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/
trapz has joined #mlpack
trapz has quit [Quit: trapz]
wasiqm has joined #mlpack
chenzhe has quit [Ping timeout: 256 seconds]
trapz has joined #mlpack
mikeling has joined #mlpack
trapz has quit [Quit: trapz]
trapz has joined #mlpack
trapz has quit [Quit: trapz]
shivakrishna9 has joined #mlpack
AndroUser has joined #mlpack
< AndroUser> @zoq thanks for merging my pull request (#27). I am interested in the better benchmarking project listed on project ideas for gsoc 2017. How to proceed now.
vinayakvivek has joined #mlpack
ironstark has joined #mlpack
AndroUser has quit [Remote host closed the connection]
govg_ has joined #mlpack
govg has quit [Ping timeout: 240 seconds]
vivekp has quit [Ping timeout: 260 seconds]
vivekp has joined #mlpack
Nax_ has joined #mlpack
ironstark has quit [Ping timeout: 246 seconds]
HoloIRCUser has joined #mlpack
HoloIRCUser has quit [Client Quit]
HoloIRCUser has joined #mlpack
HoloIRCUser is now known as ironstark
slayerjain has joined #mlpack
chenzhe has joined #mlpack
ironstark has quit [Ping timeout: 246 seconds]
HoloIRCUser has joined #mlpack
HoloIRCUser has quit [Ping timeout: 246 seconds]
HoloIRCUser has joined #mlpack
HoloIRCUser has quit [Ping timeout: 246 seconds]
Nax_ has quit [Ping timeout: 260 seconds]
HoloIRCUser has joined #mlpack
HoloIRCUser has quit [Client Quit]
mikeling has quit []
mikeling has joined #mlpack
vivekp has quit [Ping timeout: 264 seconds]
vivekp has joined #mlpack
Nax_ has joined #mlpack
slayerjain has left #mlpack []
vss has joined #mlpack
vinayakvivek has quit [Quit: Connection closed for inactivity]
chenzhe has quit [Ping timeout: 256 seconds]
trapz has joined #mlpack
trapz has quit [Client Quit]
trapz has joined #mlpack
trapz has quit [Client Quit]
jarvis_ has joined #mlpack
< jarvis_> Hi mentors, I have submitted a draft proposal based on our earlier discussions to implement 3 cnn architectures. Let me know your views!
jarvis_ has quit [Client Quit]
trapz has joined #mlpack
vinayakvivek has joined #mlpack
s1998 has joined #mlpack
vss has quit [Quit: Page closed]
Trion has joined #mlpack
sumedhghaisas has joined #mlpack
thyrix has joined #mlpack
Trion has quit [Remote host closed the connection]
Trion has joined #mlpack
thyrix has quit [Ping timeout: 260 seconds]
vss has joined #mlpack
< Nax_> zoq: https://github.com/Microsoft/vcpkg is in preview. That's why it is not included in visual studio yet.
< zoq> Nax_: I like the idea, so all we have to do is to create a port file for armadillo, and mlpack right since there is already an port for boost
jenkins-mlpack has quit []
jenkins-mlpack has joined #mlpack
jenkins-mlpack has quit [Client Quit]
jenkins-mlpack has joined #mlpack
jenkins-mlpack has quit [Client Quit]
< zoq> hm, is masterblaster moving?
jenkins-mlpack has joined #mlpack
jenkins-mlpack has quit []
jenkins-mlpack has joined #mlpack
< rcurtin> zoq: nope, I was working with the authentication
< rcurtin> now you should be able to log into masterblaster via the github account
< rcurtin> can you test and make sure that works for you also?
< zoq> it works
< rcurtin> great
< rcurtin> this will make it easier to give, e.g., GSoC students access to Jenkins without needing to make individual accounts on masterblaster
< zoq> so any org member can access jenkins right?
< rcurtin> yes, should be
< zoq> neat
< rcurtin> the list of administrators can be configured on the 'configure global security' page
< Nax_> zoq: Exactly!
nish21 has joined #mlpack
trapz has quit [Quit: trapz]
< rcurtin> zoq: I have matlab + toolboxes installed on dealgood, next I will work on getting that running on the benchmarking systems
< zoq> rcurtin: sounds great :)
s1998 has quit [Quit: Page closed]
kris1 has joined #mlpack
Trion has quit [Quit: Have to go, see ya!]
