class SetOptimizer(object):
"""docstring for SetOptimizer"""
minimizer_kwargs = {"method": "BFGS"}
stepsize = 20
def __init__(self, model_name):
super(SetOptimizer, self).__init__()
self.model = ItemSetModel()
self.model.load(model_name)
self.item_ids = [x.item_id for x in Item.objects().all()]
self.item_ids.sort()
features = 25
start_ind = 13
input_array = np.zeros(25)
def objective_function(self, x):
self.input_array[self.start_ind:] = x
return -self.model.predict(self.input_array)[0]
class TakeStep(object):
def __init__(self, items, stepsize=40):
self.stepsize = stepsize
self.item_ids = items
def __call__(self, x):
s = self.stepsize
x_new = x.copy()
for i in range(len(x)):
curr = self.item_ids.index(x[i])
start = max(0, round(curr - s / 2))
end = min(len(self.item_ids) - 1, round(curr + s / 2))
x_new[i] = self.item_ids[random.randint(start, end)]
return x_new
def optimize(self, cid, other_champion_stats):
champion = Champion.objects(champion_id=cid).get()
self.input_array[0] = cid
self.input_array[1:7] = np.array(champion.class_data)
self.input_array[7:self.start_ind] = np.array(other_champion_stats)
x0 = [random.choice(self.item_ids) for x in range(12)]
ret = basinhopping(self.objective_function,
x0,
take_step=self.TakeStep(self.item_ids,
self.stepsize),
minimizer_kwargs=self.minimizer_kwargs,
niter=100,
stepsize=100,
T=10.0)
return ret
class SetOptimizer(object):
"""docstring for SetOptimizer"""
minimizer_kwargs = {"method": "BFGS"}
stepsize = 20
def __init__(self, model_name):
super(SetOptimizer, self).__init__()
self.model = ItemSetModel()
self.model.load(model_name)
self.item_ids = [x.item_id for x in Item.objects().all()]
self.item_ids.sort()
features = 25
start_ind = 13
input_array = np.zeros(25)
def objective_function(self, x):
self.input_array[self.start_ind:] = x
return -self.model.predict(self.input_array)[0]
class TakeStep(object):
def __init__(self, items, stepsize=40):
self.stepsize = stepsize
self.item_ids = items
def __call__(self, x):
s = self.stepsize
x_new = x.copy()
for i in range(len(x)):
curr = self.item_ids.index(x[i])
start = max(0, round(curr - s/2))
end = min(len(self.item_ids)-1, round(curr + s/2))
x_new[i] = self.item_ids[random.randint(start, end)]
return x_new
def optimize(self, cid, other_champion_stats):
champion = Champion.objects(champion_id=cid).get()
self.input_array[0] = cid
self.input_array[1:7] = np.array(champion.class_data)
self.input_array[7:self.start_ind] = np.array(other_champion_stats)
x0 = [random.choice(self.item_ids) for x in range(12)]
ret = basinhopping(
self.objective_function,
x0,
take_step=self.TakeStep(self.item_ids, self.stepsize),
minimizer_kwargs=self.minimizer_kwargs,
niter=100,
stepsize=100,
T=10.0
)
return ret
def _generate_model(self, filename, num_rows, cached=False):
print("Generating model...")
model = ItemSetModel()
if(not cached):
print("Assembling data...")
X, Y = model.get_data_sets(num_rows,cache=True)
else:
print("Loading cached data...")
X, Y = model.get_cached_data(num_rows)
model.train(X, Y, train_ratio=0.8)
print("Saving model to file...")
model.save(filename)
def __init__(self, model_name):
super(SetOptimizer, self).__init__()
self.model = ItemSetModel()
self.model.load(model_name)
self.item_ids = [x.item_id for x in Item.objects().all()]
self.item_ids.sort()
def __init__(self, model_name):
super(SetOptimizer, self).__init__()
self.model = ItemSetModel()
self.model.load(model_name)
self.item_ids = [x.item_id for x in Item.objects().all()]
self.item_ids.sort()