Currently the RNG classes combine uniform and Gaussian random number generators, each with a separate buffer. We could split them, with a parameter controlling whether the RNG is uniform or Gaussian. This can be done once #26 is closed, because then we will be able to remove the separate Gaussian RNG currently used in the optimization classes.
Currently the RNG classes combine uniform and Gaussian random number generators, each with a separate buffer. We could split them, with a parameter controlling whether the RNG is uniform or Gaussian. This can be done once #26 is closed, because then we will be able to remove the separate Gaussian RNG currently used in the optimization classes.