diff --git a/recommender.tex b/recommender.tex new file mode 100644 index 0000000000000000000000000000000000000000..3ff45f9c43468c88ca46cea38b4f88170ca9fd9f --- /dev/null +++ b/recommender.tex @@ -0,0 +1,10 @@ +\section{A Study on Recommender Systems} +This section explains the basics of \textit{recommender systems} necessary for the essential understanding of the paper presented. Besides the general definition of the \textit{recommender problem}, the corresponding solution approaches are presented. Furthermore, the main focus will be on the solution approach of \textit{matrix factorization}. + +\subsection{Recommender Problem} +The \textit{recommender problem} consists of the entries of the sets $\mathcal{U}$ and $\mathcal{I}$, where $\mathcal{U}$ represents the set of all \textit{users} and $\mathcal{I}$ the set of all \textit{items}. +Each of the \textit{users} in $\mathcal{U}$ gives \textit{ratings} from a set $\mathcal{S}$ of possible scores for the available \textit{items} in $\mathcal{I}$. The resulting \textit{rating-matrix} $\mathcal{R}$ is composed of $\mathcal{R} = \mathcal{U} \times \mathcal{I}$. The entries in $\mathcal{R}$ indicate the \textit{rating} from \textit{user} $u \in \mathcal{U}$ to \textit{item} $i \in \mathcal{I}$. This entry is then referred to as $r_{ui}$. Due to incomplete \textit{item-ratings}, $\mathcal{R}$ may also be incomplete. In the following, the subset of all \textit{users} who have rated a particular \textit{item} $i$ is referred to as $\mathcal{U}_i$. Similarly, $\mathcal{I}_u$ refers to the subset of \textit{items} that were rated by \textit{user} $u$. Since $\mathcal{R}$ is not completely filled, there are missing values for some \textit{user-item relations}. The aim of the \textit{recommender system} is to estimate the missing \textit{ratings} $\hat{r}_{ui}$ using a \textit{prediction-function} $p(u,i)$. The \textit{prediction-function} consists of $p: \mathcal{U} \times \mathcal{I} \rightarrow \mathcal{S}$ \citep{DeKa11}. In the further course of the work different methods are presented to determine $p(u,i)$. + +In the following, the two main approaches of \textit{collaborative-filtering} and \textit{content-based} \textit{recommender systems} will be discussed. In addition, it is explained how \textit{matrix factorization} can be integrated into the two ways of thinking. + +\subsection{Collaborative-Filtering} \ No newline at end of file diff --git a/references.bib b/references.bib index 71e2152dacc11067a3378e702565dab6b4ab7f4f..5a3c6e906b53990b574ae5187ca47722ec945787 100644 --- a/references.bib +++ b/references.bib @@ -27,3 +27,14 @@ biburl = {https://dblp.org/rec/bib/journals/corr/abs-1905-01395}, bibsource = {dblp computer science bibliography, https://dblp.org} } +@inproceedings{DeKa11, +author = {Desrosiers, Christian and Karypis, George}, +year = {2011}, +month = {01}, +pages = {107--144}, +title = {A Comprehensive Survey of Neighborhood-Based Recommendation Methods}, +booktitle = {Recommender Systems Handbook}, +editor = {P.B. Kantor and F. Ricci and L. Rokach and B. Shapira}, +publisher={Springer}, +doi = {10.1007/978-0-387-85820-3_4} +} \ No newline at end of file diff --git a/submission.pdf b/submission.pdf index 1a19cab2d18e8f5381775cda9beea5fa47bce19e..f37eae4f5c4d863d89944736146a898a6f2bdde4 100644 Binary files a/submission.pdf and b/submission.pdf differ diff --git a/submission.tex b/submission.tex index d98a4976b53cd02cfbd01c3e416791c0ca0d28fb..b0c8b343ac024d67a46e1065424c6b1b258669d9 100644 --- a/submission.tex +++ b/submission.tex @@ -31,7 +31,10 @@ \parindent 6pt \usepackage[]{titlesec} \titlespacing*{\section} -{0pt}{0pt}{6pt} +{0pt}{6pt}{6pt} +\usepackage[]{titlesec} +\titlespacing*{\subsection} +{0pt}{6pt}{6pt} \usepackage{footmisc} \setlength{\abovedisplayskip}{0pt} \renewcommand{\footrulewidth}{0.5pt} @@ -62,7 +65,7 @@ A Study on Recommender Systems} % Hier beginnt der Inhalt! % %%%%%%%%%%%%%%%%%%%%%%%%%%%%%% \input{introduction} - +\input{recommender} \newpage \bibliography{references} \bibliographystyle{plainnat}