meeting 2024 02 02 n315 - JacobPilawa/TriaxSchwarzschild_wiki_5 GitHub Wiki
Context
A few plots instead of sending them in an email exchange with Emily.
Related to the getting the convex hull sampling correct!
Here's the result of running the convex hull samplign script for a variable number of points included in the convex hull fit. I noticed when using the cutoff of 500 we discussed in the meeting, that the resulting proprosed set of points was skewed heavily toward low Tmin, whereas this is disfavored pretty heavily by our chi2 landscapes. The same is true for the other shapes to a much more minor extent, where we end up dumping points *near* the minimum, but not really *centered* on the minimum.
K=75
K=100
K=150
K=200
K=250
K=500
K=750
K=1000
A quick follow up after our email exchange -- I re-ran the convex hull sampling, restricting the input points to those with Tmin>0.4. It looks like we still run into this issue where the highest density of proposed points is offset from where our best performing models currently are. I'm not sure if this is a catastrophic issue, but it definitely stands out to me as being....strange.
