Where's the finish line?

Once again falling behind my own crazy schedules, this time out of my past laziness.

I have spent far too long, over a month, on the approximation work due to my use of time, and as a result, I am falling further and further behind as I did not leave enough room to fix the problems I/we/the universe knew I was going to have. To add to the time dilemma, its also Iivigilation time and as it happens, exams love me so I have a lot of invigilation.

I am now fully concentrating on the approximations completeness and am changing my deadline from end of the week to weekend/early next week. Engines are on go and, oh damn, making this blog post is using that time up....

Looks like I will be updating as soon as I can, in the mean time, here's a picture of the grid approximation (the squares edges are based on the coordinate boundaries within each square and not just a load of lines I've put on top of the image). Barnes Hut is being a little more difficult.

As I havnt shown any pictures for a while, here are the outputs from my approximations at the moment; the grid variant shows  boxes on top of the graph (these are drawn based on the boundaries of each box, not just lines scattered wherever). Each box contains a list of vertices within it, boundaries for the coordinates of those vertices (if they go over they move box) and a centre of mass.

The image below shows my bugged up version of the Barnes Hut approximation. The boundaries on lower levels are wrong which suggests the problem is when the boundaries of the parent node are passed to the child node to make their own boundaries, but its not clear in my code.

NB: the size of the grid can be changed but requires some experimentation to find the ideal parameters (as mentioned by Fruchterman and Reingold).

<EDIT> Figures I would find the cause after I post the problem; this shift to the left is due to the lower levels having boundaries adjacent to the lower x boundary of the entire grid. This had been overlooked as I had assumed the problem was due to the values I had been passing as the boundaries, when instead it was ignoring these inputs and assigning incorrect values. Bad that I didnt notice this before but now it works so I'm happy.

Few more changes to my coarsening approximation scheme and it will be time to get the results.

Below is a new pic of  the Barnes Hut working in 2D (top cut off by accident due to the size of the viewing window so ignore the missing boundaries please).</EDIT>