Its been a while...

Hope you had a good Easter, unfortunately it is now over and the work has begun again, although only for 3 days (for most people).

Just noticed I havnt published anything since 15th April, meaning I must be getting lazy about this, so from now on I will attempt to give updates more often to keep my progression relatively linear.

This week I am concentrating on my seminar/presentation, specifically making it so the presentation is not too complex but not too simplistic either (I don't want to insult anyone). I have also made a realisation that I have been getting lazy, and its now my intention to stop being lazy and be awesome hard-working instead.

Approximations
Implementation wise, the approximations are at a relative standstill, little work has been done and whats even worse, there is no reason as to why I havn't done it. I have spent much of my time investigating other techniques for approximation (especially within celestial body problems) and doing some minor testing to estimate the loss of accuracy is quicker models and whether its all worth it (as time is important here).

A method using the coarsening scheme as a structure for the approximations has been discussed with my supervisors, however, I feel this is too similar to the work done by Veldhuizen. However, this is faster in that it does not require the need for an additional structure to be calculated (the Barnes-Hut Tree), whereas the difference between the Barnes Hut method and the Grid method is predictable; the BH giving more accurate results at a cost and the grid being quick.

In addition to the comparisons between the two, and the coarsening approximation I have mentioned, I would also like to compare between them and a dynamic approach, where the accuracy of the approximation is inversely proportional to the convergance (much like the cooling schedule; the accuracy of the approximation increases as the graph nears its end). This would, in theory, result to a fast initial drawing, which gradually gets a better layout as time goes on (aesthetically, people prefer feedback on a task as soon as possible, regardless of the state of the task, so this initial feedback and layout is my target).

Drawing
Sidetracking to the more graphical side, I have decided that for graphs over a certain size, the spheres representing vertices will not be drawn. This will speed up the drawing part of the scheme and in some cases, will make the graphs more pleasing.

For the selection and movement of vertices in these graphs, its likely the user will zoom in on an area to select the nodes, so when this occurs (based on the cameras position from the vertex), the spheres may be shown to allow for such selection, still not rendering those vertices too far away.

Libraries, Latency and Code Optimisation
Previously I have mentioned the latency of some libraries may take too long, a helpful comment from my supervisor gave me more accurate testing material which showed the latency is negligible in some cases. As this is a task which can take forever, all forms of code optimisation and concerns about it are to be ignored until a final implementation is created.

Literature Review
Several months passed my original set deadline and this is still going as a part time mini-task. This literature survey/review, which will hopefully be the base of my final literature review, continues to be edited with more and more information appended to it. Recently I have attempted to finish the review, and started to check it over ensuring I didnt miss anything, but I keep making small additions and removals.

Experimentation
I have found myself looking over much of the work I have undertaken previously in order to begin writing up all the little and large tests I have done with the work in order to have them available should I need them.

Future
Over these last few months, I have been getting much lazier than I should be, leaving tasks half complete or taking too long to do implement simple changes or tests. This post/blog entry represents an end point to that period, with my focus now on completing my research to as high a quality as I can, as soon as possible (much like when I started).

Already 8 months in, its time to get a move on methinks.

Errors: A comeback special

As with everything I do/attempt, I have encountered a great deal of problems with my implementation of Barnes Hut and Grid approximations when it comes to splitting them into their groups (theory is correct as far as can be told, just some issues when putting vertices in their correct group after a move). Other problems also exist in my method of using coarsening as an approximation (forces need to be scaled and further experimentation needed to identify to which extent).

Other than these few set backs, I am looking to have a simple implementation of different approximation methods by next week. On top of this, I have also been gathering notes for my seminar (nervous much?) which I will be putting together next week as well.

There is not much left to mention as of yet, I will detail my findings when my implementation of the approximation techniques are complete. For now, have a good weekend!

Approximate Universe?

This week I will be exploring the wonder that is the Barnes Hut approximation.

When I say exploring, I mean implementing, and when I say wonder, I mean hellish firestorm from which nothing can survive (oh ok, its not that bad). I will be implementing this, the FR grid variant implemented by Walshaw in his multilevel graph drawing algorithm, and also my own adaptation using the coarsening scheme as a basing structure, in order to compare the effectiveness and speed of each approach (also, if time allows, I may look into other approximation methods if anything is quicker).

As always, it comes down to speed, which in turn comes down to how quickly a graph converges, which again is speed, measured by how quickly a graph converges, known as spee.. STOP (hammertime).

The approximations have been proven to speed up the n-body problem, but other qualities such as how the approximation affects the layout is scarcely covered. In order to prove my own adaptation is quicker, I will need the BH tree to compare against.

Other than this, there is no other developments to declare in this post, so all I can say is so long, and thanks for all the fish.

Marking: The Nightmare

It is true what they say, marking is a repetitive boring and nightmarish task that will eventually get everyone. Unfortunately it is a necessity for tutors, making it harder to balance work and research (which others would incorrectly identify as the work vs. life balance).

For the past few days I have unfortunately, been trapped by the monster than is marking, and therefore my graph work is at a complete standstill.

I have also yet to write up the minutes of the previous meeting, which will be uploaded in my next post.

After this hellish nightmare of doom and darkness has finished, I will return to the realm of beautiful fluffy clouds, joyful sunshine and perfectly golden fields of glory and happiness they call research, and take revenge on the evil below - by ignoring it until the end of Easter.

All the best!