The perturber function can/should be different depending on the model you're working on. So in the current state where there's just one example, it should be moved back in with SpikeSlab, and passed as an argument to stuff, like the log likelihood function. Also, randh is useful for perturb functions in general so maybe put that in Utils.
The perturber function can/should be different depending on the model you're working on. So in the current state where there's just one example, it should be moved back in with SpikeSlab, and passed as an argument to stuff, like the log likelihood function. Also, randh is useful for perturb functions in general so maybe put that in Utils.