02:54 < saiprasanth> hello everyone. I wish to start contributing to mlpack. Please help me get started

16:39 < rcurtin> zoq: do you think it would work to do something like this?

16:40 < rcurtin> template<typename eT> class RunningStatWrapper : public arma::running_stat<eT> {

16:40 < rcurtin> // add a constructor or two

16:40 < rcurtin> // serialize the object

16:40 < rcurtin> template<typename Archive> void serialize(Archive& ar, const unsigned int version);

16:40 < rcurtin> };

16:41 < rcurtin> I haven't tested that, it's just an idea that popped into my head. If we replaced the use of running_stat with RunningStatWrapper in the codebase, I think we could "hack" in serialization like that

16:44 < zoq> I think that should work, not sure it's necessary here; computing the variance/mean is really simple. Let me open another PR with the current solution I use for testing.

16:48 < zoq> https://github.com/mlpack/mlpack/pull/1508/files#diff-5d66d090a2742cdc6da01a47364dcf23R91-94

16:49 < zoq> don't mind to switch back to the running_stats method

16:56 < rcurtin> I think that could work just fine too, since you already have the code working

16:57 < rcurtin> even if we did submit something upstream, it would take some time before it was in the next release, and it wouldn't be easy to backport

16:57 < rcurtin> so it would probably be some years before we were able to fully switch back to running_stat...

17:01 < zoq> Right, we would have to hold a copy of the whole class. If somebody prefers the running_stats class or if we see a need in the future I guess it makes sense to think about the wrapper class; nonetheless I guess it makes sense to submit a patch to Conrad.

18:47 < rcurtin> it looks like you have it all solved in #1508, so I am not sure it's worth it to take the extra effort to find a different solution :)