class Trainer(object):
def __init__(self, dataset):
"""
dataset: object of DataInputer
"""
self.dataset = dataset
self.model = LambdaRank()
def __call__(self):
self.run()
def run(self):
"""
record is a string line
"""
val_maps = []
for val_idx in xrange(TRAIN_SET_NUM):
# user ith dataset as a validate dataset
self.val_idx = val_idx
set_indexs = set(range(TRAIN_SET_NUM))
set_indexs.discard(val_idx)
self.train(set_indexs)
val_res = self.validate()
show_status(".. get map: " + str(val_res))
val_maps.append(val_res)
map_res = sum(val_maps) / TRAIN_SET_NUM
show_status(".. get avage map: " + str(map_res))
self.model.dataspace.tofile(data_path('models', str(map_res)))
def train(self, set_indexs):
# train using the rest dataset
for i in list(set_indexs):
show_status(">>>" * 20)
show_status(".. training %dth dataset" % i)
for dataset in self.dataset.get_dataset(i):
dataset_len = len(self.dataset.train_pairs[i])
show_status("dataset len: %d" % dataset_len)
for i, (X1, X2) in enumerate(dataset):
print "train %dth line" % i
show_status(">> training data", i, dataset_len)
X1 = np.array([float(i) for i in X1.split()])
X2 = np.array([float(i) for i in X2.split()])
self.model.study_line(X1, X2)
def validate(self):
"""
validate and save best MAP
"""
def mysort(l1, l2):
if l1[1] == l2[1]:
return 0
if l1[1] > l2[1]:
return -1
return 1
# TODO how to validate?
vali_set = self.dataset.trainset[self.val_idx]
uid, p_papers, n_papers = vali_set.split(',')
uid = int(uid)
p_papers = [int(i) for i in p_papers]
n_papers = [int(i) for i in n_papers]
predicts = []
for p in p_papers + n_papers:
p_feature = self.dataset.get_data_line(uid, p)
score = self.model.predict(p_feature)
predicts.append((p, score))
predicts.sort(mysort)
return cal_map(p_papers, predicts)
def __init__(self, dataset):
"""
dataset: object of DataInputer
"""
self.dataset = dataset
self.model = LambdaRank()
class Trainer(object):
def __init__(self, dataset):
"""
dataset: object of DataInputer
"""
self.dataset = dataset
self.model = LambdaRank()
def __call__(self):
self.run()
def run(self):
"""
record is a string line
"""
val_maps = []
for val_idx in xrange(TRAIN_SET_NUM):
# user ith dataset as a validate dataset
self.val_idx = val_idx
set_indexs = set(range(TRAIN_SET_NUM))
set_indexs.discard(val_idx)
self.train(set_indexs)
val_res = self.validate()
show_status(".. get map: " + str(val_res))
val_maps.append(val_res)
map_res = sum(val_maps) / TRAIN_SET_NUM
show_status(".. get avage map: " + str(map_res))
self.model.dataspace.tofile(data_path('models', str(map_res)))
def train(self, set_indexs):
# train using the rest dataset
for i in list(set_indexs):
show_status(">>>" * 20)
show_status(".. training %dth dataset" % i)
for dataset in self.dataset.get_dataset(i):
dataset_len = len(self.dataset.train_pairs[i])
show_status("dataset len: %d" % dataset_len)
for i, (X1, X2) in enumerate(dataset):
print "train %dth line" % i
show_status(">> training data", i, dataset_len)
X1 = np.array([float(i) for i in X1.split()])
X2 = np.array([float(i) for i in X2.split()])
self.model.study_line(X1, X2)
def validate(self):
"""
validate and save best MAP
"""
def mysort(l1, l2):
if l1[1] == l2[1]:
return 0
if l1[1] > l2[1]:
return -1
return 1
# TODO how to validate?
vali_set = self.dataset.trainset[self.val_idx]
uid, p_papers, n_papers = vali_set.split(',')
uid = int(uid)
p_papers = [int(i) for i in p_papers]
n_papers = [int(i) for i in n_papers]
predicts = []
for p in p_papers + n_papers:
p_feature = self.dataset.get_data_line(uid, p)
score = self.model.predict(p_feature)
predicts.append((p, score))
predicts.sort(mysort)
return cal_map(p_papers, predicts)
