Experimental parameters should all be stored in a struct that's saved with the results files.

[duvenaud]

Done. It would be nice to also specify the set of kernels being used in the experiment struct.