def train_model(args, tr_sparse):
"""Instantiate WALS model and use "simple_train" to factorize the matrix.
Args:
args: a list of parameters
tr_sparse: sparse training matrix
Returns:
the row and column factors in numpy format.
"""
tf.logging.info('Train Start: {:%Y-%m-%d %H:%M:%S}'.format(
datetime.datetime.now()))
# generate model
input_tensor, row_factor, col_factor, model = wals.wals_model(
tr_sparse, args.latent_factors, args.regularization, args.unobs_weight,
args.weights, args.wt_type, args.feature_wt_exp,
args.feature_wt_factor)
# factorize matrix
session = wals.simple_train(model, input_tensor, args.num_iters)
tf.logging.info('Train Finish: {:%Y-%m-%d %H:%M:%S}'.format(
datetime.datetime.now()))
# evaluate output factor matrices
output_row = row_factor.eval(session=session)
output_col = col_factor.eval(session=session)
# close the training session
session.close()
return output_row, output_col
def train(params,train_sparse): latent_factors = params['latent_factors'] num_iters = params['num_iters'] reg = params['reg'] unobserved_weight = params['unobserved_weight'] use_weight=params['use_weight'] weight_type = params['weight_type'] feature_weight_exp = params['feature_weight_exp'] feature_weight_lin = params['feature_weight_lin'] input_tensor, row_factor, col_factor, model= wals.wals_model(train_sparse,latent_factors,reg, unobserved_weight,use_weight, weight_type, feature_weight_exp, feature_weight_lin) sess=wals.simple_train(model,input_tensor,num_iters) user_factor=row_factor.eval(session=sess) item_factor=col_factor.eval(session=sess) sess.close() return user_factor,item_factor
def process(args):
if args.format == "adjlist":
G = graph.load_adjacencylist(args.input, undirected=args.undirected)
elif args.format == "edgelist":
G = graph.load_edgelist(args.input, undirected=args.undirected)
elif args.format == "mat":
G = graph.load_matfile(args.input,
variable_name=args.matfile_variable_name,
undirected=args.undirected)
else:
raise Exception(
"Unknown file format: '%s'. Valid formats: 'adjlist', 'edgelist', 'mat'"
% args.format)
print("Number of nodes: {}".format(len(G.nodes())))
nxG = nx.Graph(G)
Gmat = nx.adjacency_matrix(nxG).tocoo()
num_iters = args.num_iters
tf.logging.info('Train Start: {:%Y-%m-%d %H:%M:%S}'.format(
datetime.datetime.now()))
# generate model
input_tensor, row_factor, col_factor, model = wals.wals_model(
Gmat, args.dim // 2, args.reg, args.unobs)
# factorize matrix
session = wals.simple_train(model, input_tensor, num_iters)
tf.logging.info('Train Finish: {:%Y-%m-%d %H:%M:%S}'.format(
datetime.datetime.now()))
# evaluate output factor matrices
output_row = row_factor.eval(session=session)
output_col = col_factor.eval(session=session)
# close the training session now that we've evaluated the output
session.close()
embedding = np.concatenate((output_row, output_col), axis=1)
print(embedding)
# save trained model to job directory
np.savetxt(args.output, embedding)
# log results
train_rmse = wals.get_rmse(output_row, output_col, Gmat)
log_info = 'train RMSE = %.2f' % train_rmse
tf.logging.info(log_info)
print(log_info)
def train_model(args, tr_sparse):
"""Instantiate WALS model and use "simple_train" to factorize the matrix.
Args:
args: training args containing hyperparams
tr_sparse: sparse training matrix
Returns:
the row and column factors in numpy format.
"""
dim = args['latent_factors']
num_iters = args['num_iters']
reg = args['regularization']
unobs = args['unobs_weight']
wt_type = args['wt_type']
feature_wt_exp = args['feature_wt_exp']
obs_wt = args['feature_wt_factor']
tf.logging.info('Train Start: {:%Y-%m-%d %H:%M:%S}'.format(datetime.datetime.now()))
# generate model
input_tensor, row_factor, col_factor, model = wals.wals_model(tr_sparse,
dim,
reg,
unobs,
args['weights'],
wt_type,
feature_wt_exp,
obs_wt)
# factorize matrix
session = wals.simple_train(model, input_tensor, num_iters)
tf.logging.info('Train Finish: {:%Y-%m-%d %H:%M:%S}'.format(datetime.datetime.now()))
# evaluate output factor matrices
output_row = row_factor.eval(session=session)
output_col = col_factor.eval(session=session)
# close the training session now that we've evaluated the output
session.close()
return output_row, output_col
def train_model(args, tr_sparse):
"""Instantiate WALS model and use "simple_train" to factorize the matrix.
Args:
args: training args containing hyperparams
tr_sparse: sparse training matrix
Returns:
the row and column factors in numpy format.
"""
dim = args['latent_factors']
num_iters = args['num_iters']
reg = args['regularization']
unobs = args['unobs_weight']
wt_type = args['wt_type']
feature_wt_exp = args['feature_wt_exp']
obs_wt = args['feature_wt_factor']
tf.logging.info('Train Start: {:%Y-%m-%d %H:%M:%S}'.format(datetime.datetime.now()))
# generate model
input_tensor, row_factor, col_factor, model = wals.wals_model(tr_sparse,
dim,
reg,
unobs,
args['weights'],
wt_type,
feature_wt_exp,
obs_wt)
# factorize matrix
session = wals.simple_train(model, input_tensor, num_iters)
tf.logging.info('Train Finish: {:%Y-%m-%d %H:%M:%S}'.format(datetime.datetime.now()))
# evaluate output factor matrices
output_row = row_factor.eval(session=session)
output_col = col_factor.eval(session=session)
# close the training session now that we've evaluated the output
session.close()
return output_row, output_col
def train_model(tr_sparse):
"""Instantiate WALS model and use "simple_train" to factorize the matrix.
Args:
args: training args containing hyperparams
tr_sparse: sparse training matrix
Returns:
the row and column factors in numpy format.
"""
dim = 5
num_iters = 20
reg = 0.07
unobs = 0.01
wt_type = 0
feature_wt_exp = 0.08
obs_wt = 130
tf.logging.info('Train Start: {:%Y-%m-%d %H:%M:%S}'.format(
datetime.datetime.now()))
# generate model
input_tensor, row_factor, col_factor, model = wals.wals_model(
tr_sparse, dim, reg, unobs, True, wt_type, feature_wt_exp, obs_wt)
# factorize matrix
session = wals.simple_train(model, input_tensor, num_iters)
print('end')
tf.logging.info('Train Finish: {:%Y-%m-%d %H:%M:%S}'.format(
datetime.datetime.now()))
# evaluate output factor matrices
output_row = row_factor.eval(session=session)
output_col = col_factor.eval(session=session)
# close the training session now that we've evaluated the output
session.close()
return output_row, output_col
