def run(
features,
item_features_matrix,
interactions,
user_features,
user_features_matrix,
cf_model,
lsiup_model,
n_iter,
test_size,
cold_start,
learning_rate,
no_components,
a_alpha,
b_alpha,
epochs,
):
logger.debug("Fitting the model with %s", locals())
no_interactions = len(interactions.data)
if cf_model:
logger.info("Fitting the CF model")
modelfnc = lambda: CFModel(dim=no_components)
elif lsiup_model:
logger.info("Fitting the LSI-UP model")
modelfnc = lambda: LsiUpModel(dim=no_components)
else:
modelfnc = lambda: LightFM(
learning_rate=learning_rate, no_components=no_components, item_alpha=a_alpha, user_alpha=b_alpha
)
model, auc = fit_model(
interactions=interactions,
item_features_matrix=item_features_matrix,
n_iter=n_iter,
epochs=epochs,
modelfnc=modelfnc,
test_size=test_size,
cold_start=cold_start,
user_features_matrix=user_features_matrix,
)
logger.debug("Average AUC: %s", auc)
if not cf_model and not lsiup_model:
model.add_item_feature_dictionary(features.feature_ids, check=False)
try:
# Can only get similar tags if we have tag features
test_features = ("tag:bic", "tag:survival", "tag:regression", "tag:mcmc")
for test_feature in test_features:
logger.debug(
"Features most similar to %s: %s", test_feature, model.most_similar(test_feature, "item", number=10)
)
except KeyError:
pass
return auc
def run(features,
item_features_matrix,
interactions,
user_features,
user_features_matrix,
cf_model,
lsiup_model,
n_iter,
test_size,
cold_start,
learning_rate,
no_components,
a_alpha,
b_alpha,
epochs):
logger.debug('Fitting the model with %s', locals())
no_interactions = len(interactions.data)
if cf_model:
logger.info('Fitting the CF model')
modelfnc = lambda: CFModel(dim=no_components)
elif lsiup_model:
logger.info('Fitting the LSI-UP model')
modelfnc = lambda: LsiUpModel(dim=no_components)
else:
modelfnc = lambda: LightFM(learning_rate=learning_rate,
no_components=no_components,
item_alpha=a_alpha,
user_alpha=b_alpha)
model, auc = fit_model(interactions=interactions,
item_features_matrix=item_features_matrix,
n_iter=n_iter,
epochs=epochs,
modelfnc=modelfnc,
test_size=test_size,
cold_start=cold_start,
user_features_matrix=user_features_matrix)
logger.debug('Average AUC: %s', auc)
if not cf_model and not lsiup_model:
model.add_item_feature_dictionary(features.feature_ids, check=False)
try:
# Can only get similar tags if we have tag features
test_features = ('tag:bic',
'tag:survival',
'tag:regression',
'tag:mcmc')
for test_feature in test_features:
logger.debug('Features most similar to %s: %s',
test_feature,
model.most_similar(test_feature, 'item', number=10))
except KeyError:
pass
return auc
def run(features,
item_features_matrix,
interactions,
cf_model,
lsiup_model,
n_iter,
test_size,
cold_start,
learning_rate,
no_components,
a_alpha,
b_alpha,
epochs):
logger.debug('Fitting the model with %s', locals())
no_interactions = len(interactions.data)
if cf_model:
logger.info('Fitting the CF model')
modelfnc = lambda: CFModel(dim=no_components)
elif lsiup_model:
logger.info('Fitting the LSI-UP model')
modelfnc = lambda: LsiUpModel(dim=no_components)
else:
modelfnc = lambda: LightFM(learning_rate=learning_rate,
no_components=no_components,
item_alpha=a_alpha,
user_alpha=b_alpha)
model, auc = fit_model(interactions=interactions,
item_features_matrix=item_features_matrix,
n_iter=n_iter,
epochs=epochs,
modelfnc=modelfnc,
test_size=test_size,
cold_start=cold_start)
logger.debug('Average AUC: %s', auc)
if not cf_model and not lsiup_model:
model.add_item_feature_dictionary(features.feature_ids, check=False)
features.add_latent_representations(model.item_features)
titles = ('Lord of the Rings: The Two Towers, The (2002)',
'Toy Story (1995)',
'Terminator, The (1984)',
'Europa Europa (Hitlerjunge Salomon) (1990)')
for title in titles:
logger.debug('Most similar movies to %s: %s', title,
features.most_similar_movie(title, number=20))
# Can only get similar tags if we have tag features
test_features = ('genome:art house',
'genome:dystopia',
'genome:bond')
for test_feature in test_features:
try:
logger.debug('Features most similar to %s: %s',
test_feature,
model.most_similar(test_feature, 'item', number=10))
except KeyError:
pass
return auc
