this is for holding javascript data
Roland Szabo edited experiments.tex
almost 10 years ago
Commit id: 118ce1a402686c77bae25b6eeeffe4500627edf0
deletions | additions
diff --git a/experiments.tex b/experiments.tex
index 9579a46..03c4497 100644
--- a/experiments.tex
+++ b/experiments.tex
...
Table Y contains the average, maximum and minimum values obtained for the F1 measure of the character segmentation problem. The F1 measure is used instead of the accuracy because the two classes are imbalanced: there are 11475 data points which indicate a segmentation point, while there are 19402 points which are not segmentation points, almost twice as many.
\begin{table}[h]
\caption{The errors for the character recognition experiment}
\label{table:seg_values}
\begin{tabular}{llllll}
\hline
Kernel type & Regularization & Min & Max & Mean & Std. dev. \\ \hline
RBF & 0.01 & 0.09141 & 0.09207 & 0.09165 & 0.00030 \\
Linear & 0.01 & 0.71585 & 0.71768 & 0.71672 & 0.00075 \\
RBF & 1 & 0.59146 & 0.60671 & 0.60130 & 0.00697 \\
Linear & 1 & 0.90372 & 0.90610 & 0.90490 & 0.00097 \\
RBF & 100 & 0.90854 & 0.91159 & 0.91018 & 0.00126 \\
Linear & 100 & 0.89695 & 0.90006 & 0.89860 & 0.00128 \\
RBF & 1000 & 0.90183 & 0.91341 & 0.90795 & 0.00475 \\
Linear & 1000 & 0.89207 & 0.89762 & 0.89453 & 0.00231 \\
RBF & 10000 & 0.90671 & 0.91220 & 0.90957 & 0.00225 \\
Linear & 10000 & 0.89146 & 0.89762 & 0.89494 & 0.00258 \\ \hline
\end{tabular}
\end{table}
Table Z contains the confusion matrix for the best experiment on the character recognition problem.
Table W contains the confusion matrix for the best experiment on the character segmentation problem.