def __init__(self, p=None):
"""
构建模型参数，加载数据
把前90%分为8:1用作train和valid，来选择超参数, 不用去管剩下的10%.
把前90%作为train，剩下的是test，把valid时学到的参数拿过来跑程序.
valid和test部分，程序是一样的，区别在于送入的数据而已。
:param p: 一个标示符，没啥用
:return:
"""
# 1. 建立各参数。要调整的地方都在 p 这了，其它函数都给写死。
if not p:
t = 't' # 写1就是valid, 写0就是test
assert 't' == t or 'v' == t # no other case
p = OrderedDict([
('dataset', 'user_buys.txt'),
('mode', 'test' if 't' == t else 'valid'),
('split',
[0.8, 1.0] if 't' == t else [0.6, 0.8]), # no third case
('at_nums', [5, 10, 15, 20]),
('epochs', 100),
('latent_size', 20),
('alpha', 0.01),
('lambda', 0.001), # 实测表明，不需要为权重单独建立lambda。
('window_x', 5), # 2,3,4,5
('window_h', 5),
('zoneout_cell', 0.5),
('zoneout_hidd', 0.05),
('mini_batch', 0), # 0:one_by_one【用这个】, 1:mini_batch
('hcagru', 4), # 0:gru, 1:zoneout
# 2:hca_x, 3:hca_h, 4:hca
('batch_size_train', 4), # 4比较合适。
('batch_size_test', 768), # user * item 矩阵太大了，分成多次计算。 768
])
for i in p.items():
print(i)
# 2. 加载数据
# 因为train/set里每项的长度不等，无法转换为完全的(n, m)矩阵样式，所以shared会报错.
global PATH
[(user_num, item_num), (tra_buys, tes_buys)] = \
load_data(os.path.join(PATH, p['dataset']), p['mode'], p['split'])
# 正样本加masks
tra_buys_masks, tra_masks = fun_data_buys_masks(tra_buys,
tail=[item_num
]) # 预测时算用户表达用
tes_buys_masks, tes_masks = fun_data_buys_masks(tes_buys,
tail=[item_num
]) # 预测时用
# 负样本加masks
tra_buys_neg_masks = fun_random_neg_masks_tra(
item_num, tra_buys_masks) # 训练时用（逐条、mini-batch均可）
tes_buys_neg_masks = fun_random_neg_masks_tes(item_num, tra_buys_masks,
tes_buys_masks) # 预测时用
# 3. 创建类变量
self.p = p
self.user_num, self.item_num = user_num, item_num
self.tra_buys_masks, self.tra_masks, self.tra_buys_neg_masks = tra_buys_masks, tra_masks, tra_buys_neg_masks
self.tes_buys_masks, self.tes_masks, self.tes_buys_neg_masks = tes_buys_masks, tes_masks, tes_buys_neg_masks
def __init__(self, p=None):
"""
构建模型参数，加载数据
把前90%分为8:1用作train和valid，来选择超参数, 不用去管剩下的10%.
把前90%作为train，剩下的是test，把valid时学到的参数拿过来跑程序.
valid和test部分，程序是一样的，区别在于送入的数据而已。
:param p: 一个标示符，没啥用
:return:
"""
# 1. 建立各参数。要调整的地方都在 p 这了，其它函数都给写死。
if not p:
t = 't' # 写1就是valid, 写0就是test
assert 't' == t or 'v' == t # no other case
p = OrderedDict(
[
('dataset', 'user_buys.txt'),
('mode', 'test' if 't' == t else 'valid'),
('split', [0.8, 1.0] if 't' == t else [0.6, 0.8]), # no third case
('at_nums', [5, 10, 15, 20]),
('epochs', 100),
('latent_size', 20),
('alpha', 0.01),
('lambda', 0.001),
('set_len', 2), # 每时刻建模几个item
('layer', 5), # ResNet的层数
('mini_batch', 1), # 0:one_by_one, 1:mini_batch
('marank', 0), # 0:MARank
('batch_size_train', 4), # 先试试16
('batch_size_test', 768), # user * item 矩阵太大了，分成多次计算。 768
])
for i in p.items():
print(i)