< kris1> zoq: Could you have a look at GaussianInit PR.
< zoq> kris1: Just looked over the PR can you adresse the comments?
chenzhe has joined #mlpack
chenzhe has quit [Ping timeout: 240 seconds]
chenzhe has joined #mlpack
nish21 has quit [Ping timeout: 260 seconds]
nish21 has joined #mlpack
< nish21> hello! could someone explain me how serialize works?
vss has quit [Quit: Page closed]
< nish21> when i try to open a model.txt file, is it the output of serialize call?
mikeling has quit [Quit: Connection closed for inactivity]
trapz has joined #mlpack
chenzhe has quit [Quit: chenzhe]
chenzhe has joined #mlpack
< rcurtin> nish21: yes, that is using boost::serialization
< rcurtin> you can serialize as .bin, .txt, or .xml
< rcurtin> xml is the easiest to understand but the files are very large
< rcurtin> since each value in the matrix gets its own tag...
nish21 has quit [Ping timeout: 260 seconds]
nish21 has joined #mlpack
< nish21> i like how easy to understand the xml serialization is :)
Nax_ has left #mlpack []
chvsp has joined #mlpack
< chvsp> Hi zoq: I have tried typecasting, it didn't work. I also creates another variable of size_t and passed it, still I get the same error in AppVeyor.
wasiqm is now known as DDOOOqq
nish21 has quit [Ping timeout: 260 seconds]
nish21 has joined #mlpack
< chvsp> zoq: Reference to BatchNorm layer PR error. PR# 955
chenzhe1 has joined #mlpack
chenzhe has quit [Read error: Connection reset by peer]
chenzhe1 is now known as chenzhe
< zoq> chvsp: Strange, can you remove the "default" Constructor BatchNorm() and try again?
< chvsp> zoq: Ok will try.
chenzhe has quit [Quit: chenzhe]
nish21 has quit [Ping timeout: 260 seconds]
nish21 has joined #mlpack
< nish21> rcurtin: this is regarding adaboost error fix, i have looked at the discussed issue. we need to push w as well as alphat before terminating for the model to classify, does it make sense to push alpha caclulated using DBL_MAX into the model and then terminate?
< nish21> let me describe the problem more clearly.
< rcurtin> yes, I think I understand what you mean, let me think about it
< nish21> alright
< rcurtin> in Classify(), you need two things: the vector of weak learners wl[], and the vector of alpha values alpha[]
< rcurtin> in the situation where the weak learner perfectly fits the data on the first iteration, you simply need that single weak learner in wl[] and then the value '1' for alpha[]
< rcurtin> since alpha[] is the weight applied to each weak learner, if you have only one learner and it got everything right, then its weight should be 1
< rcurtin> actually, technically, its weight does not matter at all since the vector of probabilities for each point across each weak learner (cMatrix) is then normalized
< rcurtin> thus 1 is a reasonable value to use
< nish21> rcurtin: yes, that works. so push 1 to alpha and terminate
< chvsp> zoq: No it also doesn't work. It is giving the same error.
< rcurtin> nish21: yeah, this should be a reasonable solution
< nish21> rcurtin: i just confirmed locally that it works. i'll push the changes soon.
DDOOOqq has quit [Quit: Leaving]
< rcurtin> nish21: great, thanks
nish21 has quit [Ping timeout: 260 seconds]
chvsp has quit [Quit: Page closed]
trapz has quit [Quit: trapz]
nish21 has joined #mlpack
< nish21> rcurtin: actually, after that change, one test, WeakLearnerErrorIris_DS fails
joshua__ has joined #mlpack
< joshua__> hello there
< zoq> joshua__: Hello!
< joshua__> i just wanted to know what will be the type of projects be done through gsoc 2017 in mlpack c++
jenkins-mlpack has quit [Remote host closed the connection]
< zoq> joshua__: Not sure what you mean with type of project; have you seen the ideas page? https://github.com/mlpack/mlpack/wiki/SummerOfCodeIdeas
< kris1> zoq: I have updated the changes in GaussianInit. Please let me know if there are some other changes. I have some time right now. Thanks
< joshua__> @zoc will be checking on that .Thank you for the prompt reply
< kris1> Also for my proposal i am thinking about ssRBM, RBM,BRNN and GAN, Stack GAN.
