The complication is that extrapolate_to_env_present.R currently batches over new input rows, and each input row generates a similarity with every site from the original gfbootstrap environment.
predict_gfbootstrap would need additional logic to batch over the row pairs.
For example:
sites 1:10 run against sites 1:10, 11:20, 21:30...
sites 11:20 run against 11:20, 21:30, 31:40 ....
extrapolate_to_env_present.R could benefit from this logic too, but extra changes would be needed to find the clusters most similar to each new site.
The complication is that extrapolate_to_env_present.R currently batches over new input rows, and each input row generates a similarity with every site from the original gfbootstrap environment.
predict_gfbootstrap would need additional logic to batch over the row pairs.
For example:
sites 1:10 run against sites 1:10, 11:20, 21:30... sites 11:20 run against 11:20, 21:30, 31:40 ....
extrapolate_to_env_present.R could benefit from this logic too, but extra changes would be needed to find the clusters most similar to each new site.