ChanServ changed the topic of #mlpack to: "mlpack: a fast, flexible machine learning library :: We don't always respond instantly, but we will respond; please be patient :: Logs at http://www.mlpack.org/irc/
wiking has joined #mlpack
wiking has quit [Ping timeout: 276 seconds]
wiking has joined #mlpack
wiking has quit [Ping timeout: 240 seconds]
< rcurtin> I just had a really fascinating and exciting conversation with Sylvain Corlay from QuantStack about C++ and notebooks
< rcurtin> I didn't know this exists (maybe I live under a rock?) but there is a working C++ notebook environment via 'xeus':
< rcurtin> given that virtually 100% of data science is in notebooks, and that's native C++ in a notebook, I think it should not be hard to get mlpack running in a notebook!
< rcurtin> I'm hoping to adapt one of our examples later this week or maybe tomorrow if time permits
< rcurtin> but I think this could be a huge improvement to our documentation/examples, and could also make it *way* easier for users to *use* mlpack
< rcurtin> (since they wouldn't have to think about a compiler at all)
< rcurtin> don't know if anyone else here has ever played with xeus, but, this is the first I've seen it, and it's really really cool and I'm extremely excited
wiking has joined #mlpack
< rcurtin> Sylvain specifically told me that they are looking for users of xeus that they can demonstrate at conferences, etc., so I think that there is some exciting potential here
wiking has quit [Ping timeout: 265 seconds]
< zoq> I tested xeus cling, but run into som redefenition issues
< rcurtin> yeah just hit that, kind of unfortunate, but maybe fixable
wiking has joined #mlpack
wiking has quit [Ping timeout: 240 seconds]
wiking has joined #mlpack
< rcurtin> maybe some workaround can happen, I'll talk to Sylvain tomorrow
wiking has quit [Ping timeout: 265 seconds]
< rcurtin> that is kind of a blocker though... notebooks are a bit unusable if you can't screw up
< zoq> yeah, I used .undo a lot, but that wasn't really a solution; ended up restarting the kernel more often
< rcurtin> yeah, I see, that's not a great workaround though
< rcurtin> one of the commenters writes:
< rcurtin> I have written a similar interactive kernel for Fortran (https://docs.lfortran.org/) and there I simply shadow the previous definition, the example notebook has an example for that, you can also launch it in Binder to play with that interactively.
< rcurtin> but perhaps cling has just merged support for this:
< rcurtin> I'm not yet clear on what xeus provides that a regular cling notebook wouldn't, but I haven't looked too far yet
wiking has joined #mlpack
wiking has quit [Ping timeout: 265 seconds]
wiking has joined #mlpack
wiking has quit [Ping timeout: 240 seconds]
wiking has joined #mlpack
< zoq> Not sure there is a regular cling notebook, I think xeus-cling implements the Jupyter kernel and uses cling under the hood.
< zoq> I really like the idea of running mlpack in a notebook, it's a super nice way to show examples.
< jenkins-mlpack2> Yippee, build fixed!
< jenkins-mlpack2> Project docker ensmallen nightly build build #70: FIXED in 48 min: http://ci.mlpack.org/job/docker%20ensmallen%20nightly%20build/70/
wiking has quit [Remote host closed the connection]
wiking has joined #mlpack
wiking has quit [Remote host closed the connection]
< KimSangYeon-DGU> I just saw xeus-cling, it's so interesting.
< KimSangYeon-DGU> Wow, the mlpack logo on the poster is really nice :)
< jenkins-mlpack2> Project docker mlpack nightly build build #511: UNSTABLE in 2 hr 59 min: http://ci.mlpack.org/job/docker%20mlpack%20nightly%20build/511/
wiking has joined #mlpack
wiking has quit [Remote host closed the connection]
roberto_hueso has joined #mlpack
robertohueso has quit [Ping timeout: 265 seconds]
wiking has joined #mlpack
ImQ009 has joined #mlpack
wiking has quit [Remote host closed the connection]
wiking has joined #mlpack
wiking has quit [Remote host closed the connection]
wiking has joined #mlpack
wiking has quit [Ping timeout: 265 seconds]
wiking has joined #mlpack
wiking has quit [Ping timeout: 276 seconds]
wiking has joined #mlpack
jeffin143 has joined #mlpack
< jeffin143> Why don't we have kernel svm ???
