_history = []
# オブジェクト生成
obj1 = exchange.Exchange()
obj2 = penalty.Penalty()
obj3 = arrangement.Arrangement()
# 先生配置
teacher = copy.deepcopy(arg_teacher)
obj3.arrange_teacher(arg_timetable, teacher)
# ペナルティ計算
cost_current = obj2.calc_penalty_soft(arg_timetable, teacher)
# 最小ペナルティ
min_cost = cost_current
# 初期水位
water_level = cost_current
# ペナルティ記録
penalty_history.append(cost_current)
# ループ
while min_cost > 0 and cnt_stag < MAX_STAG1:
### 2-opt ###
# コピー
timetable = copy.deepcopy(arg_timetable)
teacher = copy.deepcopy(arg_teacher)
# 交換
obj1.exchange3(timetable)
# 先生配置
obj3.arrange_teacher(timetable, teacher)
# 現在のペナルティ
cost_neigh = obj2.calc_penalty_soft(timetable, teacher)
# 改善 or 水位以下
if cost_current >= cost_neigh or water_level >= cost_neigh:
# 交換を許可
arg_timetable = copy.deepcopy(timetable)
cost_current = cost_neigh
# 改善
if cost_current < min_cost:
# ペナルティ
min_cost = cost_current
min_timetable = copy.deepcopy(arg_timetable)
cnt_stag = 0
# 水位以下
if cost_current < water_level:
# 水位を下げる
water_level -= DEC_WATER1
# 同じ
if cost_current == min_cost or cost_current == water_level:
cnt_stag += 1
else:
cnt_stag += 1
# 再上昇
if cnt_stag % FREQ_RERISE1 == 0:
water_level += RERISE1
# 表示
print('min_pena', min_cost)
print('cur_pena', cost_current)
print('water_level', water_level)
print('cnt_stag', cnt_stag)
print()
# ペナルティ記録
penalty_history.append(cost_current)
obj3.arrange_teacher(min_timetable, arg_teacher)
return arg_timetable, teacher, min_cost, penalty_history
def greatdelugels(self, arg_timetable, arg_teacher):
# 初期化 cnt_stag:停滞回数
cnt_stag = 0
penalty_history = []
# オブジェクト生成
obj1 = exchange.Exchange()
obj2 = penalty.Penalty()
obj3 = arrangement.Arrangement()
# 先生配置
teacher = copy.deepcopy(arg_teacher)
obj3.arrange_teacher(arg_timetable, teacher)
# ペナルティ計算
cost_current = obj2.calc_penalty_soft(arg_timetable, teacher)
# 最小ペナルティ
min_cost = cost_current
# 初期水位
water_level = cost_current
# ペナルティ記録
penalty_history.append(cost_current)
# ループ
while min_cost > 0 and cnt_stag < MAX_STAG2:
### LocalSearch ###
# リスト
penal = []
timetable = []
teacher = []
# 近傍探索
for x in range(5):
# コピー
timetable.append(copy.deepcopy(arg_timetable))
# 交換
obj1.exchange3(timetable[x])
# 先生配置
teacher.append(copy.deepcopy(arg_teacher))
obj3.arrange_teacher(timetable[x], teacher[x])
# 現在のペナルティ
penal.append(obj2.calc_penalty_soft(timetable[x], teacher[x]))
# 最小要素
min_element = penal.index(min(penal))
timetable = copy.deepcopy(timetable[min_element])
teacher = copy.deepcopy(teacher[min_element])
cost_neigh = min(penal)
# 改善 or 水位以下
if cost_current >= cost_neigh or water_level >= cost_neigh:
# 交換を許可
arg_timetable = copy.deepcopy(timetable)
cost_current = cost_neigh
# 改善
if cost_current < min_cost:
# ペナルティ
min_cost = cost_current
min_timetable = copy.deepcopy(arg_timetable)
cnt_stag = 0
# 水位以下
if cost_current < water_level:
# 水位を下げる
water_level -= DEC_WATER2
# 同じ
if cost_current == min_cost or cost_current == water_level:
cnt_stag += 1
else:
cnt_stag += 1
# ペナルティ
penalty_history.append(cost_current)
# 再上昇
if cnt_stag % FREQ_RERISE2 == 0:
water_level += RERISE2
# 表示
print('min_pena', min_cost)
print('cur_pena', cost_current)
print('water_level', water_level)
print('cnt_stag', cnt_stag)
print()
# ペナルティ記録
penalty_history.append(cost_current)
obj3.arrange_teacher(min_timetable, arg_teacher)
return min_timetable, arg_teacher, min_cost, penalty_history