qpipattern / phase-of-matter-by-machine-learning Goto Github PK

Hi

Some paper told that when T<Tc, there are vortices pair, and when T>Tc, there is no vortices pair.
I thought this is one method to distinguish low T phase, or high T phase, and this statement is what I expected CNN can do well.

I run the program from this website, after few updates, I get the same result from initialization state is "hot"(means spin are all start from random, just like your code).

But when I keep update spin, vortices pairs are annihilated below Tc, there is no vortex pair - free vortex transition at Tc, It seen both of these two spin configuration (higher and lower around Tc) are similar(in human eyes), this result may lead big error in CNN algorithm.

If you feed the spin configuration below Tc which have vortex pair(In traditional method, it has not achieve equilibrium, the observable like m or other should not be put into average), the machine learning result well be better then feed the spin configuration which achieve equilibrium?

(Sorry for that I'm not good at English)

Dear QPIpattern, your work is very interesting: do you have by chance any paper/preprint describing your experiments/techniques?

Thanks a lot!
Giuseppe Jurman