ANN can be used to combine (or hybridize) the input from several recommen-
dation modules or data sources. Hsu et al. [30], for instance, build a TV recom-
mender by importing data from four different sources: user profiles and stereo-
types; viewing communities; program metadata; and viewing context. They use the
back-propagation algorithm to train a three-layered neural network. Christakou and Stafylopatis [19] also built a hybrid content-based CF RS. The content-based rec-
ommender is implemented using three neural networks per user, each of them cor-
responding to one of the following features: “kinds”, “stars”, and “synopsis”. They
trained the ANN using the Resilient Backpropagation method.