00:05 < zoq> At least there is progress.

01:08 < travis-ci> mlpack/mlpack#3945 (master - 5662319 : Ryan Curtin): The build has errored.

01:08 < travis-ci> Change view : https://github.com/mlpack/mlpack/compare/253ba103c4b5...5662319dc25d

01:08 < travis-ci> Build details : https://travis-ci.org/mlpack/mlpack/builds/340744296

01:49 < travis-ci> ShikharJ/mlpack#73 (ResizeLayer - 02bf568 : Shikhar Jaiswal): The build has errored.

01:49 < travis-ci> Change view : https://github.com/ShikharJ/mlpack/compare/fcbb2cce1f18...02bf568ffe43

01:49 < travis-ci> Build details : https://travis-ci.org/ShikharJ/mlpack/builds/340767183

04:39 < sshkhrnwbie> @zoq : this is regarding the parameter modetype (fan-in/fan-out/fan-avg) for the variance scaling initializer class. I tried to implement it the same way Kris was doing for Normal/Uniform.

04:41 < sshkhrnwbie> I tried creating a class template which accepts as parameter one of 3 empty classes fan-in, fan-out, fan-avg and then using (std::enable_if_t<std::is_same<Mode, template parameter >::value>* = 0) in the class constructor

04:44 < sshkhrnwbie> However I was running into build errors for the init_test which I wrote based on the above definitions. I am not very thorough with template programming so could use some comments if this is a correct approach or to try something else.

04:45 < sshkhrnwbie> Thanks

05:11 < coolprinshu> \join

09:26 < KishoreKaushal> Hey Everyone ! I am Kaushal Kishore a second year undergraduate at the department of Computer Science & Engineering at Indian Institute of Technology, Palakkad. And I would like to contribute to the MLPACK in the GSOC 2018.

09:31 < KishoreKaushal> I have read some of the project Ideas posted on you github page. I would like to contribute to the Deep Learning Section.

09:57 < zoq> KishoreKaus: Thanks for getting in touch, there has been some discussion about the project idea on the mailing list, so my recommendation is to search the mailing list first: http://mlpack.org/gsoc.html. Also, the models listed on the ideas page are just suggestions, so if you have something interesting in mind, feel free to start a discussion.

10:01 < zoq> sshkhrnwbie: So it sounds like you did something like this: https://github.com/mlpack/mlpack/tree/master/src/mlpack/methods/ann/convolution_rules? I think that is a good way to tackle the problem if you like I can take a look at the code and I'm certain together we can figure out what went wrong.

10:34 < rajeshdm9> Hey everyone, I'm Rajesh from International Institute of Information Technology, Bangalore (IIIT-B) in my final year of Masters. I want to contribute to mlpack as part of GSoC 2018. I have gone through the list of potential ideas. I would love to contribute to the Reinforcement learning project. I feel I have the right background for this as I am doing my thesis in the field of reinforcement learning and have already gone th

10:34 < rajeshdm9> (Double DQN and Proximal Policy Optimization Algorithms) as part of my thesis literature review. I also have a strong background in C++ as I've done multiple projects in C++ in the last 5 years. So please let me know what is expected of me to be able to contribute to GSoC 2018.

10:41 < zoq> rajeshdm9: Hello and welcome, getting familiar with the codebase especially src/mlpack/methods/reinforcement_learning/ is the first step. Also Shangtong's blog posts might be helpful: http://www.mlpack.org/gsocblog/ShangtongZhangPage.html

10:42 < zoq> rajeshdm9: Another step would be to understand and run the existing rl tests: (rl_components_test.cpp) 'bin/mlpack_test -t RLComponentsTest' and (q_learning_test.cpp) bin/mlpack_test -t QLearningTest, if you like you can implement another task and see if you could solve it with the existing code.

10:42 < zoq> rajeshdm9: As always if you have any question, please don't hestiate to ask.

10:47 < rajeshdm9> Sure, I will go through the resources and get back to you.

10:48 < zoq> rajeshdm9: Sounds good :)

10:50 < sshkhrnwbie> @zoq : So should I submit a pr with the code i have a written? You can suggest me the changes which I will add to it later.

10:51 < zoq> sshkhrnwbie: Yeah, that would be the easiest solution, I'll take a look at the code once I have a chance.

10:52 < sshkhrnwbie> and yes I was trying something similar to the border modes for the mode of the initializer

10:52 < sshkhrnwbie> Okay I will go ahead and submit a pr

10:52 < zoq> okay, great.

16:31 < bhuwnesh> I've decided to work for the "Essential Deep Learning Modules" project. Take me through the all requirements.

17:36 < s1998_> rcurtin: binary_space_tree_impl.hpp uses the fact that metrictype::evaluate is a static function. But in case of hdbscan_metric the function does not remain static. Do you think there is a way around or the binary space tree code should be refactored ?

17:36 < s1998_> https://github.com/mlpack/mlpack/blob/5662319dc25d78107e470f4695678956c3f98a95/src/mlpack/core/tree/binary_space_tree/binary_space_tree_impl.hpp#L797

17:37 < s1998_> Regarding union_find code duplication, I think that is doable and will push the changes soon.

17:38 < s1998_> *w.r.t union_find code duplication comment, I think the duplication can be removed and I will push the changes soon.

17:48 < rcurtin> s1998: I think modifying binary_space_tree is fine

20:33 < rcurtin> Tucson reports that all of the build infrastructure is rackmounted now... but none of it is accessible and I still have no external IPs for any of it

20:33 < rcurtin> but I have a call with the team in 30 minutes, so maybe they can shed some light on what is going on...

21:04 < kaushik_> Hi for the "String Processing Utilities" project are the mappings already decided or that need be included in proposal based on research?

21:14 < rcurtin> kaushik_: the proposal is up to the student, so you should do what you think is most appropriate

21:14 < rcurtin> for that project keep in mind there will be users of the software you make, so you should be designing the software from their perspective

21:17 < kaushik_> rcurtin: hmm, i will read up on some common mappings, (I am familiar with TF-IDF) and then discuss here

21:18 < rcurtin> I think TF-IDF is a good candidate, but it is certainly not the only one that could be chosen

21:18 < rcurtin> (and likely more than one could be implemented)

21:54 < rcurtin> ok... I have the management console on masterblaster and am getting the system up with its new internal IP, but there is still no external IP yet so it will not be publically accessible

21:54 < rcurtin> someone changed all the management credentials on the benchmarking systems (slake/scrooloos/savannah/gekko/skyfish) so I can't get a console on those yet

21:55 < rcurtin> and I think the datacenter did not know how to configure the IP of the management software on the three Sun systems (aunty/ironbar/collector) so those are not accessible at all at the moment

21:56 < zoq> hm, hopefully they can figure it out ...

21:56 < rcurtin> yeah, the biggest issue is the external IPs; if I can get masterblaster back up today then we have like 90% of what we need

21:57 < zoq> right

22:33 < rcurtin> zoq: I made the font for the author tag a little larger, what do you think? http://mlpack.org/docs/mlpack-2.2.5/doxygen/dettutorial.html

22:35 < zoq> rcurtin: Looks good to me.