Random shower thought: when doing importance sampling for Monte-Carlo integration, we can approximate the resulting variance based on the variance of the relative error between our probability distribution and the ideal one!
Probably deserves a small blog post...
@lisyarus that would be great (because don't exactly understand what is happening from the few lines), definitely sounds interesting :)
@lisyarus sounds a bit like re-stir, but I don’t actually remember how that one works.