def get_predictor(args):
from neural_networks.update_manager_k import get_update_manager
from neural_networks.recurrent_layers import get_recurrent_layers
updater = get_update_manager(args)
recurrent_layer = get_recurrent_layers(args)
sequence_noise = get_sequence_noise(args)
target_selection = get_target_selection(args)
from neural_networks.rnn_oh_keras import RNNOneHotK
backend = 'tensorflow'
return RNNOneHotK(mem_frac=args.mem_frac,
backend=backend,
max_length=args.max_length,
regularization=args.regularization,
updater=updater,
target_selection=target_selection,
sequence_noise=sequence_noise,
recurrent_layer=recurrent_layer,
batch_size=args.batch_size,
active_f=args.act,
tying=args.tying,
temperature=args.temp,
gamma=args.gamma,
iter=args.iter,
tying_new=args.tying_new,
attention=args.att)
def get_predictor(args):
from neural_networks.update_manager_k import get_update_manager
from neural_networks.recurrent_layers import get_recurrent_layers
updater = get_update_manager(args)
recurrent_layer = get_recurrent_layers(args)
from neural_networks.rnn_core_keras import RNNCore
backend = 'tensorflow'
return RNNCore(mem_frac=args.mem_frac,
backend=backend,
max_length=args.max_length,
regularization=args.regularization,
updater=updater,
recurrent_layer=recurrent_layer,
batch_size=args.batch_size,
active_f=args.act)
def get_predictor(args):
args.layers = map(int, args.layers.split('-'))
updater = get_update_manager(args)
recurrent_layer = get_recurrent_layers(args)
sequence_noise = get_sequence_noise(args)
target_selection = get_target_selection(args)
if args.method == "MF":
return Factorization()
elif args.method == "BPRMF":
return BPRMF(k=args.hidden, reg = args.regularization, learning_rate = args.learning_rate, annealing=args.cooling, init_sigma = args.init_sigma, adaptive_sampling=(not args.no_adaptive_sampling), sampling_bias=args.fpmc_bias)
elif args.method == "FISM":
if args.clusters > 0:
return FISMCluster(h=args.hidden, reg=args.regularization, alpha=args.fism_alpha, loss=args.loss, interactions_are_unique=(not args.repeated_interactions), predict_with_clusters=(not args.ignore_clusters), sampling_bias=args.sampling_bias, sampling=args.sampling, cluster_sampling=args.c_sampling, init_scale=args.init_scale, scale_growing_rate=args.scale_growing_rate, max_scale=args.max_scale, n_clusters=args.clusters, cluster_type=args.cluster_type, updater=updater, target_selection=target_selection, sequence_noise=sequence_noise, recurrent_layer=recurrent_layer, use_ratings_features=args.rf, use_movies_features=args.mf, use_users_features=args.uf, batch_size=args.batch_size)
else:
return FISM(k=args.hidden, reg = args.regularization, learning_rate = args.learning_rate, annealing=args.cooling, init_sigma = args.init_sigma, loss=args.loss, alpha=args.fism_alpha)
elif args.method == "Fossil":
return Fossil(k=args.hidden, order=args.fossil_order, reg = args.regularization, learning_rate = args.learning_rate, annealing=args.cooling, init_sigma = args.init_sigma, alpha=args.fism_alpha)
elif args.method == "FPMC":
return FPMC(k_cf = args.k_cf, k_mc = args.k_mc, reg = args.regularization, learning_rate = args.learning_rate, annealing=args.cooling, init_sigma = args.init_sigma, adaptive_sampling=(not args.no_adaptive_sampling), sampling_bias=args.fpmc_bias)
elif args.method == "LTM":
return LTM(k = args.hidden, alpha = args.ltm_damping, window = args.ltm_window, learning_rate=args.learning_rate, use_trajectory=(not args.ltm_no_trajectory))
elif args.method == "UKNN":
return UserKNN(neighborhood_size=args.ns)
elif args.method == "POP":
return Pop()
elif args.method == "MM":
return MarkovModel()
elif args.method == 'RNN':
if args.clusters > 0:
return RNNCluster(interactions_are_unique=(not args.repeated_interactions), max_length=args.max_length, cluster_selection_noise=args.csn, loss=args.loss, predict_with_clusters=(not args.ignore_clusters), sampling_bias=args.sampling_bias, sampling=args.sampling, cluster_sampling=args.c_sampling, init_scale=args.init_scale, scale_growing_rate=args.scale_growing_rate, max_scale=args.max_scale, n_clusters=args.clusters, cluster_type=args.cluster_type, updater=updater, target_selection=target_selection, sequence_noise=sequence_noise, recurrent_layer=recurrent_layer, use_ratings_features=args.rf, use_movies_features=args.mf, use_users_features=args.uf, batch_size=args.batch_size)
elif args.loss == 'CCE':
return RNNOneHot(interactions_are_unique=(not args.repeated_interactions), max_length=args.max_length, diversity_bias=args.diversity_bias, regularization=args.regularization, updater=updater, target_selection=target_selection, sequence_noise=sequence_noise, recurrent_layer=recurrent_layer, use_ratings_features=args.rf, use_movies_features=args.mf, use_users_features=args.uf, batch_size=args.batch_size)
elif args.loss in ['hinge', 'logit', 'logsig']:
return RNNMargin(interactions_are_unique=(not args.repeated_interactions), loss_function=args.loss, balance = args.balance, popularity_based = args.pb, min_access = args.min_access, target_selection=target_selection, sequence_noise=sequence_noise, recurrent_layer=recurrent_layer, max_length=args.max_length, updater=updater, use_ratings_features=args.rf, use_movies_features=args.mf, use_users_features=args.uf, batch_size=args.batch_size)
elif args.loss in ['BPR', 'TOP1', 'Blackout']:
return RNNSampling(interactions_are_unique=(not args.repeated_interactions), loss_function=args.loss, diversity_bias=args.diversity_bias, sampling=args.sampling, sampling_bias=args.sampling_bias, recurrent_layer=recurrent_layer, max_length=args.max_length, updater=updater, target_selection=target_selection, sequence_noise=sequence_noise, use_ratings_features=args.rf, use_movies_features=args.mf, use_users_features=args.uf, batch_size=args.batch_size)
else:
raise ValueError('Unknown loss for the RNN model')
elif args.method == "SDA":
return StackedDenoisingAutoencoder(interactions_are_unique=(not args.repeated_interactions), layers = args.layers, input_dropout=args.input_dropout, dropout=args.dropout, updater=updater, batch_size=args.batch_size, use_ratings_features=args.rf)
def predictor_command_parser(parser):
parser.add_argument('-m', dest='method', choices=['RNN', 'SDA', 'BPRMF', 'FPMC', 'FISM', 'Fossil', 'LTM', 'UKNN', 'MM', 'POP'],
help='Method', default='RNN')
parser.add_argument('-b', dest='batch_size', help='Batch size', default=16, type=int)
parser.add_argument('-l', dest='learning_rate', help='Learning rate', default=0.01, type=float)
parser.add_argument('-r', dest='regularization', help='Regularization (positive for L2, negative for L1)', default=0., type=float)
parser.add_argument('-g', dest='gradient_clipping', help='Gradient clipping', default=100, type=int)
parser.add_argument('-H', dest='hidden', help='Number of hidden neurons (for LTM and BPRMF)', default=20, type=int)
parser.add_argument('-L', dest='layers', help='Layers (for SDA)', default="20", type=str)
parser.add_argument('--db', dest='diversity_bias', help='Diversity bias (for RNN with CCE, TOP1, BPR or Blackout loss)', default=0.0, type=float)
parser.add_argument('--rf', help='Use rating features.', action='store_true')
parser.add_argument('--mf', help='Use movie features.', action='store_true')
parser.add_argument('--uf', help='Use users features.', action='store_true')
parser.add_argument('--ns', help='Neighborhood size (for UKNN).', default=80, type=int)
parser.add_argument('--cooling', help='Simulated annealing', default=1., type=float)
parser.add_argument('--init_sigma', help='Sigma of the gaussian initialization (for MF)', default=1, type=float
parser.add_argument('--no_adaptive_sampling', help='No adaptive sampling (for MF)', action='store_true')
parser.add_argument('--fpmc_bias', help='Sampling bias (for MF)', default=100., type=float)
parser.add_argument('--ltm_no_trajectory', help='Do not use users trajectory in LTM, just use word2vec', action='store_true')
parser.add_argument('--max_length', help='Maximum length of sequences during training (for RNNs)', default=30, type=int)
parser.add_argument('--repeated_interactions', help='The model can recommend items with which the user already interacted', action='store_true')
update_manager_command_parser(parser)
recurrent_layers_command_parser(parser)
sequence_noise_command_parser(parser)
target_selection_command_parser(parser)
def get_predictor(args):
args.layers = map(int, args.layers.split('-'))
updater = get_update_manager(args)
recurrent_layer = get_recurrent_layers(args)
sequence_noise = get_sequence_noise(args)
target_selection = get_target_selection(args)
if args.method == "MF":
return BPRMF(k=args.hidden, reg = args.regularization, learning_rate = args.learning_rate, annealing=args.cooling, init_sigma = args.init_sigma, adaptive_sampling=(not args.no_adaptive_sampling), sampling_bias=args.fpmc_bias)
elif args.method == "UKNN":
return UserKNN(neighborhood_size=args.ns)
elif args.method == "MM":
return MarkovModel()
elif args.method == 'RNN':
return RNNOneHot(interactions_are_unique=(not args.repeated_interactions), max_length=args.max_length, diversity_bias=args.diversity_bias, regularization=args.regularization, updater=updater, target_selection=target_selection, sequence_noise=sequence_noise, recurrent_layer=recurrent_layer, use_ratings_features=args.rf, use_movies_features=args.mf, use_users_features=args.uf, batch_size=args.batch_size)
def get_predictor(args):
# args.layers = map(int, args.layers.split('-'))
updater = get_update_manager(args)
recurrent_layer = get_recurrent_layers(args)
sequence_noise = get_sequence_noise(args)
target_selection = get_target_selection(args)
if args.method == 'RNN':
if args.clusters > 0:
return RNNCluster(interactions_are_unique=(not args.repeated_interactions), max_length=args.max_length,
cluster_selection_noise=args.csn, loss=args.loss,
predict_with_clusters=(not args.ignore_clusters), sampling_bias=args.sampling_bias,
sampling=args.sampling, cluster_sampling=args.c_sampling, init_scale=args.init_scale,
scale_growing_rate=args.scale_growing_rate, max_scale=args.max_scale,
n_clusters=args.clusters, cluster_type=args.cluster_type, updater=updater,
target_selection=target_selection, sequence_noise=sequence_noise,
recurrent_layer=recurrent_layer, use_ratings_features=args.rf,
use_movies_features=args.mf, use_users_features=args.uf, batch_size=args.batch_size)
elif args.loss == 'CCE':
return RNNOneHot(interactions_are_unique=(not args.repeated_interactions), max_length=args.max_length,
diversity_bias=args.diversity_bias, regularization=args.regularization, updater=updater,
target_selection=target_selection, sequence_noise=sequence_noise,
recurrent_layer=recurrent_layer, use_ratings_features=args.rf, use_movies_features=args.mf,
use_users_features=args.uf, batch_size=args.batch_size)
elif args.loss in ['hinge', 'logit', 'logsig']:
return RNNMargin(interactions_are_unique=(not args.repeated_interactions), loss_function=args.loss,
balance=args.balance, popularity_based=args.pb, min_access=args.min_access,
target_selection=target_selection, sequence_noise=sequence_noise,
recurrent_layer=recurrent_layer, max_length=args.max_length, updater=updater,
use_ratings_features=args.rf, use_movies_features=args.mf, use_users_features=args.uf,
batch_size=args.batch_size)
elif args.loss in ['BPR', 'TOP1', 'Blackout']:
return RNNSampling(interactions_are_unique=(not args.repeated_interactions), loss_function=args.loss,
diversity_bias=args.diversity_bias, sampling=args.sampling,
sampling_bias=args.sampling_bias, recurrent_layer=recurrent_layer,
max_length=args.max_length, updater=updater, target_selection=target_selection,
sequence_noise=sequence_noise, use_ratings_features=args.rf, use_movies_features=args.mf,
use_users_features=args.uf, batch_size=args.batch_size)
else:
raise ValueError('Unknown loss for the RNN model')