< jeffin143> Or did I overlooked ???
wiking has quit [Remote host closed the connection]
< jeffin143> Wow , just went through the logs
< jeffin143> Xeus-cling is really great
< jeffin143> Will probably try that out
< rcurtin> jeffin143: nobody's implemented kernel SVM that's competitive with libsvm yet :)
wiking has joined #mlpack
< jeffin143> @rcurtin : is it specifically for Mlpack or as a general libsvm wins over kernelsvm
< rcurtin> jeffin143: I just mean that if we add kernel SVM, we should make sure it outperforms libsvm :)
< rcurtin> otherwise we can't claim that mlpack is "fast" :)
< jeffin143> That means over linear_svm is faster than liblinear ??
< rcurtin> I think it's definitely possible---it would just need some careful implementation
< rcurtin> jeffin143: right, I think there are some benchmarks somewhere in the PRs where we compared the two
< rcurtin> (or, what we compared was mlpack's `mlpack_linear_svm` vs. scikit's implementation, which is just a liblinear/libsvm wrapper)
< rcurtin> I seem to remember that our implementation was quite a bit faster, but I don't have the exact numbers
< jeffin143> And how do we know that our implementation of kernelsvm is slow , did someone implement it till now , and that didn't work out ?
< rcurtin> oh, we don't have an implementation of kernel svm at all
< rcurtin> I don't think anyone has tried to implement it (just because of lack of time or interest maybe?)
< rcurtin> but if someone does implement it, we just need to make sure it is competitive before merging it :)
< jeffin143> Ahaa ok :)
wiking has quit [Remote host closed the connection]
< jeffin143> On the way to learn rust , long term plan to come up with rust binding for Mlpack , until some one beat me :)
wiking has joined #mlpack
< rcurtin> jeffin143: nice, I've heard good things about rust :)
< jeffin143> Yes :) that's y most loved language according to stack overflow developer survey 🚀
< rcurtin> :)
wiking has quit [Remote host closed the connection]
jeffin143 has quit [Quit: AndroIRC - Android IRC Client ( http://www.androirc.com )]
wiking has joined #mlpack
wiking has quit [Remote host closed the connection]
travis-ci has joined #mlpack
< travis-ci> mlpack/mlpack#8256 (master - 84dadf0 : Ryan Curtin): The build has errored.
travis-ci has left #mlpack []
Nakul has joined #mlpack
< Nakul> Hi please have a look at https://github.com/mlpack/mlpack/pull/2075
< Nakul> I have sended the PR directly to the master branch is there any problem with that ?
wiking has joined #mlpack
< zoq> Nakul: Hello, that works.
< Nakul> how should i download libarmadilo , i am using sudo apt-get install libarmadilo and persistantly getting this error https://pastebin.com/41s6K1t6
< Nakul> previously i got this working but for some reason i uninstalled and then i got this error
< Nakul> after instally once again
wiking has quit [Remote host closed the connection]
wiking has joined #mlpack
< rcurtin> Nakul: can you tell us a little about how you configured mlpack?
< Nakul> previously i downloaded zip file , but now i git cloned fron the repo and maded build dir. and then run cmak ../ and after tha make
< Nakul> after running every methods got builded successfully but at the end i got that error
wiking has quit [Ping timeout: 268 seconds]
wiking has joined #mlpack
< rcurtin> Nakul: try removing the old armadillo that you downloaded, I think that could help
< Nakul> ok..
