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/
vinayakvivek has quit [Quit: Connection closed for inactivity]
mikeling has joined #mlpack
vivekp has quit [Ping timeout: 240 seconds]
AK has joined #mlpack
AK has quit [Client Quit]
AK has joined #mlpack
AK has quit [Ping timeout: 260 seconds]
trapz has joined #mlpack
trapz has quit [Quit: trapz]
trapz has joined #mlpack
sumedhghaisas has quit [Ping timeout: 240 seconds]
< zoq>
sumedhghaisas: Hello, just post your questions here or over the mailing list and I answer all of them as soon as I can.
trapz has quit [Quit: trapz]
vivekp has joined #mlpack
chvsp has joined #mlpack
chvsp has quit [Ping timeout: 260 seconds]
nax_ has joined #mlpack
vinayakvivek has joined #mlpack
govg_ has quit [Ping timeout: 240 seconds]
govg has joined #mlpack
govg has quit [Ping timeout: 264 seconds]
nish21 has joined #mlpack
< nish21>
rcurtin: regarding adaboost fix, we need to save weak_learner and alpha only if we terminate when rt >= 1, so i introduced a separate check after line 135. is that fine?
nish21 has quit [Quit: Page closed]
mikeling has quit [Quit: Connection closed for inactivity]
s1998 has joined #mlpack
vinayakvivek has quit [Quit: Connection closed for inactivity]
govg has joined #mlpack
nax_ has quit [Ping timeout: 260 seconds]
nax_ has joined #mlpack
< nax_>
zoq: http://coapp.org CoApp is better alternative to NuGet for C++
< cult->
hello, i have some questions: 1) is it possible to extract eigenvalues from Kernel PCA? 2) are there any algorithms in the library that would give me a so-called Robust PCA?
trapz has quit [Quit: trapz]
vss has quit [Ping timeout: 260 seconds]
< cult->
please disregard the first question, take only the second
trapz has joined #mlpack
trapz has quit [Client Quit]
vss has joined #mlpack
mikeling has joined #mlpack
trapz has joined #mlpack
trapz has quit [Quit: trapz]
trapz has joined #mlpack
< rcurtin>
nish21: yes, that's fine
< rcurtin>
cult-: no, there isn't any robust PCA implementation
< rcurtin>
the alternating minimization algorithm by Netrapalli and others (in NIPS 2014) would need to be implemented by hand if that is what you wanted to do
< cult->
rcurtin: thank you for this. can we treat kernel pca as robust anyway?
chvsp has joined #mlpack
< rcurtin>
cult-: if by 'robust' you mean 'tolerant to noise', then probably you can get ok results with KPCA
< rcurtin>
but if you mean the NIPS formulation where you have signal corrupted with large sparse noise, I am not sure how it will perform in that setting
< cult->
thanks, we will see
< rcurtin>
:)
nax_ has quit [Ping timeout: 260 seconds]
shikhar has quit [Quit: Page closed]
Trion has joined #mlpack
trapz has quit [Quit: trapz]
nax_ has joined #mlpack
< Trion>
On calling env.step(action), observation_space will update automatically?
< zoq>
Trion: yes
trapz has joined #mlpack
sagarbhathwar has joined #mlpack
< zoq>
rcurtin: Any idea what's wrong with jenkins job's some plugin error?
< rcurtin>
yeah, I saw that, I am going to try restarting jenkins, but I want the matrix build to finish first
< rcurtin>
it seems to have to do with the authentication being changed to github authentication
< rcurtin>
but it seems like the matrix build is hanging again... I'll just restart now
< zoq>
rcurtin: oh, okay
< zoq>
nax_: I'm not sure I get the benefit, is the process simpler?
< zoq>
nax_: Eager to hear what the appveyor guys have to say.
< nax_>
zoq: Me too, everything is nearly ready. Will need some testing though but as soon as appveyors reply we ll be ready to go!
< rcurtin>
poor appveyor, we must be running up a big heating bill for them :(
< rcurtin>
I think that appveyor has been building mlpack PRs and commits continuously probably for months now
< rcurtin>
wait, nevermind, that's not true at all now that I look at the build history...
