The effect of Hierarchy and Overlap on routing
Following on from the previous post, I implemented the algorithms that evaluate the structure of the hierarchy, and plotted this for all datasets against the main routing metrics.
However, this showed a confusing picture, so Pádraig suggested I run multiple parameters to HGCE over one dataset, and see if it made a difference. We chose the Studivz-three-B dataset, as it is large enough to give a nice general picture.
The parameters were as follows:
K=3,4
M (threshold) = 0,1e-05,2e-05,3e-05,4e-05,6e-05,6e-05,7e-05,8e-05,9e-05,6.83085891329e-05,2.04848750968e-05,8.6207182699e-05,5.97475523656e-06
the other parameters were kept at their default (see here)
Again, on first look, this seemed confusing, but by removing the two very low delivery ratio results (leaving 24 out of 26), the visualizations seemed to show a trend towards increased hierarchical structure meaning increased delivery ratio.
It is quite interesting to explore the data by filtering out different runs, e.g removing all values where K=3, or 4 makes a difference: Hierarchy Data Studivz-B
I suspect the two clumps in delivery ratio form either side of the point an important Threshold parameter.
Alright y’all, found lago777comloginregister.com. Looks like it’s a place to log in and register for something. If you’re looking for that, lago777comloginregister might be what you need.
Checked out 78brlbet the other day. It’s okay. Nothing special, but it does the job. Interface is a little clunky,but functional. Might be worth checking if you are out of other choices anyway: 78brlbet