Slides here
DLM: Slide5, Objective function compared to 1.3.0, can be different depending on the number of molecules.
CBy: What’s the difference in run1 & run2.
TG: Run1 was a bit buggy. But, the take away is that I am doing better in run2.
CBy: Slide8, from the table you exclude the r3-r3 bonds with smarts pattern and the description is saying to keep r3-r3 bonds?
TG: The generic parent parameter keeps whatever left. I am inverting the process of smarts pattern hierarchy.
CaB: So, it all depends on the training set, right?
TG: Yeah.
CB: The splits decrease the objective fn, and has less number of parameters than 1.3.0, so does this improve our FF?
TG: I will look into that.
CC: So the patterns are generated per step?
TG: Yeah.
CB: So, when you have the smarts pattern, it excludes things, but I can’t see the hierarchy from the table.
Why ring4 bonds before ring3? This is still a bit puzzling.
TG: Okay, I will get back with more details.