< rcurtin>
looks like 10-12 builds a day, whereas continuously building would be more like 48/day
< rcurtin>
nax_: I just merged #953, thanks; if you want to add your name to the list of contributors in src/mlpack/core.hpp and COPYRIGHT.txt feel free to open another PR
chvsp has quit [Ping timeout: 260 seconds]
nish21 has joined #mlpack
travis-ci has joined #mlpack
< travis-ci>
mlpack/mlpack#2169 (master - f1a30c8 : Marcus Edel): The build was broken.
< sagarbhathwar>
Hello! I was looking at "Essential Deep Learning Modules" under GSoC ideas. For the proposal, should I pick only one of them(say Restricted Boltzmann Machine) and detail about it? Or multiple?
< zoq>
sagarbhathw: Hello there, depending on the model, one is probably not enough work for the summer.
< sagarbhathwar>
zoq: Is RBM + GAN a good combination?
< zoq>
sagarbhathw: Sounds reasonable. I think if you implement RBM, DBN is an easy extension that you could propose to do if there is time left.
< sagarbhathwar>
zoq: Thanks a lot for the info!
chvsp has joined #mlpack
< zoq>
sagarbhathw: You’re welcome, here to help.
< sagarbhathwar>
zoq: Which two of them would you suggest for optimal time utilization during summer?
< rcurtin>
sagarbhathwar: thanks for closing #941, I'd forgotten to do that
< sagarbhathwar>
rcurtin: No problem :)
< zoq>
sagarbhathw: That's hard to say, since I don't know your background; take a look at every model and choose the one you find the most interesting and manageable.
< sagarbhathwar>
zoq: I will look into it. Thanks
< nish21>
what's the link to mailing list archives?
< rcurtin>
Feodor is awesome, he is always quick to reply
< rcurtin>
when we didn't have enough processing power to build mlpack on AppVeyor, he simply just upgraded our free open-source account to an account that had higher limits, free of charge
< rcurtin>
I remember how difficult Windows testing was before AppVeyor, it is a really nice service
vss has quit [Quit: Page closed]
nax_ has quit [Ping timeout: 260 seconds]
< zoq>
rcurtin: yeah, really great support and service
< rcurtin>
zoq: I was listening to the first Moderat album last night, great album, thanks for the recommendation :)
trapz has quit [Ping timeout: 246 seconds]
< zoq>
rcurtin: I like it too, not every track but the overall album is really enjoyable.
chvsp_ has joined #mlpack
chvsp_ has quit [Client Quit]
< chvsp>
Hi zoq: Are there any other solutions which come to your mind for the BatchNorm error?
< zoq>
chvsp: I'll have to take closer look ... do that tomorrow I'll keep you updated
sagarbhathwar has left #mlpack []
Trion has joined #mlpack
Trion2 has joined #mlpack
< chvsp>
Sure. :)
< Trion2>
After reading paper on Synthetic Gradients even IRC upgraded me to Trion2
Trion has quit [Ping timeout: 264 seconds]
trapz has joined #mlpack
Trion2 has quit [Quit: Have to go, see ya!]
trapz has quit [Quit: trapz]
trapz has joined #mlpack
trapz has quit [Quit: trapz]
< cult->
how can I overwrite the constructor of KernelPCA to force the centerTransformedData to be true? ( is that actually centering the data or its just a flag? )
< cult->
nevermind it was easy :)
govg has quit [Ping timeout: 256 seconds]
trapz has joined #mlpack
govg has joined #mlpack
nish21 has quit [Ping timeout: 260 seconds]
kris2 has joined #mlpack
< cult->
whatever kernel im using for kernelPCA, i get error: arma::memory::acquire(): requested size is too large after the memory goes above 90%
< cult->
this is with 2.0.1
trapz has quit [Ping timeout: 258 seconds]
trapz has joined #mlpack
trapz has quit [Quit: trapz]
trapz has joined #mlpack
< chvsp>
quit
< chvsp>
Sorry, it was a mistake.
chvsp has quit [Quit: Page closed]
trapz has quit [Quit: trapz]
sumedhghaisas has quit [Ping timeout: 260 seconds]
Skywalker has joined #mlpack
< Skywalker>
Hey there! I would like to apply to the GSoC ... Does anyone know or has some advice about what should I write in the application form for mlpack?