if __name__ == '__main__':
import Recommender_System.utility.gpu_memory_growth
from tensorflow.keras.optimizers import Adam
from Recommender_System.data import kg_loader, data_process
from Recommender_System.algorithm.MKR.model import MKR_model
from Recommender_System.algorithm.MKR.train import train
n_user, n_item, n_entity, n_relation, train_data, test_data, kg, topk_data = data_process.pack_kg(
kg_loader.ml1m_kg_MKR)
model_rs, model_kge = MKR_model(n_user,
n_item,
n_entity,
n_relation,
dim=8,
L=1,
H=1,
l2=1e-6)
train(model_rs,
model_kge,
train_data,
test_data,
kg,
topk_data,
kge_interval=3,
optimizer_rs=Adam(0.02),
optimizer_kge=Adam(0.01),
epochs=20,
batch=4096)
'''
n_user, n_item, n_entity, n_relation, train_data, test_data, kg, topk_data = data_process.pack_kg(kg_loader.lastfm_kg_MKR)
model_rs, model_kge = MKR_model(n_user, n_item, n_entity, n_relation, dim=4, L=2, H=1, l2=1e-6)
if __name__ == '__main__':
import Recommender_System.utility.gpu_memory_growth
from Recommender_System.algorithm.KGCN.tool import construct_undirected_kg, get_adj_list
from Recommender_System.algorithm.KGCN.model import KGCN_model
from Recommender_System.algorithm.KGCN.train import train
from Recommender_System.data import kg_loader, data_process
import tensorflow as tf
n_user, n_item, n_entity, n_relation, train_data, test_data, kg, topk_data = data_process.pack_kg(
kg_loader.ml1m_kg1m, negative_sample_threshold=4)
neighbor_size = 16
adj_entity, adj_relation = get_adj_list(construct_undirected_kg(kg),
n_entity, neighbor_size)
model = KGCN_model(n_user,
n_entity,
n_relation,
adj_entity,
adj_relation,
neighbor_size,
iter_size=1,
dim=16,
l2=1e-7,
aggregator='sum')
train(model,
train_data,
test_data,
topk_data,
optimizer=tf.keras.optimizers.Adam(0.01),
from Recommender_System.algorithm.RippleNet.train import train
from Recommender_System.data import kg_loader, data_process
import tensorflow as tf
'''
n_user, n_item, n_entity, n_relation, train_data, test_data, kg, topk_data = data_process.pack_kg(kg_loader.ml20_kg_KGCN, split_ensure_positive=True)
hop_size, ripple_size = 2, 32
ripple_set = get_ripple_set(hop_size, ripple_size, get_user_positive_item_list(train_data), construct_directed_kg(kg))
model = RippleNet_model(n_entity, n_relation, dim=32, hop_size=hop_size, ripple_size=ripple_size, kge_weight=0.01, l2=1e-7)
train(model, train_data, test_data, topk_data, ripple_set, optimizer=tf.keras.optimizers.Adam(0.01), epochs=10, batch=65536)
'''
n_user, n_item, n_entity, n_relation, train_data, test_data, kg, topk_data = data_process.pack_kg(
kg_loader.ml1m_kg_RippleNet, split_ensure_positive=True)
hop_size, ripple_size = 2, 32
ripple_set = get_ripple_set(hop_size, ripple_size,
get_user_positive_item_list(train_data),
construct_directed_kg(kg))
model = RippleNet_model(n_entity,
n_relation,
dim=16,
hop_size=hop_size,
ripple_size=ripple_size,
kge_weight=0,
l2=1e-7)