Using T1 Transitions instead of reconnect?

[sniffo]

I am runnning some simulations where I would like the merge-split based reconnection system implemented by default in the euler solver with a simple T1 transition - since the settings I'm using causes rosettes to form a lot faster than they can be resolved. I would also like to analyze the difference between rosette based transitions and simple T1 transitions for a set of simulations I have already performed and analyzed. Is there an easy what to do this?

[glyg]

Hi @sniffo

Three parameters govern reorganization:
The threshold length (sheet.settings["threshold_length"]), and the reconnect frequencies sheet.settings["p_4"] and sheet.settings["p_5"]

If you set p_4 to 1, this means that the 'split' from a range 4 vertex will be instantaneous, corresponding to a type 1 transition

If you have time, could you test the dev branch here? I merged you PR and @sophietheis 's and changed the packaging?

[sniffo]

Hi Guillaume,

Thanks for the suggestion. But is it not the case that the split has a 50% chance of it reforming the original edge that collapsed to form the rosette in the first place? This would still be a significant improvement regardless.

Yes, I would be happy to help with the test later this week.

[glyg]

Thanks for the suggestion. But is it not the case that the split has a 50% chance of it reforming the original edge that collapsed to form the rosette in the first place? This would still be a significant improvement regardless.

Yes you're right, I hadn't thought about that. At the very least it should prevent rosettes. The less easy alternative is to change the reconnect behavior that is called by the solver (so sub-classing EulerSolver to run the T1s

Yes, I would be happy to help with the test later this week.

Great!
You need to :

git fetch
git switch dev
pip install .[dev] # editable install is broken don't use -e
make tests

And then try your code :)

Best

Guillaume