Nakul has quit [Ping timeout: 260 seconds]
jeffin143 has joined #mlpack
< jeffin143> Rcurtin : can you take a look at : https://github.com/pybind/pybind11
< jeffin143> Umm we could add it as optional dependency and then user may be able to pass Lambda function from python and then we can use that function in our c++ class
< jeffin143> To be specific as of now , a use case is the string binding where tokenizer could be a lambda function , for binding we have used one internal tokenizer but using this we could give user more power to customise it
< jeffin143> @lozhnikov : I will go through ngrams implementation and try to implement this week , most probably will start from Wednesday , since I haven't used that , and hence to read a lil bit
jeffin143 has quit [Remote host closed the connection]
Nakul has joined #mlpack
Nakul has quit [Ping timeout: 260 seconds]
jeffin143 has joined #mlpack
< jeffin143> Also check this out
< jeffin143> Linux source code 😂 🚀
MatrixBridge has joined #mlpack
MatrixBridge has left #mlpack []
< rcurtin> jeffin143: I see what you mean, I think it would be exceedingly difficult to recreate our python bindings to make that work, unfortunately
MatrixBridge has joined #mlpack
MatrixBridge has left #mlpack []
rcurtin[m] has joined #mlpack
< jeffin143> Ohh I see :) no issues
<@rcurtin> like I said, maybe it's possible, it would just be a lot of work and thinking
<@rcurtin> in the end it would be super useful though, so I agree it is a cool idea
< rcurtin[m]> test message from riot.im / #mlpack:matrix.org
<@rcurtin> rcurtin[m]: hey, it worked! good job
< jeffin143> Y two two rcurtin are there ??
< rcurtin[m]> yeah, I am testing the Matrix service... so this message is sent from a riot.im client in one tab :)
<@rcurtin> and this message is sent from my IRC client which is in another tab :)
< jeffin143> How did you manage to get that @ symbols?
<@rcurtin> Cadair from AstroPy/SunPy was nice enough to show this to me, so I am playing with it now to see how it works
<@rcurtin> ohh, oops, I had to op myself to accept the channel integration, but it's not needed anymore so let me deop myself
< rcurtin[m]> Here's a picture of the stickers table at the NumFOCUS summit (I'm seeing what happens when I send an image...)
< rcurtin> nice! that looks good from the IRC end too
GitterIntegratio has joined #mlpack
< jeffin143> We get to know many things through these summits :)
< jeffin143> It's awesome
< jeffin143> Came across many things :)
< jeffin143> Cpp con also :)
jeffin143 has quit [Quit: AndroIRC - Android IRC Client ( http://www.androirc.com )]
< rcurtin> definitely, yeah, the only thing is that I now need to find time to check out all the cool stuff I learned about :-O
ImQ009 has quit [Quit: Leaving]
wiking has quit [Remote host closed the connection]
test-webchat has joined #mlpack
< rcurtin[m]> <rcurtin "nice! that looks good from the I"> This is a response to a thread!
< rcurtin> this is an IRC response
jeffin143 has joined #mlpack
< jeffin143> Rcurtin : did we remove sparse svm ??
< jeffin143> If yes why ?
jeffin143 has quit [Client Quit]
test-webchat has quit [Remote host closed the connection]
< rcurtin> jeffin143: we realized that the sparse svm implementation was more accurately named linear svm :)
jeffin143 has joined #mlpack
< jeffin143> Hahaha
< jeffin143> Ok ok rcurtin :)
< jeffin143> Sparse svm has linear complexity or may be quadratic so yes linear_svm sounds good :-p
jeffin143 has quit [Quit: AndroIRC - Android IRC Client ( http://www.androirc.com )]
< rcurtin> :) yeah, in this case our linear_svm implementation is very good for sparse data (or it should be), but it's still sparse linear svm
wiking has joined #mlpack
wiking has quit [Ping timeout: 240 seconds]