this is for holding javascript data
Dylan Freedman added results3.tex
about 9 years ago
Commit id: 64390f3fe08eeac3c3d1ebb528d8c96b4c4c203a
deletions | additions
diff --git a/results3.tex b/results3.tex
new file mode 100644
index 0000000..db9c85a
--- /dev/null
+++ b/results3.tex
...
\subsection{Clustering}
One natural experiment after obtaining data corresponding to all the pairwise combinations of songs in a dataset, is to attempt clustering. I used the Markov Cluster Algorithm (MCL)\footnote{http://micans.org/mcl/} with custom-tuned parameters to attempt to find groups of Billboard 2014 songs that were connected well with one another. After some experimentation with parameters, the largest cluster result returned contained:\\
\{\textit{Call Me Maybe}\text{ by Carly Rae Jepsen}, \textit{Heart Attack}\text{ by Demi Lovato}, \textit{Whistle}\text{ by FloRide}, \textit{Cruise}\text{ by Florida Georgia Line}, \textit{Stereo Hearts ft. Adam Levine}\text{ by Gym Class Heroes}, \textit{The Edge of Glory}\text{ by Lady Gaga}, \textit{Live While We're Young}\text{ by One Direction}, \textit{Try}\text{ by P!nk}, and \textit{If I Die Young}\text{ by The Band Perry}\}
\subsection{Visualization}
\subsection{Ranking Fully Connected Pairwise Comparisons}
\subsection{Ranking Random N-Gram Search}
\subsection{Key-Finding Accuracy}
\section{Smith-Waterman Results}
\subsection{Normalization}
\item Problem, just given a raw score independent of song length or anything else
\item Attempt to solve: $normalization$ $equation$ $used$
\subsection{Clustering}
\item Tried out using graph clustering algorithm
\item Future: try k-means
\section{Visualization}
\subsection{Motivation}
\item Given over 100 pairwise (triangle) number comparisons between sets of songs, it is important to have a useful means of visualizing the data
\item Initially tried using Graphos and Circos and other graphing libraries
\item Demanded a level of interactivity unachievable with other programs
\subsection{Using D3.js}
D3, a Javascript library, if used to create the visualization. All songs are displayed as radial nodes around a circle, and pairwise connections as edges.
\item Using popular javascript graphing library, can construct a visualization based on hierarchical edge bundling, as used in other modern graphing libraries (curved edges using tension factors)
\item Can add interactivity using web server
\subsection{Using Twistd webserver}
\item Facile, easy use web server all in Python. Link C libraries to Python and have a rapid coding environment that has fast native C-code at its base
\item Can compare align results in a visual interface based on reconstructing the aligned data
\subsection{Audio playback}
\item Audio transposition problem, and solution using fewest localized transpositions
\item Following along using popcorn.js
\subsection{Public implementation}
\item Webserver available at chordmatch.com