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/
ImQ009 has joined #mlpack
marko has joined #mlpack
< marko> hi
< marko> this throws
< marko> read access violation
< marko> arma::mat features(11, 100);
< marko> arma::rowvec targets(1, 100);
< marko> mlpack::regression::LARS glm (false, .5, .5);
< marko> std::vector<mlpack::regression::LARS> inventoryofModels;
< marko> inventoryofModels.emplace_back(glm);
< marko> inventoryofModels[0].Train(features, targets);
< marko> LARS object is placed into a vector
< marko> and than accessed in the vector
< marko> in order to apply "Train" method
< marko> and this throws an exception
< marko> W10, VS2017
< marko> the debug is indicating line 235 in LARS.cpp.
< marko> matGramActive(i, j) = (*matGram)(activeSet[i], activeSet[j]);
< marko> anyone over there?
< marko> :)
< Manav-KumarGitte> Does mlpack have rule base classifier like ripple method?
< marko> trees are the closest you can find to rule based
< marko> browse here
< Manav-KumarGitte> But I guess they don't have rule based classifier which applies fuzzy logic in them?
< marko> if you cannot find it in docs, its not there
< marko> ;)
< rcurtin> marko: looks like maybe LARS needs a correct copy constructor? you could try something like this as a workaround:
< rcurtin> basically the idea would be to use std::vector<mlpack::regression::LARS*> instead (i.e. a vector of pointers) so that you can avoid the copy constructor
< rcurtin> it's not a great solution, but it's at least a workaround
< rcurtin> if you want to report this as a bug on github that would be great :)
< marko> let me try if that works
< rcurtin> you'll want to allocate each LARS object with `new` and make sure to `delete` it when the program is over :)
< marko> rcurtin many thanks
< marko> lets try
< rcurtin> sure; sorry that the copy constructor is not correct to begin with :)
< marko> khm...
< marko> * seems not to help
< marko> can you please clarify the second approach
< rcurtin> sure, hang on
< rcurtin> (then don't forget to call `delete inventoryofModels[0]` when you're done with everything)
< rcurtin> another approach would be to call Train() before emplace_back(), but I don't know if that works with your situation
< marko> is push_back mandatory)
< marko> ?
< rcurtin> hmmm, maybe emplace_back() will work with the pointer, but basically I want to avoid calling the copy constructor of LARS
< rcurtin> give it a shot, see what happens :)
< marko> i have placed the error in the github
< marko> there must be a workaround
< marko> Train does not work
< marko> Predict does work, that is funny
< marko> rcurtin
< marko> 'initializing': cannot convert from 'mlpack::regression::LARS' to '_Objty'
conrad41 has joined #mlpack
< rcurtin> does it say what _Objty is?
< rcurtin> also, what line gives that error?
< rcurtin> I just guessed at my workaround code, I didn't try it. so maybe some extra debugging is needed :)
< marko> thanks anyway
< marko> :D
< marko> here is the github error
< marko> rcurtin so we may continue over there
< marko> thanks
marko has left #mlpack []
< rcurtin> marko: sounds good---I'll see if I can find some time to look later today
Yashwants19 has joined #mlpack
< Yashwants19> Hi rcurtin, I saw something here https://www.mlpack.org/doc/mlpack-3.3.0/julia_documentation.html#sparse_coding. I think so we are not providing these options (`copy_all_inputs=false, help=false, info="", version=false`) with julia.
< Yashwants19> Or may be I misunderstood something here.
Yashwants19 has quit [Remote host closed the connection]
< rcurtin> Yashwants19: interesting, `verbose` is valid, but the other three aren't; I'll look into why those are being printed
< rcurtin> the `info`, `help`, and `copy_all_inputs` inputs aren't actually a part of the Julia binding in the .jl, so it must be a printing bug
Yashwants19 has joined #mlpack
conrad41 has left #mlpack []
< Yashwants19> I think so we are allowing all persistent there.
< Yashwants19> We must allow `verbose` only.
