To Hell with it All!

Bad day, I guess.... but moving on to more important issues;

I/we had my monthly meeting last Thursday, which went well. We spoke about the results I have collected and the general consensus is that there should be something wrong with my implementation of the Quad Tree algorithm (should...). It was therefore decided that I need to check the code and rerun the tests in order to find any discrepencies. Another plan is to implement the scaling forces into the normal multilevel implementations (no approximation).

To start the week off well, I had left the tests running over the weekend. One of the quad tree retests failed to output altogether (due to the way Java stores output until theres enough to save, the file was left empty, suggesting an infinite loop or a buggy machine - which has happened before). This has been restarted with some more stable code (which worked for other graphs).

Continuing on with the dissapointment, the finan512 tests for CA still have not surpassed the size 100 tests (starts at 50, incremenets by 50 until 500, so very very slow).

Furthermore, the results I did manage to collect now show, wait for it... no difference.... The number of edge crossings against scalar value is linear (instead of quadratic like my own CA). I am still awaiting the results for 4elt before I can say for all cases but 3elt and 55grid show no change to the results collected before.

This leaves me with waiting for the results of the multilevel implementation with scalable forces, which previously hit a Java Heap Space error shortly after runtime.Another chance for redemption is to continue with the matching tests (compare matching with 4 vertices in CA against QT with matchings of 4 - the only real true test).

I guess its true to say if something can fail, it will all fail at once, but I'm sure there are some silver lining to these grey clouds somewhere in here. It just might be harder to see through the lightning and rain... heh. I will update again when I have more news. Have a fun week!