Προσωποποιημένες συνεργατικές μέθοδοι συστάσεων
View/
Keywords
Συνεργατική μάθησηAbstract
This thesis presents federated learning as an appropriate way to create machine learning models where user data is protected even if it is distributed on remote devices. In addition, there are presented the recommender systems and the way in which they are trained and decide the appropriate proposals for each user. Then, it is described the theoretical framework of the implementation which concerns the creation of a federated recommender system that consists of an embedding neural network that will be trained in a federated way. The data set that was used consists of the ratings given by some users to a number of movies. After the training, the model is able to predict the rating that a user is likely to give to a particular movie, so that if it is high enough the movie will be recommended to this user to watch. Furthermore, a personalized approach is used to calculate the degree to which each user’s local model is mixed with the global model in order to optimize the performance of the model.
Postgraduate Studies Programme
Πληροφοριακά Συστήματα και ΥπηρεσίεςDepartment
Σχολή Τεχνολογιών Πληροφορικής και Επικοινωνιών. Τμήμα Ψηφιακών ΣυστημάτωνNumber of pages
72Language
GreekCollections
Except where otherwise noted, this item's license is described as
Αναφορά Δημιουργού-Μη Εμπορική Χρήση-Όχι Παράγωγα Έργα 3.0 Ελλάδα
Αναφορά Δημιουργού-Μη Εμπορική Χρήση-Όχι Παράγωγα Έργα 3.0 Ελλάδα