Unit:
Department of Informatics and TelecommunicationsΠληροφορική
Author:
PALAIOLOGOU THEONI
Supervisors info:
Ιωάννης Ιωαννίδης, Καθηγητής, Τμήμα Πληροφορικής και Τηλεπικοινωνιών, ΕΚΠΑ
Original Title:
Αναπαράσταση Τεχνικών Μηχανικής Μάθησης για Σύσταση Αντικειμένων στην Άλγεβρα Μονοπατιών RecGraph
Translated title:
Representing Machine Learning Techniques for Data Recommendation in the RecGraph Algebra
Summary:
A recommendation system is a subclass of information filtering systems that offer content to users that is relevant to their interests. Studying the anarchic recommendation systems space, the abundance of different recommendation approaches available, the problems associated with the complex, heterogeneous data domain it requires, and the consequences of this complexity for both the user and the system: the algebraic model “RecGraph” was proposed. A different approach to the problem of recommendations which redefines the recommendation as a path computational problem in a graph data model. In this thesis we will examine the expressivity of the RecGraph language and its suitability for representing algorithmic recommendations based on machine learning techniques. We will focus on machine learning techniques for recommendation with a hidden level of neurons and we will show their reduction in RecGraph language. This reduction proves that the RecGraph system can describe, with the appropriate extensions, a range of recommendation algorithms.
Main subject category:
Science
Keywords:
Recommender Systems, Graph Databases, Funk-Singular Value Decomposition, Machine Learning
thesis_TheoniPalaiologou_tbf.pdf (1 MB)
Open in new window
thesis_mainpack.zip
5 KB
File access is restricted.