# 2. 加载数据
# 因为train/set里每项的长度不等，无法转换为完全的(n, m)矩阵样式，所以shared会报错.
global PATH
[(user_num, item_num), (tra_buys, tes_buys)] = \
load_data(os.path.join(PATH, p['dataset']), p['mode'], p['split'])
# 正样本加masks
tra_buys_masks, tra_masks = fun_data_buys_masks(tra_buys, tail=[item_num]) # 预测时算用户表达用
tes_buys_masks, tes_masks = fun_data_buys_masks(tes_buys, tail=[item_num]) # 预测时用
# 负样本加masks
tra_buys_neg_masks = fun_random_neg_masks_tra(item_num, tra_buys_masks) # 训练时用（逐条、mini-batch均可）
tes_buys_neg_masks = fun_random_neg_masks_tes(item_num, tra_buys_masks, tes_buys_masks) # 预测时用
# 训练序列给做成集合的形式，每个usr在每个时刻都是处理多个items。做成mask形式，predict要用。
tra_set_masks = fun_tra_set(tra_buys_masks, tail=[item_num], set_len=p['set_len'])
# 3. 创建类变量
self.p = p
self.user_num, self.item_num = user_num, item_num
self.tra_buys_masks, self.tra_masks, self.tra_buys_neg_masks = tra_buys_masks, tra_masks, tra_buys_neg_masks
self.tes_buys_masks, self.tes_masks, self.tes_buys_neg_masks = tes_buys_masks, tes_masks, tes_buys_neg_masks
self.tra_set_masks = tra_set_masks
def __init__(self, p=None):
if not p:
t = 't'
assert 't' == t or 'v' == t or 's' == t
p = OrderedDict(
[
('dataset', 'Foursquare.txt'),
# ('dataset', 'Gowalla.txt'),
('mode', 'test' if 't' == t else 'valid' if 'v' == t else 's'),
('load_epoch', 0),
('save_per_epoch', 100),
('split', -2 if 'v' == t else -1),
('at_nums', [5, 10, 15, 20]),
('epochs', 101),
('latent_size', 20),
('alpha', 0.01),
('lambda', 0.001),
('loss_weight', [0.5, 0.5]),
('dd', 200),
('UD', 40),
('mini_batch', 0),
('gru', 0),
('batch_size_train', 1),
('batch_size_test', 32),
])
for i in p.items():
print(i)
dist_num = int(p['UD'] * 1000 / p['dd']) # 1520 = 38*1000/25。idx=[0, 1519+1]
[(user_num, item_num), pois_cordis, (tra_buys, tes_buys), (tra_dist, tes_dist)] = \
load_data(os.path.join(PATH, p['dataset']), p['mode'], p['split'], p['dd'], dist_num)
tra_buys_masks, tra_dist_masks, tra_masks = fun_data_buys_masks(tra_buys, tra_dist, [item_num], [dist_num])
tes_buys_masks, tes_dist_masks, tes_masks = fun_data_buys_masks(tes_buys, tes_dist, [item_num], [dist_num])
tra_buys_neg_masks = fun_random_neg_masks_tra(item_num, tra_buys_masks)
tes_buys_neg_masks = fun_random_neg_masks_tes(item_num, tra_buys_masks, tes_buys_masks)
tra_dist_neg_masks = fun_compute_dist_neg(tra_buys_masks, tra_masks, tra_buys_neg_masks, pois_cordis, p['dd'], dist_num)
usrs_last_poi_to_all_intervals = fun_compute_distance(tra_buys_masks, tra_masks, pois_cordis, p['dd'], dist_num)
self.p = p
self.user_num, self.item_num, self.dist_num = user_num, item_num, dist_num
self.pois_cordis = pois_cordis
self.tra_buys_masks, self.tra_masks, self.tra_buys_neg_masks = tra_buys_masks, tra_masks, tra_buys_neg_masks
self.tes_buys_masks, self.tes_masks, self.tes_buys_neg_masks = tes_buys_masks, tes_masks, tes_buys_neg_masks
self.tra_dist_masks = tra_dist_masks
self.tes_dist_masks = tes_dist_masks
self.tra_dist_neg_masks = tra_dist_neg_masks
self.ulptai = usrs_last_poi_to_all_intervals
def __init__(self, p=None):
"""
构建模型参数，加载数据
把前90%分为8:1用作train和valid，来选择超参数, 不用去管剩下的10%.
