summaryrefslogtreecommitdiff
diff options
context:
space:
mode:
authorAndré Nusser <andre.nusser@googlemail.com>2020-02-09 18:37:04 +0100
committerAndré Nusser <andre.nusser@googlemail.com>2020-02-09 18:37:04 +0100
commit386dc049ed661feffd72d53c5e1045b142764eb2 (patch)
tree09099cacc4417c8f481a7e6dbd1f148e531e8463
parent630de07ec76cea8d7dff038afa05a90e3ec8967a (diff)
Add figures for some of the experiments.
-rw-r--r--sampling_alg_lac2020/LAC-20.tex38
-rw-r--r--sampling_alg_lac2020/figures/112_new.pdfbin0 -> 12060 bytes
-rw-r--r--sampling_alg_lac2020/figures/112_old.pdfbin0 -> 12476 bytes
-rw-r--r--sampling_alg_lac2020/figures/16_new.pdfbin0 -> 11895 bytes
-rw-r--r--sampling_alg_lac2020/figures/16_old.pdfbin0 -> 12218 bytes
-rw-r--r--sampling_alg_lac2020/figures/48_new.pdfbin0 -> 12024 bytes
-rw-r--r--sampling_alg_lac2020/figures/48_old.pdfbin0 -> 12722 bytes
-rw-r--r--sampling_alg_lac2020/figures/80_new.pdfbin0 -> 11860 bytes
-rw-r--r--sampling_alg_lac2020/figures/80_old.pdfbin0 -> 11918 bytes
-rw-r--r--sampling_alg_lac2020/figures/sweep_new.pdfbin0 -> 14089 bytes
-rw-r--r--sampling_alg_lac2020/figures/sweep_old.pdfbin0 -> 14335 bytes
11 files changed, 34 insertions, 4 deletions
diff --git a/sampling_alg_lac2020/LAC-20.tex b/sampling_alg_lac2020/LAC-20.tex
index 4acd2db..93b63dd 100644
--- a/sampling_alg_lac2020/LAC-20.tex
+++ b/sampling_alg_lac2020/LAC-20.tex
@@ -541,7 +541,9 @@ Note that by traversing the samples in order of their distance to $p$ (which is
To conduct the experiments, we use the implementation of the new sampling algorithm in DrumGizmo.
As a base-line for comparison, we use the previous sample selection algorithm of DrumGizmo.
We want to evaluate how the sample selection algorithm performs in practice, therefore we use the drum kits of DrumGizmo. In particular, we use the power value distribution of the samples of those kits.
-We do experiments on three different drum kits. See Table \ref{tab:drumkit_data} for some information about the samples of the drum kits and see Figure \ref{fig:power_level_distribution} for a visualization of the power level distribution of the kits.
+We do experiments on three different drum kits.
+As normally the snare drum is the instrument with the highest number of samples, we choose this drum for our experiments.
+See Table \ref{tab:drumkit_data} for some information about the samples of the drum kits and see Figure \ref{fig:power_level_distribution} for a visualization of the power level distribution of the kits.
\begin{table}
\caption{Drum kit data.}
@@ -574,12 +576,39 @@ We want to test the following hypotheses with our experiments:
% \todoandre{Experiments are: playing fast sweeps (with multiple hits per velocity); playing a single note over and over again at the same velocity; sound examples that people can listen to online?}
To test the above hypotheses, we conduct the following experiments:
\begin{enumerate}
- \item Playing fast sweeps over the whole velocity range.
- \item Playing a single note at high velocity.
- \item Audio recordings of the processes
+ \item Playing fast sweeps from MIDI velocity 0 to MIDI velocity 127, 8 times
+ \item Playing a single velocity fast for 1015 times
+ % \item Audio recordings of the processes
\end{enumerate}
\todoandre{Do beautiful tables and plots here}
+To be able to compare the results of those experiments, we have to somehow visualize the selections. We choose to plot histograms. More concretely, for a run of the experiment, we plot the number of times a sample was selected over its index in the array of ordered samples. For the first experiment, i.e., the sweeps, you can see the histograms in Figure \ref{fig:sweep}. \todo{describe and explain plot}
+
+\begin{figure*}
+ \includegraphics[width=.47\linewidth]{figures/sweep_old.pdf}
+ \hfill
+ \includegraphics[width=.47\linewidth]{figures/sweep_new.pdf}
+ \caption{The histogram of the old algorithm (left) and the new algorithm (right) of playing fast sweeps from MIDI velocity 0 to MIDI velocity 127, 8 times.}
+ \label{fig:sweep}
+\end{figure*}
+
+The second experiment we conducted for two different MIDI velocities: MIDI velocity 80 (Figure \ref{fig:80}) and MIDI velocity 112 (Figure \ref{fig:112}). \todo{describe and explain plot}
+
+\begin{figure*}
+ \includegraphics[width=.47\linewidth]{figures/80_old.pdf}
+ \hfill
+ \includegraphics[width=.47\linewidth]{figures/80_new.pdf}
+ \caption{The histogram of the old algorithm (left) and the new algorithm (right) of playing MIDI velocity 80 fast for 1015 times.}
+ \label{fig:80}
+\end{figure*}
+
+\begin{figure*}
+ \includegraphics[width=.47\linewidth]{figures/112_old.pdf}
+ \hfill
+ \includegraphics[width=.47\linewidth]{figures/112_new.pdf}
+ \caption{The histogram of the old algorithm (left) and the new algorithm (right) of playing MIDI velocity 112 fast for 1015 times.}
+ \label{fig:112}
+\end{figure*}
% \todoandre{Also do an experiment regarding the adaptive search}
To also get an idea of the performance of the new sampling algorithm, we want to see how many power values of samples are evaluated per query.
