def non_convexity(params):
t = 1
x = 0
confs = list(ModelConfig(
gamma=1,
lead_time=0,
info_state_rvs=None,
holding_cost=params["h"],
backlogging_cost=params["b"],
setup_cost=0,
unit_price=0,
usage_model=BinomUsageModel(n=params["usage_n"], p=params["usage_p"]),
increments=1,
horizon=n,
info_rv=rv_0_1,
label="base_case_non_convex_bh_100",
label_index=i
) for n in [0, 1, 2])
model_0 = get_model(confs[0])
model_1 = get_model(confs[1])
model_2 = get_model(confs[2])
j0 = mean([model_0.j_function(t, x, info_state) for info_state in model_0.info_states()])
j1 = mean([model_1.j_function(t, x, info_state) for info_state in model_1.info_states()])
j2 = mean([model_2.j_function(t, x, info_state) for info_state in model_2.info_states()])
metric = 1 if j0 == j1 and j1 == j2 else (j1 - j2) / (j0 - j1)
return metric
i = 0
usage_models = [PoissonUsageModel(scale=1)]
for rv in rvs:
for horizon in [0, 1, 2, 3, 4]:
for u in usage_models:
for b in [0.001, 1, 10, 100, 1000, 10000]:
for h in [1]:
configs.append(ModelConfig(
gamma=1,
lead_time=0,
info_state_rvs=None,
holding_cost=h,
backlogging_cost=b,
setup_cost=0,
unit_price=0,
usage_model=u,
increments=1,
horizon=horizon,
info_rv=rv,
label="k0_geometric_bookings",
label_index=i)
)
i += 1
if __name__ == "__main__":
xs = list(range(0, 30))
ts = list(range(0, 21))
run_configs(configs, ts, xs, pools=8)
configs = []
i = 0
for info_rv in [rv_7_13, rv_6_14, rv_5_15]:
for horizon in [0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10]:
for k in [40, 50, 60, 70, 80]:
for b in [10]:
i += 1
configs.append(ModelConfig(
gamma=0.9,
lead_time=0,
info_state_rvs=None,
holding_cost=1,
backlogging_cost=b,
setup_cost=k,
unit_price=0,
usage_model=DeterministUsageModel(scale=1),
increments=1,
horizon=horizon,
info_rv=info_rv,
label="Non-Convex_Search_Det_Usage",
label_index=i)
)
if __name__ == "__main__":
xs = list(range(0, 1))
ts = list(range(0, 20))
run_configs(configs, ts, xs, pools=8)
from scm_optimization.model import ModelConfig, run_configs, PoissonUsageModel, BinomUsageModel, DeterministUsageModel
import pacal
from decimal import *
configs = []
i = 0
for horizon in [0, 1, 2, 3, 4]:
for b in [10, 1000]:
for d in [1, 5, 10, 20]:
configs.append(
ModelConfig(gamma=1,
lead_time=0,
info_state_rvs=None,
holding_cost=1,
backlogging_cost=b,
setup_cost=0,
unit_price=0,
usage_model=PoissonUsageModel(1),
horizon=horizon,
info_rv=pacal.BinomialDistr(d, 0.5),
label="demand_scale_experiment",
label_index=i))
i += 1
configs = [configs[35], configs[39]]
if __name__ == "__main__":
xs = list(range(0, 1))
ts = list(range(0, 21))
run_configs(configs, ts, xs, pools=8)
configs = []
i = 0
for demand_n in [1, 2, 3, 4]:
for usage_n in [1, 2, 3, 4]:
for p in [0.25, 0.5, 0.75]:
for q in [0.25, 0.5, 0.75]:
for horizon in [0, 1, 2, 3, 4]:
for b in [0.001, 0.05, 0.1, 0.5, 1, 2, 4, 100]:
configs.append(
ModelConfig(
gamma=1,
lead_time=0,
info_state_rvs=None,
holding_cost=1,
backlogging_cost=b,
setup_cost=0,
unit_price=0,
usage_model=BinomUsageModel(n=usage_n, p=p),
increments=1,
horizon=horizon,
info_rv=pacal.BinomialDistr(demand_n, q),
label="numerical_experiments_binomial_model",
label_index=i))
i += 1
if __name__ == "__main__":
xs = list(range(0, 20))
ts = list(range(0, 21))
run_configs(configs, ts, xs, pools=8)
configs = []
i = 0
usage_models = list(BinomUsageModel(n=1, p=p) for p in [0.6])
for rv in rvs:
for horizon in [0, 1, 2, 3, 4, 5, 6]:
for u in usage_models:
for b in [100]:
for h in [100]:
configs.append(
ModelConfig(gamma=1,
lead_time=0,
info_state_rvs=None,
holding_cost=h,
backlogging_cost=b,
setup_cost=0,
unit_price=0,
usage_model=u,
increments=1,
horizon=horizon,
info_rv=rv,
label="base_case_non_convex_bh_100",
label_index=i))
i += 1
if __name__ == "__main__":
xs = list(range(0, 3))
ts = list(range(0, 20))
run_configs(configs, ts, xs, pools=8)
BinomUsageModel(n=5, p=0.2),
BinomUsageModel(n=10, p=0.1),
]
b = 1000
poisson_usage_model = PoissonUsageModel(1, trunk=1e-10)
booking_model = pacal.BinomialDistr(10, 0.5)
for horizon in [0, 1, 2, 3]:
configs.append(ModelConfig(
gamma=1,
lead_time=0,
info_state_rvs=None,
holding_cost=1,
backlogging_cost=b,
setup_cost=0,
unit_price=0,
usage_model=poisson_usage_model,
horizon=horizon,
info_rv=booking_model,
label="poisson_usage_policy",
label_index=i)
)
i += 1
binomial_usage_configs = []
i = 0
for horizon in [0, 1, 2, 3]:
for usage_model in binomial_usage_models:
binomial_usage_configs.append(ModelConfig(
gamma=1,
PoissonUsageModel(scale=1),
BinomUsageModel(n=4, p=0.25),
BinomUsageModel(n=2, p=0.5),
DeterministUsageModel(1)
]
for rv in rvs:
for horizon in [0, 1, 2, 3, 4, 5, 6]:
for u in usage_models:
for b in [100, 50, 10, 5, 1]:
configs.append(
ModelConfig(gamma=0.9,
lead_time=0,
info_state_rvs=None,
holding_cost=1,
backlogging_cost=b,
setup_cost=0,
unit_price=0,
usage_model=u,
increments=1,
horizon=horizon,
info_rv=rv,
label="k0_searches",
label_index=i))
i += 1
if __name__ == "__main__":
xs = list(range(0, 30))
ts = list(range(0, 21))
run_configs(configs, ts, xs, pools=8)
rv_4_16 = pacal.DiscreteDistr([4, 16], [0.5, 0.5])
rv_2_18 = pacal.DiscreteDistr([2, 18], [0.5, 0.5])
rv_0_20 = pacal.DiscreteDistr([0, 20], [0.5, 0.5])
rv_8_16 = pacal.DiscreteDistr([8, 16], [0.5, 0.5])
configs = []
i = 0
def poisson_usage(o):
return pacal.PoissonDistr(o, trunk_eps=1e-3)
for horizon in [0, 1, 2, 3, 4]:
for inc in [1, Decimal("0.5"), Decimal("0.2"), Decimal("0.1")]:
configs.append(
ModelConfig(gamma=0.9,
lead_time=0,
info_state_rvs=None,
holding_cost=1,
backlogging_cost=10,
setup_cost=50,
unit_price=0,
usage_model=poisson_usage,
increments=inc,
horizon=horizon,
info_rv=rv_6_14,
label="Action_Increment",
label_index=i))
i += 1
rv_8_16 = pacal.DiscreteDistr([8, 16], [0.5, 0.5])
configs = []
i = 0
for horizon in [0, 1, 2, 3, 4, 5, 6, 7, 8]:
for l in [0, 1, 2, 3]:
for b in [10]:
i += 1
configs.append(ModelConfig(
gamma=0.9,
lead_time=l,
info_state_rvs=None,
holding_cost=1,
backlogging_cost=b,
setup_cost=50,
unit_price=0,
usage_model=DeterministUsageModel(scale=1),
increments=1,
horizon=horizon,
info_rv=rv_6_14,
label="Leadtime",
label_index=i)
)
if __name__ == "__main__":
xs = list(range(0, 1))
ts = list(range(0, 16))
run_configs(configs, ts, xs, pools=8)
usage_models = [
PoissonUsageModel(scale=1),
BinomUsageModel(n=4, p=0.25),
BinomUsageModel(n=2, p=0.5),
DeterministUsageModel(1)
]
for horizon in [0, 1, 2, 3, 4]:
for k in [50, 55, 60]:
for u in usage_models:
configs.append(
ModelConfig(gamma=0.9,
lead_time=0,
info_state_rvs=None,
holding_cost=1,
backlogging_cost=10,
setup_cost=k,
unit_price=0,
usage_model=u,
increments=1,
horizon=horizon,
info_rv=rv_6_14,
label="Non-Convex_Search_Usage_Model",
label_index=i))
i += 1
if __name__ == "__main__":
xs = list(range(0, 30))
ts = list(range(0, 16))
run_configs(configs, ts, xs, pools=8)