def __init__(self,
data,
train_ratio=0.99,
ratio=0.1,
pool_size=10,
hash_count=8):
super().__init__(data, train_ratio, ratio)
self.lsh_family = []
self.hash_count = hash_count
for i in range(self.hash_count):
self.lsh_family.append(LSH(pool_size))
self.lsh_family[i].fit(self.train_data.shape[1])
def __init__(self,
data,
train_ratio=0.95,
ratio=0.1,
pool_size=10,
hash_count=8):
super().__init__(data, train_ratio, ratio)
self.num_of_services = self.data.shape[1]
self.lsh_family = []
self.hash_count = hash_count
for i in range(self.hash_count):
self.lsh_family.append(LSH(pool_size))
##这里拟合的是和UCFRecommender的标准不一样
self.lsh_family[i].fit(self.num_of_train)
def __init__(self,
data,
ratio=0.1,
num_hash_functions=10,
num_hash_tables=8):
self.raw_data = data
self.num_hash_functions = num_hash_functions
self.num_hash_tables = num_hash_tables
#random erased
self.data = self.randomErased()
#initialize lsh_family
self.lsh_family = []
for i in range(num_hash_tables):
self.lsh_family.append(LSH(num_hash_functions))
#initialize the parameters of hash functions
self.lsh_family[i].fit(self.data.shape[1])
def __init__(self, data, num_of_functions=4, num_of_tables=8, seed=1):
'''
:param data: shape (num_of_users, num_of_services, num_of_time_slices)
:param ratio: erase ratio of data[:, :, 63], the erased element will be set to 0
:param parameters: parameters of local sensitive hash function, in consistent with num_of_hash_tables
'''
self.data = data
(self.num_of_users, self.num_of_services) = data.shape
#initialize lsh tables
self.num_of_tables = num_of_tables
self.lsh_family = []
for i in range(self.num_of_tables):
self.lsh_family.append(LSH(num_of_functions))
self.lsh_family[i].fit(self.num_of_users, seed)
seed += 1
def __init__(self, data, num_of_functions=4, num_of_tables=8, seed=1):
'''
:param data: shape (num_of_users, num_of_services, num_of_time_slices)
:param num_of_functions:
:param num_of_tables:
:param seed: 随机函数的种子
'''
self.data = data
(self.num_of_users, self.num_of_services) = data.shape
#initialize lsh tables
self.num_of_tables = num_of_tables
self.lsh_family = []
for i in range(self.num_of_tables):
self.lsh_family.append(LSH(num_of_functions))
self.lsh_family[i].fit(self.num_of_users, seed)
seed += 1
def __init__(self, data, num_of_functions = 4, num_of_tables = 8, seed = 1, threshold = 0):
'''
:param data: 测试数据
:param num_of_functions:
:param num_of_tables:
:param seed:
:param threshold: 找相似用户的阈值，默认为0， 与传统的LSH方法一致
'''
self.data = data
(self.num_of_users, self.num_of_services) = data.shape
#initialize lsh tables
self.num_of_tables = num_of_tables
self.lsh_family = []
for i in range(self.num_of_tables):
self.lsh_family.append(LSH(num_of_functions))
self.lsh_family[i].fit(self.num_of_users, seed)
seed += 1
def __init__(self, data, ratio = 0.8,
pool_size = 10, hash_count = 8):
'''
:param data: 从文件中读取的评分数据，矩阵形式，行数为用户数，列数为项目数
:param user：测试用户的索引
:param ratio: 为了让测试预测的效果，设定保留的数据的比例，剩余的数据抹零
:param pool_size: 每个LSH返回的hash值的位数（二进制）
:param hash_count: LSH的个数
'''
# self.num_of_users = self.data.shape[0]
self.data = data
self.processed_data = np.copy(self.data)
self.sample(ratio)
self.user = np.random.randint(0, self.data.shape[0])
self.remove_similar_users()
self.lsh_family = []
self.hash_count = hash_count
for i in range(self.hash_count):
self.lsh_family.append(LSH(pool_size))
#这里是查找用户的相似度
self.lsh_family[i].fit(self.data.shape[1])