13:41 < nilkel> zoq: Sorry for the late reply. Regarding what I asked yesterday, I've built everything via cmake and I'm trying to compile kfn_main.cpp using g++.

13:42 < rcurtin> nilkel: why are you trying to compile kfn_main.cpp by hand? the CMake build system is there so that you can build mlpack_kfn by typing 'make mlpack_kfn'

13:45 < kaushik_> rcurtin: hi, I have been going through the code and think I will start with a working on a prototype for tf-idf.

13:46 < kaushik_> dictionary_encoding.hpp/cpp and dictionary_encoding_main be in mlpack/core/data right?

13:49 < kaushik_> rcurtin: I had a question, why is the dataset_mapper implementation in .hpp file. Everywhere else the class declaration is in .hpp and implementation in .cpp. also the module_main.cpp handles the commandline.

13:50 < travis-ci> mlpack/mlpack#4039 (master - 2bb965d : Ryan Curtin): The build has errored.

13:50 < travis-ci> Change view : https://github.com/mlpack/mlpack/compare/617a2acb5261...2bb965d23b72

13:50 < travis-ci> Build details : https://travis-ci.org/mlpack/mlpack/builds/343816574

13:51 < nilkel> rcurtin: I was just trying to do it using the command line instructions given in the tutorial. It's working properly with make.

13:59 < rcurtin> nilkel: can you tell me where the documentation says to compile the command-line programs by hand? that's not a thing that is easy to do (you have to define lots of macros and use specific options), so if the documentation told you to do that we should clarify it :)

14:00 < rcurtin> kaushik_: all template classes must be available in header files

14:02 < nilkel> rcurtin:"On this page, several simple mlpack examples are contained, in increasing order of complexity. If you compile from the command-line, be sure that your compiler is in C++11 mode. With gcc and clang, this can be accomplished by adding the -std=c++11 option."

14:02 < nilkel> I thought I'd give the command line a shot

14:03 < nilkel> http://www.mlpack.org/docs/mlpack-2.2.5/doxygen/sample.html

14:21 < rcurtin> nilkel: thanks; I'll clarify this to point out that the command-line programs like `kfn_main.cpp` and similar shouldn't be compiled by hand since they require special arguments

14:21 < rcurtin> the idea there was just that you'd take the simple code examples and compile them directly with gcc

15:38 < travis-ci> mlpack/mlpack#4041 (master - 6c3f84b : Ryan Curtin): The build has errored.

15:38 < travis-ci> Change view : https://github.com/mlpack/mlpack/compare/2bb965d23b72...6c3f84b50fc4

15:38 < travis-ci> Build details : https://travis-ci.org/mlpack/mlpack/builds/343855342

15:39 < satyam_2401> Hello all. I am interested in contributing to particle swarm optimisation. I have been working on PSO and NSGA for past few months. For single objective, pso has been quite successful. But for multi-objective, I have been trying to combine pso with nsga-3,(reference based approach, ranking and crowding distance) to improve the diversity of the solution.

15:40 < satyam_2401> From where should I start, if I am interested to contribute to pso?

15:42 < zoq> satyam_2401: Hello, there is an open discussion on the mailing list about the PSO project, with some additional pointers: http://knife.lugatgt.org/pipermail/mlpack/2018-January/thread.html#3438

15:43 < zoq> satyam_2401: NSGA 2/3 is also an intersting method.

15:45 < satyam_2401> Yes, combining nsga3 with pso provides very interesting results. A swarm can be divided into multiple swarms and we can assign a global guide to each reference line as mentioned in nsga 3 to improve pso to a large extent.

15:46 < zoq> Agreed, I have seen some really good results.

15:49 < crest> Hi ! I'm trying to use mlpack but an error occurs when i try to execute "mlpack_knn --help" . "libmlpack.so.2 : cannot open shared object file . No such file or directory "

15:50 < zoq> crest: Hello can you search for "libmlpack.so.2" on https://github.com/mlpack/mlpack/blob/master/README.md

15:51 < zoq> crest: Let me know if I should clarify anything

15:54 < satyam_2401> @zoq , Should I direct my query on discussion about PSO to marcus on mailing list?

