just draw randomly from a prior and report to a csv. This should be a straightforward extension of the grid-based sampling you already have, and would allow you to easily "increase the sampling size" without recomputing everything on the grid. It'll also remove any weird artifacts associated with precise grid spacing.
You should do this for both investigate-gpr*hyperparams scripts, which means you'll need to wrap the gpr-gpr stuff into a function call. But you were going to do that anyway for the emcee stuff, right?
just draw randomly from a prior and report to a csv. This should be a straightforward extension of the grid-based sampling you already have, and would allow you to easily "increase the sampling size" without recomputing everything on the grid. It'll also remove any weird artifacts associated with precise grid spacing.
You should do this for both investigate-gpr*hyperparams scripts, which means you'll need to wrap the gpr-gpr stuff into a function call. But you were going to do that anyway for the emcee stuff, right?