把前90%作为train，剩下的是test，把valid时学到的参数拿过来跑程序.
valid和test部分，程序是一样的，区别在于送入的数据而已。
:param p: 一个标示符，没啥用
:return:
"""
# 1. 建立各参数。要调整的地方都在 p 这了，其它函数都给写死。
if not p:
t = 't' # 写1就是valid, 写0就是test
assert 't' == t or 'v' == t or 's' == t # no other case
p = OrderedDict([
('dataset', 'Foursquare.txt'),
# ('dataset', 'Gowalla.txt'),
('mode', 'test' if 't' == t else 'valid' if 'v' == t else 's'),
('load_epoch', 0),
('save_per_epoch', 100),
('split', -2 if 'v' == t else -1), # test预测最后一个。
('at_nums', [5, 10, 15, 20]),
('epochs', 101),
('latent_size', 20),
('alpha', 0.01),
('lambda', 0.001),
('loss_weight', [0.5, 0.5]),
('dd', 200), # 25m
('UD', 40), # 截断距离38km，lambda_s的维度。
('mini_batch', 0), # 0:one_by_one, 1:mini_batch. 全都用逐条。
('gru', 3), # 0:bpr, 1:gru, 2:spatial-gru, 3: ca-rnn
('batch_size_train', 1), #
('batch_size_test', 32), # user * item 矩阵太大了，分成多次计算。 768
])
for i in p.items():
print(i)
dist_num = int(p['UD'] * 1000 /
p['dd']) # 1520 = 38*1000/25。idx=[0, 1519+1]
# 2. 加载数据
# 因为train/set里每项的长度不等，无法转换为完全的(n, m)矩阵样式，所以shared会报错.
[(user_num, item_num), pois_cordis, (tra_buys, tes_buys), (tra_dist, tes_dist)] = \
load_data(os.path.join(PATH, p['dataset']), p['mode'], p['split'], p['dd'], dist_num)
# 正样本加masks
tra_buys_masks, tra_masks = fun_data_buys_masks(tra_buys,
tail=[item_num
]) # 预测时算用户表达用
tes_buys_masks, tes_masks = fun_data_buys_masks(tes_buys,
tail=[item_num
]) # 预测时用
tra_dist_masks, _ = fun_data_buys_masks(tra_dist, tail=[dist_num])
tes_dist_masks, _ = fun_data_buys_masks(tes_dist, tail=[dist_num])
# 负样本加masks
tra_buys_neg_masks = fun_random_neg_masks_tra(
item_num, tra_buys_masks) # 训练时用（逐条、mini-batch均可）
tes_buys_neg_masks = fun_random_neg_masks_tes(item_num, tra_buys_masks,
tes_buys_masks) # 预测时用
# 计算负样本与上一个正样本的距离间隔，并加masks
tra_dist_neg_masks = fun_compute_dist_neg(tra_buys_masks, tra_masks,
tra_buys_neg_masks,
pois_cordis, p['dd'],
dist_num)
# 每个user最后一个poi和all pois的距离落在哪个区间里。
usrs_last_poi_to_all_intervals = fun_compute_distance(
tra_buys_masks, tra_masks, pois_cordis, p['dd'], dist_num)
# print(tra_dist[0][:5]) # [1520, 274, 0, 428, 142], 38km/25m=1520
# 3. 创建类变量
self.p = p
self.user_num, self.item_num, self.dist_num = user_num, item_num, dist_num
self.pois_cordis = pois_cordis
self.tra_buys_masks, self.tra_masks, self.tra_buys_neg_masks = tra_buys_masks, tra_masks, tra_buys_neg_masks
self.tes_buys_masks, self.tes_masks, self.tes_buys_neg_masks = tes_buys_masks, tes_masks, tes_buys_neg_masks
self.tra_dist_masks = tra_dist_masks
self.tes_dist_masks = tes_dist_masks
self.tra_dist_neg_masks = tra_dist_neg_masks
self.ulptai = usrs_last_poi_to_all_intervals