00:17 < zoq> gauravcr7rm: For NEAT one example could be to test it on some simple task like XOR, or mountain car.

00:20 < zoq> rf_sust2018: Bindings for Java/Scala would be nice to have, Yasmine worked on Go bindings and I think Julia bindings are coming pretty soon.

00:23 < zoq> ReemGody_: Welcome, I think you already glanced over http://mlpack.org/gsoc.html? If every issue you are interested in is already taken up, you could see if you can extend or improve an existing method.

01:30 < Karl> mulx10: Hi, thanks for checking. The cmake seems to be ok

01:34 < Karl> mulx10: There several 'not find' as following.

01:34 < Karl> mulx10: -- Could NOT find OpenMP_C (missing: OpenMP_gcrt1_LIBRARY) (found version "3.1") -- Could NOT find OpenMP_CXX (missing: OpenMP_gcrt1_LIBRARY) (found version "3.1") -- Could NOT find OpenMP (missing: OpenMP_C_FOUND OpenMP_CXX_FOUND)

01:35 < Karl> mulx10: I read some post before talking about it is because the gcc version is too low. Currently, my gcc is as following.

01:35 < Karl> mulx10: -- The C compiler identification is GNU 4.8.5 -- The CXX compiler identification is GNU 4.8.5

01:35 < Karl> mulx10: Other 'not find' are following.

01:36 < Karl> mulx10: CMake Warning at CMakeLists.txt:514 (message): txt2man not found; man pages will not be generated.

01:36 < Karl> mulx10: -- Not building Markdown bindings.

01:36 < Karl> mulx10: -- Could NOT find Doxygen (missing: DOXYGEN_EXECUTABLE)

01:37 < Karl> mulx10: From my understanding, these are not crucial for mlpack being able to work, aren't they? So I think my cmake works correctly.

01:38 < Karl> mulx10: after trying what jeffin suggested, I still got the error like following.

01:39 < Karl> mulx10: [ 55%] Linking CXX executable ../../../../bin/mlpack_adaboost Undefined symbols for architecture x86_64: "boost::program_options::to_internal(std::basic_string<char, std::char_traits<char>, std::allocator<char> > const&)", referenced from: std::vector<std::basic_string<char, std::char_traits<char>, std::allocator<char> >, std::allocator<std::basic_string<char, std::char_traits<char>, std::allocator<char> > > > boost::p

01:39 < Karl> mulx10: I am sure the boost package work now, because I checked with 'brew list --versions boost', and output is 'boost 1.69.0'.

04:28 < Hemal> Is there any way to use the Reinforcement learning algorithms of mlpack for game simulations ?

04:30 < Hemal> One way I can think of is, Using mlpack's python bindings and other python packages to run a simulation.

04:30 < Hemal> But can it be done natively, i.e using c++ ?

05:16 < jeffin143> Hi, contributors and mentors,I am jeffin Sam currently doing Btech 3rd year from Karunya institute of Technology and Sciences.Today is the last day before application opens up , and i have enjoyed my time researching about the code base and also about organization. I would like to take up string processing utilities as my idea and would like to build upon it.

05:19 < jeffin143> I have already raised some PRs to afirm my knowledge in processing of string which includes one hot encoding , dictionary encoding and also have started contributing to mlpack. I have worked in ML/DL field for a couple of years now and know NLP concepts too.

05:22 < jeffin143> For time being, I am progressing to build an api and make my utility function accessible for users without any issues.A sincere thanks to all the mentors and contributors for helping me with all the issues and queries.

05:28 < mulx10> Karl: CHeck this out (https://stackoverflow.com/questions/11852568/gcc-4-8-on-mac-os-x-10-8-throws-undefined-symbols-for-architecture-x86-64)

05:29 < mulx10> Karl: I guess this may help

05:31 < mulx10> Karl: I guess the gcc version is the issue

07:27 < Hsankesara> Hello community, I am facing some trouble building mlpack_test on my local system. I want to run test for a few methods only but I cannot find a way to do it on test. Building all mlpack_test took too much time and eventually led to memory error. Is there any way I can do it?

07:50 < jenkins-mlpack2> Project docker mlpack nightly build build #272: UNSTABLE in 3 hr 36 min: http://ci.mlpack.org/job/docker%20mlpack%20nightly%20build/272/

08:37 < myaow> Hi everyone! My name is Monica and I'm currently a student at Harvey Mudd College. I've taken a course on neural nets so I have a working understanding of the topic, and I have also implemented some NN and RL algorithms. In the last few weeks, we touched more on RL and the specifics of Alpha Go. I would love to work on the RL project, and I am excited to read papers that describe state of the art of RL+NN!