15:55 < zoq> satyam_2401: either mailing list or IRC is fine

15:56 < crest> @zoq found the problem. Thanks

18:29 < vss> @rcurtin hi! I looked up the ticket under "Improvement of traversers" and i noticed that the cover tree implementation needs a lot more number of comparisons than the kd tree implementation. Does that just mean the dual cover tree is not as viable as dual kd tree ?

18:39 < rcurtin> vss: that depends strongly on the properties of the dataset

19:24 < moksh> Hey, I am Moksh Jain, from NITK, India. I'd like to apply for GSoC this year. I am interested in Reinforcement Learning and I would like to take up that project. I have built and installed mlpack from the source and have written simple models using it. I have also started reading the Double DQN and Proximal Policy Optimization Algorithms papers. So can anyone please guide me regarding what I should do next? Thanks in advance :)

19:43 < rcurtin> hi Moksh, welcome to the community. There are some pointers for how to get started in http://www.mlpack.org/involved.html and http://www.mlpack.org/gsoc.html; have you taken a look at those?

19:47 < robertohueso> rcurtin: I am going to try to implement KDE #150 (Since it's been over a month since @serendipity5497 wrote his comment) but I'm still not sure if it would be too hard for me, should I write a comment in the issue?

19:52 < rcurtin> robertohueso: I guess it would be a good idea to "claim" it like that

19:52 < rcurtin> that or open a WIP PR or something like this to mark that you are working on it could be helpful

19:53 < rcurtin> but KDE is one of the more difficult things, and to be honest, there is less interest in any of the harder issues, so it is less likely that multiple people will work on them :)

19:53 < rcurtin> I would suggest, as a first pass, implement brute-force KDE

20:06 < robertohueso> rcurtin: Thank you :)

20:12 < rcurtin> sure, happy to help

21:06 < travis-ci> mlpack/mlpack#4047 (master - af3c00b : Marcus Edel): The build has errored.

21:06 < travis-ci> Change view : https://github.com/mlpack/mlpack/compare/6c3f84b50fc4...af3c00b164a9

21:06 < travis-ci> Build details : https://travis-ci.org/mlpack/mlpack/builds/344002558

21:10 < satyam_2401> https://www.irccloud.com/pastebin/RYL5q6ZV/

21:12 < satyam_2401> For unconstrained problem, I think we should try merging nsga3 technique with the pso

21:17 < zoq> satyam_2401: hm, I'll have to think about it, but maybe it makes sense to implement it as an own class or to use the policy design pattern.

21:17 < zoq> I think an own class would make sense if we also implement NSGA2.

21:40 < satyam_2401> https://www.irccloud.com/pastebin/wUxkiYbu/

21:52 < zoq> Agreed Strategy pattern is an option, however, ideally, we should be able to do the decision at compile time, to avoid any performance penalty.

21:52 < zoq> Also, I like idea number 6 and 5, do you think we could come up with an initial idea of an interface? Might be easier to go from there, let me know what you think.

22:00 < rcurtin> I still have no public IPs for the build systems

22:00 < rcurtin> I let them know it's now been 11 days of downtime, but I am not sure that anyone cares... I am getting kind of frustrated

22:04 < zoq> rcurtin: Starting to miss masterblaster.

22:05 < rcurtin> same, it is really important for processing PRs...

22:08 < zoq> ssh forwarding :)

22:12 < rcurtin> I would do that, but I don't even have external connectivity at the moment; something is wrong with the networking

22:16 < zoq> I don't think it makes sense right now to host Jenkins somewhere else and only use masterblaster and co as nodes, but maybe something to think about for the future.

22:18 < rcurtin> you're right, maybe that is a better long term solution

22:18 < rcurtin> I don't think there is a plan in place to close the Tucson datacenter but it is always a possibility I guess...

22:18 < rcurtin> the best alternative would probably be to run Jenkins on mlpack.org directly

22:21 < zoq> Right, I guess it's always nice to have something we could use in such a situation, or if the DNS breaks and nobody is going to fix it etc...