As mentioned in issue 14, we need to add more search algorithms, and each one will have its own issue to track progress. This issue will track progress implementing slice-sample MCMC with Zeus.
The Zeus paper suggests that the code scales much better to higher-dimensional problems (10 < N < 100) than emcee does, and it handles complex posteriors (e.g. multimodal or moderately non-Gaussian) much more effectively thanks to using slice sampling. They also include a nice interface for changing the type of move, which make it possible for walkers to more easily shift between different modes. First step will be to implement a simple drop-in replacement for the emcee class, and then work to unlock the more powerful features of Zeus.
As mentioned in issue 14, we need to add more search algorithms, and each one will have its own issue to track progress. This issue will track progress implementing slice-sample MCMC with Zeus.
The Zeus paper suggests that the code scales much better to higher-dimensional problems (10 < N < 100) than emcee does, and it handles complex posteriors (e.g. multimodal or moderately non-Gaussian) much more effectively thanks to using slice sampling. They also include a nice interface for changing the type of move, which make it possible for walkers to more easily shift between different modes. First step will be to implement a simple drop-in replacement for the emcee class, and then work to unlock the more powerful features of Zeus.