09:31 < Suryo> zoq, rcurtin: hey, I know you are really busy but could you please give me a ballpark figure on how long proposals to mlpack usually are? In terms of pages. Thanks...

10:07 < zoq> Suryo: We have seen proposals from 1-25+ pages; ideally you can cover everything that is mentioned here: https://github.com/mlpack/mlpack/wiki/Google-Summer-of-Code-Application-Guide.

10:37 < Suryo> zoq: okay, thanks. That's amazing.

12:14 < sreenik> zoq: Just like Linear, Relu and other layer objects can be stored in a LayerTypes instance, what can be used to store initialization, loss and optimizer objects? I have tried experimenting a lot but couldn't really figure out what to use

12:15 < sreenik> Is there anything on the lines of InitializationRuleType or OutputLayerType?

12:39 < zoq> sreenik: Not sure I get what you mean, can you give me a simple example?

12:42 < sreenik> Yeah sure. What I want to do is something like -- InitializationType init = new GlorotInitialization();

12:43 < sreenik> But there is nothing called InitializationType

12:44 < sreenik> Is there anything similar to that? I want to create an FFN object later with that init type

14:55 < mulx10> sreenik: Using same type as class may help `GlorotInitializationType<false> *init = new GlorotInitialization();`

14:56 < mulx10> sreenik: or I good hack is to use auto `auto *init = new GlorotInitialization();`

14:58 < mulx10> ShikharJ, zoq: Apologies for the late reply. With respect to Pull #1800, The cellState is output param just like lstm output.

14:58 < mulx10> ShikharJ, zoq: hence I think there is no need to serialize.

15:00 < mulx10> zoq: Also I have finding it difficult to find some test case for the functionality. I have added dimension check.

15:01 < mulx10> zoq: Please let me know what should be included in the test. Thanks for the help.

15:04 < sreenik> mulx10: Thanks for the response. Unfortunately that won't actually solve my problem. I have a function that is supposed to return a particular initialization type depending on some parameters. So I don't know before hand if it is going to return glorot initialization or random initialization or anything else. Knowing the return type will solve my problem :)

15:06 < mulx10> sreenik: So 'auto' may help

15:13 < sreenik> Functions can have an auto return type but only if another return type is specified after the declaration. Basically the return type needs to be mentioned because C++ is a statically typed language. So my only way out is knowing the class type whose objects can be instantiated by all the intitialization types. I have searched the codebase a couple of times but couldn't find a straightforward solution. This might require a roundabou

18:41 < sreenik> Regarding the problem I was facing it can also be solved if the 'init type', 'loss type' and 'optimizer' can be separately set for the FFN (i.e. afterwards and not during the FFN declaration). Is that possible? thanks :)

19:38 < rcurtin> myaow: hi there, thanks for getting in touch. Let us know if there's anything that we can clarify :)

19:38 < rcurtin> Hsankesara: I saw your message on the Github PR, I'll comment there shortly

19:39 < rcurtin> jeffin143: hi there, I've seen your PRs in the past days. I'm hoping to find time to review them today :)

20:47 < zoq> sreenik: Sorry for the slow response, does eomthing like: https://github.com/mlpack/mlpack/blob/85e00107fe3e2bb8d12f331a725cdfe11acb290b/src/mlpack/methods/kernel_pca/kernel_pca_main.cpp#L135-L176 this work for you as well?

20:48 < zoq> sreenik: The last solution you pointed out would work as well, since we only initalize the network once.

20:49 < zoq> sreenik: I can sketch something up if you like, but perhaps the same solution we used for kpca works here as well.

21:13 < Karl> mulx10: I change the compiler from gcc-4.8 to gcc-7.

21:15 < Karl> mulx10: lots of warnings are resolved, but the boost undefined error still persists. Looks like the boost complied by homebrew is using clang instead of gcc and this makes the error. https://github.com/conda-forge/staged-recipes/issues/5812

21:15 < Karl> mulx10: I guess I will try to get a linux system and install mlpack on that.

23:27 < Suryo> zoq: thanks for the comments on PR#86. I'll take care of the things you pointed out.