\begin{figure}[!t]
\begin{center}
\begin{tabular}{|r|r|r|r|r|}\cline{3-4}
 \multicolumn{2}{c}{~}& \multicolumn{2}{|c|}{Execution Time} & \multicolumn{1}{c}{~}\\\hline
dataset     & Instances & W     & W2     & W2 speedup\\\hline
telecom1    & 18        & 0.07s  & 0.04s & 1.6x\\
coc81       & 63        & 0.43s  & 0.08s & 5.3x\\
nasa93      & 93        & 0.69s  & 0.10s & 6.6x\\
china	    & 500       & 7.37s  & 0.42s & 10.8x\\\hline
\end{tabular}
\end{center}
\caption{Average execution times for the W and W2 algorithms. By removing the $O(n^2)$ kth nearest neighbor calculation from W we drastically improve performance, especially on larger datasets such as China (500 instances).}
\label{fig:w2-runtimes}
\end{figure}