Adobe PDF (1.44 MB)
Title Details:
Matrix Factorization Algorithms
Authors: Symeonidis, Panagiotis
Description:
Abstract:
In chapter five we will study the basic algorithms for matrix analysis and matrix factorization. We will define the problem of recommender items to users as a two-dimensional matrix that holds the scores that the latter give to the former. We will also define an objective function for minimizing the difference between a user's actual and predicted score for an item. Next, we will describe Singular Value Decomposition and Alternating Least Squares Table Analysis. This will be followed by the definition of higher-order tables (tensors), where more entities (users, items, item categories etc.) are involved, and the description of tensor analysis and factorization algorithms. Finally, we will define the problem as pairwise ranking, which means that we will attempt to predict the correct ranking of the items in the list of recommendations pertaining to a user. More specifically, we will define a new objective function whose objective will be to always recommend in a better ranking an item with which the user has interacted in the past compared to some other item with which there has been no interaction.
Linguistic Editors: Sakellarios, Michalis
Technical Editors: Karatzidis, Dimitrios
Graphic Editors: Symeonidis, Panagiotis
Type: Chapter
Creation Date: 29-05-2023
Item Details:
License: Attribution - NonCommercial - ShareAlike 4.0 International (CC BY-NC-SA 4.0)
Handle http://hdl.handle.net/11419/9577
Bibliographic Reference: Symeonidis, P. (2023). Matrix Factorization Algorithms [Chapter]. In Symeonidis, P. 2023. Intelligent Recommender Systems [Postgraduate textbook]. Kallipos, Open Academic Editions. https://hdl.handle.net/11419/9577
Language: Greek
Is Part of: Intelligent Recommender Systems
Publication Origin: Kallipos, Open Academic Editions