< zoq> kris1: Looks good, just some minor style issues I'll fix after the merge; I have to step out for an hour or so, once I get back I merge it in.
joshua__ has quit [Quit: Page closed]
< kris1> Thanks zoq. I will also get some sleep. I will finish the xavier init tomorrow most probably.
< zoq> kris1: hm, maybe you focus on some of the models and propose to implement another if there is time left?
< zoq> kris1: I think implementing all of them is probably to much work.
< kris1> okay i will take that into account.
< zoq> kris1: okay, sounds good
< zoq> kris1: Don't underestimate the time to write some really good tests.
nish21 has quit [Quit: Page closed]
< kris1> Yes i was thinking about those. Maybe i will discuss about that with you tomorrow. If you have some time.
kris1 has left #mlpack []
< rcurtin> ok, great! slake.mlpack.org now has matlab + toolboxes
< rcurtin> let me deploy that to the other four systems...
trapz has joined #mlpack
< rcurtin> zoq: I am working with the reuters dataset from the benchmark-data repo, but I can't seem to find the labels for the training set
< rcurtin> there is reuters_train.csv, reuters_test.csv, and reuters_labels.csv (the labels for the test set); do you have the training labels anywhere?
< zoq> rcurtin: hm, it's not the last row of the training set?
< rcurtin> oh!
< rcurtin> yep, exactly, sorry about that
trapz has quit [Quit: trapz]
trapz has joined #mlpack
< zoq> If I clone the github repo and configure a job that uses the cmake plugin (choose no build type) it builds with "-g"?
< rcurtin> where do you see that it built with -g
< rcurtin> ?
< rcurtin> it seems to me like it didn't build with -g because the valgrind stacktrace output doesn't give line numbers, i.e.
< rcurtin> "source code not available"
trapz has quit [Quit: trapz]
< zoq> I thought DEBUG=ON is the default setting for the master branch so set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} -g -O0 -ftemplate-backtrace-limit=0")
< zoq> ah, it's not
< rcurtin> this was changed sometime back to default to DEBUG=OFF after someone wrote a paper comparing against mlpack compiled with -g -O0 because they didn't know better :)
< rcurtin> I can't remember the PR, maybe I can find it...
< zoq> ah, right
< zoq> thanks!
< zoq> I've modified the mlpack_test executable to output all tests to parallelize the memory check: https://github.com/zoq/mlpack/commit/1eb20dcd3ac9f450dfbbffd7128cf9bf369dfc53 not sure that's something interesting for the master branch.
travis-ci has joined #mlpack
< travis-ci> mlpack/mlpack#2163 (master - 2dac408 : Ryan Curtin): The build passed.
travis-ci has left #mlpack []
< rcurtin> zoq: I think it's fine to pull that into the master branch
< rcurtin> it's test code anyway, so anything that makes the tests more useful I think should be in master
< zoq> rcurtin: okay, maybe someone else finds it useful, all it does is to print out the test cases: http://masterblaster.mlpack.org/job/pull-requests%20mlpack%20memory/ws/testbins.txt
< zoq> I thought it's a simple way to parallelize the work, especially if you have a multicore CPU.
< zoq> and you are on linux and can use xargs :)
< rcurtin> hmm, it looks like you can also do bin/mlpack_test --list_content and some text processing to get the same result perhaps
< rcurtin> but I think the text processing will be irritating to do
< zoq> bin/mlpack_test --list_content runs the tests on my system
< rcurtin> oh, really? it just lists them on mine; I guess it must have changed between boost versions or something
< rcurtin> yeah, it seems 1.62 lists them correctly but 1.58 does not
< rcurtin> in that case it is probably best to just provide the support
< zoq> I agree
< sumedhghaisas> zoq: Hey Marcus. I am just going through the ANN code to write my proposal. I had some quick questions if you have some time.
trapz has joined #mlpack
trapz has quit [Quit: trapz]