cameron.freenode.net changed the topic of #mlpack to: http://www.mlpack.org/ -- We don't respond instantly... but we will respond. Give it a few minutes. Or hours. -- Channel logs: http://www.mlpack.org/irc/
< stephentu>
naywhayare: have you had a chance to look @ the matrix completion API
< naywhayare>
I've been plodding through the paper today in order to understand the problem a bit better
< naywhayare>
the recht/fazel/parrilo paper
< naywhayare>
I'm not spending the time to verify the proofs or anything, just understand the nuances of the problem and algorithm a bit better
< naywhayare>
it seems like maybe we can drop this in place into the collaborative filtering model in order to do LRMA? I'll have to look at the code and think about it a bit more
< naywhayare>
but I am putting time into this today, so you can expect a response shortly :)
< stephentu>
yay
< stephentu>
what is LRMA
< naywhayare>
low-rank matrix approximation
< naywhayare>
perhaps that isn't the right term to use
< stephentu>
ah i see
< stephentu>
ya the nuclear norm minimization stuff is pretty general as long as you have a linear constraint
< stephentu>
i pulled out matrix completion since its an interesting special case
< stephentu>
anyways i'm working on an interior point solver now... might be fun to have for high accuracy solutions :)
< stephentu>
i'm honestly just curious how much a straightforward implementation differs from these closed source commerical packages
< stephentu>
in terms of performance
< naywhayare>
are you able to compare them?
< naywhayare>
maybe your lab has a copy or something?
< stephentu>
oh ya
< stephentu>
i have a copy of mosek
< naywhayare>
ah, that's right, that's what you've been comparing with
udit_s has quit [Ping timeout: 246 seconds]
shreya_ has joined #mlpack
shreya_ has quit [Client Quit]
shreya has joined #mlpack
shreya has quit [Ping timeout: 244 seconds]
< naywhayare>
stephentu: any of the comments I make are discussion points and not necessarily "you must do this", so feel free to discuss if you have differing opinions on anything I write :)
sumedh has quit [Ping timeout: 244 seconds]
govg has quit [Ping timeout: 256 seconds]
< stephentu>
naywhayare: ya i wrote a few comments back
< stephentu>
thansk for the comments though, i agree with most of it and i like the idea of putting it into CF
< stephentu>
but i'll have to study taht code a little bit to see if it makes sense
< stephentu>
and yes, being able to drop in an SDP solver is my goal before the semester starts
< stephentu>
so hopefully we'll have that in a few weeks
< naywhayare>
wow, your semester hasn't started yet?
< stephentu>
stephentu: nope it starts late
< naywhayare>
ah, nice
< stephentu>
around MLK day
< naywhayare>
the semester started here this week, but semesters don't really matter to me anymore since I'm done with classes...
< naywhayare>
just research all the time...
< stephentu>
ya i'm not quite there yet, trying to take 3 this semester
< naywhayare>
my first semester of graduate school I thought it would be a good idea to take 15 hours (5 classes)
< stephentu>
ya no
< naywhayare>
I learned that it was indeed NOT a good idea
< stephentu>
lol
< stephentu>
2-3 is a reasonable # i think
< naywhayare>
yeah, I would agree with that
< stephentu>
theres just too much maths to learn
< naywhayare>
it's true
< naywhayare>
anyway, I have to move on to other things for now
< naywhayare>
whenever you're happy enough with the changes I'll go ahead and pull them in
< naywhayare>
it can always be changed later
< naywhayare>
(or, at least, until a release, after which point we should think about maintaining a stable API :))