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/
rcurtin_ has joined #mlpack
ChanServ has joined #mlpack
ChanServ has left #mlpack []
jenkins-mlpack has quit [Ping timeout: 260 seconds]
jenkins-mlpack has joined #mlpack
vivekp has quit [Ping timeout: 240 seconds]
vivekp has joined #mlpack
prakhar_code[m] has quit [Ping timeout: 256 seconds]
killer_bee[m] has quit [Ping timeout: 248 seconds]
caladrius[m] has quit [Ping timeout: 256 seconds]
prakhar_code[m] has joined #mlpack
killer_bee[m] has joined #mlpack
caladrius[m] has joined #mlpack
dk97[m] has joined #mlpack
Nisha_ has joined #mlpack
caladrius[m] has quit [Ping timeout: 256 seconds]
dk97[m] has quit [Ping timeout: 255 seconds]
prakhar_code[m] has quit [Ping timeout: 252 seconds]
killer_bee[m] has quit [Ping timeout: 248 seconds]
prakhar_code[m] has joined #mlpack
united_ has joined #mlpack
killer_bee[m] has joined #mlpack
dk97[m] has joined #mlpack
caladrius[m] has joined #mlpack
united_ has quit [Ping timeout: 260 seconds]
united_ has joined #mlpack
< united_>
Does mlpack implement CNNs? I couldn't find them in the documentation. Also, if I want to use computer vision techniques, it would be better if mlpack has the standard networks like alexnet, VGG, Resnet, etc. Is there any way mlpack users can get access to such networks?
< rcurtin_>
if you'd like to implement one of the standard networks you were thinking of, you could submit it to the models repository for other users: https://github.com/mlpack/models
< united_>
Cool. I will check which models have been implemented and try implementing a new one. Thanks!
united_ has quit [Ping timeout: 260 seconds]
< rcurtin_>
sounds good :)
govg has quit [Ping timeout: 240 seconds]
vivekp has quit [Ping timeout: 264 seconds]
vivekp has joined #mlpack
govg has joined #mlpack
vivekp has quit [Ping timeout: 276 seconds]
vivekp has joined #mlpack
ImQ009 has joined #mlpack
manthan has joined #mlpack
Nisha_ has quit [Quit: Connection closed for inactivity]
ricklly has joined #mlpack
< ricklly>
hello, I am working on mlpack decision tree benchmark. Can some explain
< ricklly>
Unfortunately, it can not parse the options after I run it on my shell
dmatt has joined #mlpack
dmatt has left #mlpack []
sourabhvarshney1 has joined #mlpack
< zoq>
ricklly: You are right, it's used to parse the output, mainly the timing output. The first step is probably to run the command line manually and to check the output, and too adept the parsing step afterwards.
< sourabhvarshney1>
@rcurtin @zoq Can you please help me over the issue that I am not able to create PRs over many other forked repos of mlpack other than original repo?
< zoq>
sourabhvars: Have you tested to fork the repo over the github interface?
< sourabhvarshney1>
I think we can not fork the repo again if we have the original repo as such like in my case I have already forked mlpack/mlpack but I am not able to create PR on manthan/mlpack?
< zoq>
sourabhvars: you could remove the existing one, I guess you have a local copy anyway.
< sourabhvarshney1>
But I think this is not the best solution. Because whenever I need to create PR to any other fork I need to delete that and add a new one. This doesn't sounds good. Looks like there is a privacy issue with this.
< sourabhvarshney1>
@zoq Can we find another solution to it?
manthan has quit [Ping timeout: 260 seconds]
< zoq>
sourabhvars: You can test it out, if it works the github settings are fine, and it's a local issue.
dmatt has joined #mlpack
< sourabhvarshney1>
@zoq I tried but it is not working.
< sourabhvarshney1>
I think the problem lies in the personal settings of the repository. Might be the case, but not sure.
< zoq>
sourabhvars: hm, okay, I'll take a closer look later today
< sourabhvarshney1>
Thanks @zoq
manthan has joined #mlpack
< sourabhvarshney1>
@manthan for which tests in decision tree test, we need to have pruning test. We need not to have them for all. We can have it for tests like InformationGainManyPoints etc.
< sourabhvarshney1>
Your ideas?
dmatt has quit [Remote host closed the connection]
< manthan>
i was in the process of thinking about tests and I think that testing the pruning on trees formed through different fitness functions for both numeric and categorical types be sufficient
< manthan>
so we will write tests for each fitness function (for both numeric and categorical data)
< manthan>
if for all the pruned tree is small and produces less error than actual tree on validation set then the algorithm can be verified
< sourabhvarshney1>
Sounds good to me. @manthan we can choose a random training dataset and then train the tree. Then check again for a random validation data set.
< sourabhvarshney1>
This we can do for different fitness functions. What say?
