Disaster Strikes!

A wise old man once told me I was a fool of a Took, though now that I think of it, that might have been someone else...

The programs I left running over the weekend, never finished. A small mistake in my code meant that instead of moving to the next graph after the scale parameter is over 1000.0, it continued with the same graph with the same parameters. All down to one while loop; while(scale != 1000.0), a silly rookie mistake but not all is lost.

The programs did give some results for the graph 3elt, so I can use these to refine the tests further and highlight any possible issues. It has also shown me the importance of setting up remote desktop so I can see whats happening when at home (the computers are for hot-desking so do not have it set up, nor given useful machine aliases).

Unfortunately I will have to run the simulations again for those results I did not receive (losing more valuable time). I will edit this post when I've made a quick analysis of the results I do have.

<Results>

Delayed once again, I was fixing some irregularities in the different algorithms to give fairer results (one was calculating edge lengths as though in 3D, and the other had hard coded testing values), but all fixed now. Those tests have been left running and am expecting results back later today.

The results show that CA is 31.1% faster on average, than QT, but QT has less edge crossings on average. This is due to the affect of increasing the power of the repulsive forces; QT is little affected by this scalar and the edge crossings remain similar though out, whereas CA has large increases in edge crossings as the scalar increases.

These results are from running each algorithm once, with some discrepancies between the algorithms (QT had a hard coded 'size' of 100 instead of the 150 it should have been. Future results will show whether this has an effect on the edge crossings or not.

Results above are based of one runtime per scalar on each algorithm, for the graph 3elt. Numerical data from the tests will not be presented just yet until further results are finished.

</Results>