verne.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/
srirambandi has joined #mlpack
< srirambandi>
Hi!
< srirambandi>
How can i contribute
srirambandi has quit [Quit: Page closed]
srirambandi has joined #mlpack
srirambandi has quit [Client Quit]
srirambandi has joined #mlpack
srirambandi has quit [Client Quit]
daivik has joined #mlpack
manish7294 has quit [Remote host closed the connection]
manish7294 has joined #mlpack
manish7294 has quit [Remote host closed the connection]
< s1998>
rcurtin: is the project " alternative to neighbourhood -based collaborative filtering" available this year ?
< s1998>
I wanted to implement this : Collaborative deep learning for recommender systems , link : http://scholar.google.co.in/scholar_url?url=https://arxiv.org/pdf/1409.2944&hl=en&sa=X&scisig=AAGBfm2w0TNt53vw5cJFv1t9GcsV2_DCtQ&nossl=1&oi=scholarr&ved=0ahUKEwif-uP-oO7YAhVHtpQKHVPJD0YQgAMIIygAMAA
< s1998>
So CDL + RBFN would be okay for proposal ?
gaurav_ has joined #mlpack
s1998 has quit [Read error: Connection reset by peer]
vijay has quit [Ping timeout: 260 seconds]
s1998 has joined #mlpack
amrtanair has joined #mlpack
gaurav_ has quit [Ping timeout: 260 seconds]
amrtanair has quit [Ping timeout: 260 seconds]
s1998 has quit [Ping timeout: 248 seconds]
s1998 has joined #mlpack
s1998 has quit [Ping timeout: 240 seconds]
mayank1495 has joined #mlpack
< mayank1495>
I would like to start contributing to mlpack. I am really looking forward to GSOC 2018. How should I start with it ? I am good with algorithms and data structures but I dont have the knowledge of machine learning algorithms.
s1998 has joined #mlpack
ShikharJ has joined #mlpack
< rcurtin>
s1998: yes, the project is available still
< rcurtin>
but like every project it is not guaranteed that a student will be selected for it (it depends on the time available for mentors)
< rcurtin>
I think the paper you linked to is interesting, and could make a good project
< rcurtin>
I'd encourage you to think about how you can integrate the API of whatever you make with the existing CF code; it will be important that these pieces fit together well
< rcurtin>
some modification to the existing CF algorithms may be necessary
< s1998>
(rcurtin) I'd encourage you to think about how you can integrate the API of whatever you make with the existing CF code; it will be important that these pieces fit together well - sure
< s1998>
For rbfn - training strategy will be backpropagation, right ?
< rcurtin>
presumably---I am not familiar with the paper at this point, I have only glanced at it
< s1998>
Okay, thanks
< ShikharJ>
rc
< ShikharJ>
Sorry my bad!
< s1998>
One more thing, i saw in log you mesage regarding implementation of single linkage clustering. I had implemented them as part of hdbscan. Can i separate them and make a separate PR ?
< s1998>
*your
< mayank1495>
ok. Thanks.
< rcurtin>
s1998: I didn't realize that, sorry about that. the reason that PR has taken so long for me to get to is that it's hard to find time to read the paper and review it
< rcurtin>
I am hopeful to get to it this week, since it has been open so long
< rcurtin>
so we can handle it there, and if you did implement it in a way that it can be separated out, we can focus on getting HDBSCAN merged and getting both in the library
< zoq>
Didn't know you can use rdp on appveyor, quite nice.
< rcurtin>
yeah, I have gotten to the bottom of the issue
< rcurtin>
preparing a fix now
< rcurtin>
basically the %PATH% is not set right, but I can't seem to get it to work right in this session so I'll just copy the DLLs into the right directory, that will work ok...
< rcurtin>
I wish that there was a better error message so I did not need to RDP in, but this is actually nice to work with, it's a fully functioning Windows desktop
< rcurtin>
it runs the tests really fast---5 minutes total for all of them in release mode, which I think is a good bit faster than Travis
kaushik_ has quit [Quit: Connection closed for inactivity]
< zoq>
I guess the specs of the machine is slightly better, on the other side travis does run a lot more builds as appveyor.
< zoq>
Also if I remember right, they put us on a better machine.
< zoq>
Anyway, glad that you figured it out.
< rcurtin>
yeah, they upped the build time that they give us
< rcurtin>
and I think the system now has multiple cores
< rcurtin>
I am not done yet, let's see if this modification fixes it :)
< rcurtin>
and I'm still not totally sure how to extract the test results from AppVeyor or print them in a useful way
< zoq>
You can't just print the tesults as done with travis, maybe I missed something?
< rcurtin>
I'm not sure, let's see what ctest gives as output
< rcurtin>
but even if it does give that correctly, it's still like 15000 lines of MSVC build warnings before that...
< rcurtin>
I never load appveyor build results when I am on my phone because it will use up all my data :)