< Yashwants19> *persistent options
Yashwants19 has quit [Ping timeout: 240 seconds]
< rcurtin> Yashwants19: if you want to beat me to it and open a PR feel free :) I'm pretty busy today so it might be a bit
< rcurtin> I think you're right about the solution
Yashwants19 has joined #mlpack
< Yashwants19> Great I will open a PR soon :)
Yashwants19 has quit [Remote host closed the connection]
PrinceGuptaGitte has joined #mlpack
< rcurtin> PrinceGuptaGitte: looks like it found the headers for Armadillo, but not libarmadillo.so; maybe you need to specify `-DARMADILLO_LIBRARY=/path/to/libarmadillo.so` in your CMake configuration?
< PrinceGuptaGitte> Thanks that worked. Now I'm getting "Unable to find the requested Boost libraries.", I suppose it's the same problem. But why could it be happening? I've never had to explicitly specify path to all the libraries during `cmake ../`
< rcurtin> did you clean your build directory after upgrading your system? you might want to do that
< rcurtin> i.e. either remove the entire build directory and make a new one, or, `rm CMakeCache.txt` will probably work too
< PrinceGuptaGitte> I did clean my build directory when I was getting this error.
< rcurtin> if that's happening from a clean build directory, I'm not sure what the issue is; I guess you can work around it by manually specifying the location of the Boost libraries
< rcurtin> if you find something wrong in the CMake configuration, feel free to open an issue, but I don't think that anything's wrong... I don't think we've changed how dependencies are found recently?
< PrinceGuptaGitte> Yeah, I guess I'll have to stay with that for some time.
< PrinceGuptaGitte> I just saw my PR #2184 went stale. What should I do about this?
< rcurtin> you can comment on it or push commits and it'll become un-stale; I see that there are merge conflicts, so if you can fix those it would be great
< rcurtin> personally I haven't had any time to review it
< rcurtin> I'm trying to figure out what's going on with the video meeting... is anyone else successfully joined?
< rcurtin> oh, sorry, stupidity on my end
ImQ009 has quit [Quit: Leaving]
< rcurtin> ahhh, those ensmallen stickers I just ordered... delayed: "These items will ship when normal service resumes." so who knows when they'll get here :)
< zoq> I wonder if that is the same here in Germany.
< zoq> Intersting, I think they got more expensive since the last time I checked, they want 105 Euro for 1k.
< jeffin143[m]> Anybody from Russia ?
< rcurtin> yeah, I got 1k stickers for $188 USD last time, this time it was $210
< zoq> rcurtin: for the circle sticker, 25mm x 25mm?
< zoq> Maybe I should order the batch? $210 is even more expensive.
< rcurtin> 2" x 2" (so ~50mm x 50mm)
< rcurtin> the mlpack stickers that I ordered in the past are that size, so I figured I'd go with the same size
< zoq> Ahh, that is 170 Euro, here
< rcurtin> it scales sublinearly with squared mm :)
< zoq> I don't mind to order some, but if we ask NumFOCUS to reimburse us, it doesn't matter.
< rcurtin> yeah, I'll get in touch with NumFOCUS and see what they think
< rcurtin> I figured I'd mail a bunch of the ones that I got to you, so that we have two locations we can send from :)
< zoq> rcurtin: Happy to have some here and mail them out :)
< rcurtin> :)
< sreenik[m]> It was a very cool discussion today, thanks Shikhar and Ryan (both :D)
< rcurtin> :)
< zoq> The sessions terminates after some time?
< rcurtin> the video chat? it should actually just be an always-open zoom room that anyone can use at any time
< sreenik[m]> It terminates after the host leaves I guess (i.e., if the host doesn't hand over before leaving)
< rcurtin> really?? sorry about that, I thought I had it configured differently
< sreenik[m]> No let me see if I can join now
< sreenik[m]> We hanged out for around half-an hour after you left :)
< rcurtin> ahhh, ok, that's good, it would be bad if it terminated immediately
< jeffin143[m]> I remember Omar was the host
< jeffin143[m]> After rcurtin lefy
< sreenik[m]> Hehe
< jeffin143[m]> Left*
< sreenik[m]> I checked, yes, we can join anytime
< jeffin143[m]> So now the power is with shrit to terminate the room
< jeffin143[m]> May it be good , two guys working on a pr, to pair programming sharing screen on always opened zoom room
< rcurtin> yeah, the zoom room would be perfect for things like that, feel free to use it that way :)
< jeffin143[m]> 🤗
< sreenik[m]> I think it'll be very effective during the gsoc
yernar[m] has joined #mlpack