p@36: 
p@36: \section{Hybrid Music Recommender}
p@36: 
p@36: The hybrid music recommender approach is an implementation of feature augmentation and meta-level methods. One advantage of the meta-level method is the use of compressed users and songs information instead of sparse raw data.
p@36: %\begin{itemize}
p@36: %\item We model bananas and plums using a density estimation of the colour space.
p@36: %\item We fit a Gaussian Mixture Model using Expectation Maximization. We select the number of clusters using an MDL criteria.
p@36: %\item Typical image examples of bananas and plums look like this:
p@36: %\end{itemize}
p@36: 
p@36: \begin{center}
p@36: 	\resizebox*{0.9\columnwidth}{!}{\includegraphics{images/diagram_hybrid_music_recommender.eps}}\\
p@36: 	{\large \textbf{Fig. 1.} Diagram of our hybrid music recommender approach}
p@36: %\begin{tabular}{c@{ }c}
p@36: %\resizebox*{0.45\columnwidth}{!}{\includegraphics{images/diagram_hybrid_music_recommender.eps}} &
p@36: %\resizebox*{0.45\columnwidth}{!}{\includegraphics{images/fruit/plums.eps}} \\
p@36: %Bananas & Plums \\
p@36: %\end{tabular}
p@36: \end{center}