< manthan>
ya, one way could be to split the already used dataset into training and validation sets and use those for our purposes
< sourabhvarshney1>
That is also possible.
dmatt has joined #mlpack
dmatt has quit [Client Quit]
dmatt has joined #mlpack
manthan has quit [Ping timeout: 260 seconds]
manthan has joined #mlpack
sourabhvarshney1 has quit [Ping timeout: 260 seconds]
< ricklly>
what's the relationship bettween iris_train.csv and iris_labels.csv?
< ricklly>
why can not we use iris_train.csv and iris_train_labels.csv?
< zoq>
ricklly: iris_labels.csv contains the labels of the test set
< zoq>
iris_train.csv already contains the labels see the last row
< manthan>
zoq : could you have a look at the pruning PR?
< ricklly>
wait
dmatt_ has joined #mlpack
< zoq>
manthan: We are in the process of releasing mlpack 3.0, so this has to wait at least one more day.
< manthan>
ok sure, thanks :D
< manthan>
just curious to know, does the present master becomes 3.0?
< ricklly>
zoq: I get it. I have check the iris_train.csv downloaded from mlpack.org/dataset. But no labels exists in the iris dataset under the build directory
< ricklly>
waww, 3.0!
dmatt has quit [Ping timeout: 260 seconds]
< ricklly>
I think a identifiable file name is important
< ricklly>
zoq: I wonder if there is necessary to make L-BFGS faster?
dmatt has joined #mlpack
< zoq>
ricklly: A faster optimizer is always great to have, if you find something that could be improved, please feel free to open a PR.
< zoq>
manthan: it's pretty much the same, some more fixes going in as well.
< rcurtin>
yeah, in my view the big effort is not so much the code right now but the documentation and making sure that new users can figure out how to quickly use mlpack
dmatt_ has quit [Ping timeout: 276 seconds]
dmatt_ has joined #mlpack
dmatt has quit [Ping timeout: 255 seconds]
< manthan>
ohh i see, thanks.
< ricklly>
Ok, right! The top priority is sufficient documentation. I really feel bad when I first met MLPack... I have to read the code and understand the logic
< ricklly>
A perfect python notebook tutorial is welcome
< manthan>
I think an open issue for addition of documentation would have drawn some contributions for documentation. That would have perhaps helped in the process
dmatt has joined #mlpack
< rcurtin>
in my opinion it is very very hard to have new contributors write documentation, because they may not fully understand what it is they are documenting
dmatt_ has quit [Disconnected by services]
dmatt_ has joined #mlpack
dmatt has quit [Ping timeout: 260 seconds]
dmatt has joined #mlpack
dmatt_ has quit [Disconnected by services]
dmatt_ has joined #mlpack
Prabhat-IIT has joined #mlpack
dmatt has quit [Ping timeout: 260 seconds]
dmatt_ is now known as dmatt
< Prabhat-IIT>
zoq: Do we have something like std::map or something like std::vector with push_back capabilities in armadillo?
< rcurtin>
Prabhat-IIT: you can use join_rows() or join_cols() or insert_rows() or insert_cols(), but almost always it is better to allocate all the necessary space in the beginning and then fill the elements
< Prabhat-IIT>
rcurtin: Thanks! btw Is it acceptable to use something like std::vector use its push_back capabilities and later convert it into arma::vec through advanced constructor using iterator in mlpack?
< Prabhat-IIT>
or will it have some extra overheads and that's why we don't generally use std::containers
dmatt_ has joined #mlpack
dmatt has quit [Ping timeout: 276 seconds]
< rcurtin>
Prabhat-IIT: it depends a lot on the situation, but in general, I would say that it is typically better to find the size of the arma::vec you will need and then fill it
< rcurtin>
but that is not always possible---like I said, it depends on the situation
< Prabhat-IIT>
rcurtin: Thanks, I was facing this issue while implementing something!
dmatt has joined #mlpack
dmatt has quit [Remote host closed the connection]
dmatt_ has quit [Ping timeout: 255 seconds]
govg has quit [Ping timeout: 256 seconds]
govg has joined #mlpack
Prabhat-IIT has quit [Ping timeout: 260 seconds]
vivekp has quit [Ping timeout: 276 seconds]
vivekp has joined #mlpack
< manthan>
rcurtin : true, i understand the problem with new contributors
csoni has joined #mlpack
travis-ci has joined #mlpack
< travis-ci>
mlpack/mlpack#4555 (master - bd91b37 : Ryan Curtin): The build has errored.
csoni has quit [Read error: Connection reset by peer]
dmatt has joined #mlpack
dmatt has quit [Remote host closed the connection]
Prabhat-IIT has joined #mlpack
< Prabhat-IIT>
zoq: you there?
dmatt has joined #mlpack
< zoq>
Prabhat-IIT: yes
< Prabhat-IIT>
zoq: I've added the sphere function test in SAGA as well as made the modifications for SAG but still the sphereFunction test is still failing on Travis and is passing locally
< Prabhat-IIT>
This had happened with me a lot of time I don't know why
< Prabhat-IIT>
That test passes locally but will not pass on Travis
< zoq>
did you use another random seed for each run?