Keeping reviewers, and myself, happy

This week I have been continuing with my implementation of the matching; attempting to fix the many problems it seems to have with my already existing "testing suite". Although many have been resolved or patched, there remains a larger issue when attempting to extend coordinates calculated in coarser graphs to the finer graphs (the dreaded NaN has returned).

Experience suggests this is a bug somewhere within my code where I am dividing by zero - normally due to calculating repulsive forces between a vertex and itself (resulting with a distance of zero). However, this only appears once so a more thorough examination of verbose output is required. I will update again if and when this is resolved.

Additionally, I have been working to alter my iV2012 contribution, requiring some investigation into why my MGF(CA) algorithm is faster that the OT for smaller graphs, but slows for larger graphs. Originally I hypothesised that this was due to the dimensions of the tree but failed to give any real description or evidence. Having looking into this a bit deeper, I have found my hypothesis is correct (quite significantly - in fact, its surprising my algorithm keeps up with the OT, see the below table for a comparison). Further discussion of this will be included in my technical report, and due to the amount of information, it can only be summarised in my iV2012 contribution.
 

			Levels
Graph	|V|	|E|	QT	MGF
3elt	4720	13722	14	14
uk	4824	6837	15	22
4elt	15606	45878	13	17
fe_sphere	16386	49152	14	15
finan512	74752	261120	14	36
wave	156317	1059331	15	89
dime20	224843	336024	15	52

Other changes, as requested, are being made to the submission but the limited page count is beginning to limit my additions (im convinced they set it just to play with me...).

For now, its back to "work". I shall update again when I have made an acceptable amount of progress, or when I get distracted from what I should be doing. Whichever comes first.