search

j-faria / EPRVlnZ

MIT License
0 stars 0 forks source link

What should be our "computational effort" metric #3

Open j-faria opened 7 years ago

j-faria commented 7 years ago

The requirement is a integer roughly proportional to how much computational effort was used (e.g., number of iterations, number of model evaluations, CPU-seconds).

This is "effectively a label that allows one method to provide multiple estimates with more or less computational effort."

Maybe number of likelihood evaluations is the easiest metric

eggplantbren commented 7 years ago

Yeah, number of DNest iterations is probably best. This will equal the number of likelihood evaluations if you structure things in the standard way so that log_likelihood() computes the log likelihood from scratch.