search

ulissigroup / GASpy_feedback

This is a GASpy submodule that is meant to work with GASpy_regressions to determine what GASpy simulations to run next.
GNU Lesser General Public License v3.0
0 stars 3 forks source link

Feedback on surfaces or materials, not sites #6

Closed ktran9891 closed 6 years ago

ktran9891 commented 7 years ago

We are currently feeding back optimal sites. But if these "optimal sites" lie on surfaces with a lower-energy, non-optimal sites, then it's kind of a moot point (i.e., there's less value in investigating that site).

We should develop a filter for our feedback loop that penalizes sites for having neighboring sites with lower, non-optimal adsorption energies.

And if we wanted to, we should be able to extend this to material sites, not just surfaces (if we wanted to simply look at heterogeneous nanoparticles instead of specific surfaces).

ktran9891 commented 6 years ago

Note that this commit makes us use a data ball of predictions. We can actually do this filtering with the data ball before it ever gets here.