Open lkct opened 1 year ago
Another idea to extend the "random state": we accept anything that supports rand
, randn
, etc. as a general RNG. However, the implementation can also be a pre-defined deterministic generator -- enabling us to verify all the results by hand with some known "random" sequence. This is useful for unit tests, e.g., on sampling.
shall we close this one?
Ref: discussion in the comments under #58.
We shall make a decision when adding a new argument will not cause much trouble.