Open codyevers opened 1 year ago
Let's make sure Rachel is part of this code revision. She has put a lot of thought into the weights and her expertise would be useful here.
And Alan- because this might create a difference in forsys versus ForSysX.
I think this will address the problem with weights more heavily searching some portions of the state space than others. Karen Abt brought this up a few years ago as well. I do think we should maintain the original weighting code as an option to maintain compatibility with ForSysX.
Great idea about keeping this as an option. The weighting issue becomes super clear when I tried to animate this stuff where you would barely see any movement as you changed weights, then would get they drastic shifts. I'll share some examples and early work here and we can continue to develop this idea as a team.
# DRAFT FUNCTION FOR PERMUTING WEIGHTS
func <- function(x, s=3) log10(x/(1-x))/s
# plot differences in S
num_seq <- seq(0,1,.01)
plot(func(num_seq, s=2), type='l')
points(func(num_seq, s=1.33), type='l')
points(func(num_seq, s=5), type='l')
Things to improve with the weighting function: