01:50 < stephentu_> naywhayare: do the mlpack builds always fail?

03:15 < naywhayare> stephentu: oops, didn't realize you were still in the channel, so I guess I didn't need to respond to your message via email :)

03:15 < naywhayare> I'll get the matrix build fixed tomorrow...

03:15 < naywhayare> unless I discover the NFS server holding the armadillo versions ate its hard drive or something, which it probably did

03:20 < stephentu> naywhayare: its cool

03:20 < stephentu> naywhayare: fyi my goal is to get the predictor-corrector optimziation implemented tonight

03:20 < stephentu> for the IP method

03:20 < stephentu> then address the api issues

03:20 < stephentu> tomorrow

03:21 < stephentu> also its my irc fail i had two handles

03:21 < stephentu> one in the lab and one at home

03:21 < stephentu> i shoudl figure a better way to do irc

03:28 < naywhayare> :)

03:28 < naywhayare> I have an always-up host that I run GNU screen on and irssi inside of

03:28 < naywhayare> well, "mostly-always-up"

03:29 < naywhayare> the modified GaussianDistribution code looks great; thanks for taking the time to do that

03:29 < naywhayare> tomorrow (hopefully? maybe Thursday) I'll find some time to transition the other distributions to the same API

07:19 < stephentu> naywhayare: everytime i think i know c++

07:19 < stephentu> it turns out i really dont

19:32 < stephentu_> naywhayare: is there any support for sp_vec?

19:32 < stephentu_> or just use an sp_mat with one column

19:33 < stephentu_> seems like reading the doc it should just be the latter

19:42 < naywhayare> stephentu_: I thought sp_vec worked just fine

19:43 < naywhayare> should be typedeffed to be SpCol<double>

21:26 < stephentu_> naywhayare: ok, i just read the docs, so maybe the docs can be updated

21:27 < naywhayare> stephentu_: okay, I'll put that on my list

21:27 < naywhayare> I'm honestly not the biggest fan of the way Armadillo does documentation

21:27 < naywhayare> it's nice for users who have simple needs and are doing simple things

21:28 < naywhayare> but for anyone who actually wants to know what's going on under the hood and make sure that their code is being optimized the way they expect, one basically has to dive into the code

21:28 < naywhayare> and it gets pretty ugly pretty fast...

22:00 < stephentu_> naywhayare: ya i agree, i ahve a clone of armadillo sitting on my machine but my desire to look through it is very little

22:00 < stephentu_> although its cool how you guys manage to defer execution sometimes

22:00 < stephentu_> by using these glue types

22:00 < naywhayare> yeah, it's a really neat design

22:00 < naywhayare> in the Old Days he used to maintain Doxygen documentation too and make that available

22:00 < naywhayare> as well as the simpler documentation available now on the website

22:00 < naywhayare> but now if you want to know what's going on under the hood, there's not good documentation

22:01 < naywhayare> honestly I think Armadillo needs a full-time person devoted to maintenance and cleanup... the primary maintainer is also publishing papers and doing research, so there's not much motivation to clean up the code

22:01 < naywhayare> and I certainly don't have any time

22:04 < stephentu_> i wonder how eigen does it

22:04 < stephentu_> they have pretty good documentation

22:07 < naywhayare> it's the same type of approach

22:08 < naywhayare> I dislike Eigen because of its interface's complexity... it's all documented but the learning curve is quite steep

22:35 < stephentu_> naywhayare: how long do builds take for you

22:48 < naywhayare> depends on the system and what I'm building. usually a 'make -j4 mlpack_test' on my work desktop will take 3-4 minutes

22:49 < naywhayare> I think it would be nice to make the build faster, but I don't think it can be made instantaneous because it's C, not C++