The feature set is pretty similar to spearmint. We have found that the MOE GP package is much more robust than the code in spearmint though (it was open sourced by yelp and is used in production there).
In contrast to hyperopt, osprey is a little bit more geared towards ML, in that ideas about cross validation and datasets are baked in. It also doesn't require you setup your own mongo server. Both of those packages are great tools. -Robert On Sunday, November 2, 2014, federico vaggi <[email protected]> wrote: > Looks neat, but how does it differ from hyperopt or spearmint? > > On Fri, Oct 31, 2014 at 11:46 PM, Robert McGibbon <[email protected] > <javascript:_e(%7B%7D,'cvml','[email protected]');>> wrote: > >> Hey, >> >> I started working on a project for hyperparmeter optimization of sklearn >> models. >> The package is here: https://github.com/rmcgibbo/osprey. It's designed >> to be easy >> to run in parallel on clusters with minimal setup. For search strategies, >> it supports >> Gaussian process expected improvement using the MOE >> <https://github.com/yelp/moe> package, as well as >> random search and hyperopt's TPEs. The code is apache licensed. It's >> still pretty >> beta, but if anyone here is interested I'd encourage you to check it out >> and post >> any issues you have on github. >> >> -Robert >> >> >> ------------------------------------------------------------------------------ >> >> _______________________________________________ >> Scikit-learn-general mailing list >> [email protected] >> <javascript:_e(%7B%7D,'cvml','[email protected]');> >> https://lists.sourceforge.net/lists/listinfo/scikit-learn-general >> >> >
------------------------------------------------------------------------------
_______________________________________________ Scikit-learn-general mailing list [email protected] https://lists.sourceforge.net/lists/listinfo/scikit-learn-general
