Ranking under near decomposability

• Main Author: Νικολακόπουλος, Αθανάσιος Ν.
• Other Authors: Γαροφαλάκης, Ιωάννης
• Format: Thesis
• Language: English
• Published: 2016
• Subjects: Ranking; Sparsity; Decomposability; Hierarchy; Link analysis; Top-N recommendation; Collaborative filtering; Dimensionality reduction; NCDawareRank; NCDREC; EigenRec; PageRank; Markov chains; Non-negative matrix theory; 005.741; Πρόβλημα παραγωγής διανυσμάτων κατάταξης; Αραιότητα; Ιεραρχία; Αλυσίδες Μαρκόφ

In this dissertation, we study the problem of Ranking in the presence of Sparsity focusing on two of the most important and generic ranking settings; namely Link Analysis and Top-N Recommendation. Building on the intuition behind Decomposability – the seminal work of the Nobel Laureate Economist and Turing Award winner Herbert A. Simon – we introduce a novel and versatile modeling approach that results to effective algorithmic frameworks for both application areas. The models and algorithms we propose are shown to possess a wealth of useful mathematical properties that imply favorable computational as well as qualitative characteristics. A comprehensive set of experiments on several real-world datasets verify the applicability of our methods in big-data scenarios as well as their promising performance in achieving high-quality results with respect to state-of-the-art link-analysis and recommendation algorithms.