@@ -596,6 +625,7 @@ Without the smart search optimization described at the end of Section \ref{sec:i
\section{Conclusion and Future Work}
\todoandre{Recapitulate what was done in this paper. Highlight some of the difficulties and surprises.}
\todoandre{List future work: transforming the loudness space; refine the objective function; adapt algorithm to other instruments/settings; study to see what sounds good to people and do they actually hear the difference?}
+\todoandre{Change diversity parameter. Add a bound on the deviation from the requested power value to definitely avoid extrem outliers.}
% \section{Acknowledgements}
% \todo{Thank people for testing?}
diff --git a/sampling_alg_lac2020/figures/112_new.pdf b/sampling_alg_lac2020/figures/112_new.pdf
new file mode 100644
index 0000000..01c4773
--- /dev/null
+++ b/sampling_alg_lac2020/figures/112_new.pdf
Binary files differ
diff --git a/sampling_alg_lac2020/figures/112_old.pdf b/sampling_alg_lac2020/figures/112_old.pdf
new file mode 100644
index 0000000..39b2e40
--- /dev/null
+++ b/sampling_alg_lac2020/figures/112_old.pdf
Binary files differ
diff --git a/sampling_alg_lac2020/figures/16_new.pdf b/sampling_alg_lac2020/figures/16_new.pdf
new file mode 100644
index 0000000..a332945
--- /dev/null
+++ b/sampling_alg_lac2020/figures/16_new.pdf
Binary files differ
diff --git a/sampling_alg_lac2020/figures/16_old.pdf b/sampling_alg_lac2020/figures/16_old.pdf
new file mode 100644
index 0000000..94c3dcb
--- /dev/null
+++ b/sampling_alg_lac2020/figures/16_old.pdf
Binary files differ
diff --git a/sampling_alg_lac2020/figures/48_new.pdf b/sampling_alg_lac2020/figures/48_new.pdf
new file mode 100644
index 0000000..d68096a
--- /dev/null
+++ b/sampling_alg_lac2020/figures/48_new.pdf
Binary files differ
diff --git a/sampling_alg_lac2020/figures/48_old.pdf b/sampling_alg_lac2020/figures/48_old.pdf
new file mode 100644
index 0000000..8674ed8
--- /dev/null
+++ b/sampling_alg_lac2020/figures/48_old.pdf
Binary files differ
diff --git a/sampling_alg_lac2020/figures/80_new.pdf b/sampling_alg_lac2020/figures/80_new.pdf
new file mode 100644
index 0000000..a4adc3a
--- /dev/null
+++ b/sampling_alg_lac2020/figures/80_new.pdf
Binary files differ
diff --git a/sampling_alg_lac2020/figures/80_old.pdf b/sampling_alg_lac2020/figures/80_old.pdf
new file mode 100644
index 0000000..6bd427e
--- /dev/null
+++ b/sampling_alg_lac2020/figures/80_old.pdf
Binary files differ
diff --git a/sampling_alg_lac2020/figures/sweep_new.pdf b/sampling_alg_lac2020/figures/sweep_new.pdf
new file mode 100644
index 0000000..d7a645e
--- /dev/null
+++ b/sampling_alg_lac2020/figures/sweep_new.pdf
Binary files differ
diff --git a/sampling_alg_lac2020/figures/sweep_old.pdf b/sampling_alg_lac2020/figures/sweep_old.pdf
new file mode 100644
index 0000000..c623f19
--- /dev/null
+++ b/sampling_alg_lac2020/figures/sweep_old.pdf
Binary files differ