04:00 < bhuwnesh> For the decision tree section are the implementations of bagged decision trees, random forest and boosted decision trees required?

11:38 < namratamukhija> fatal error: mlpack/core/util/mlpack_main.hpp: No such file or directory compilation terminated. I checked the core/util folder and the file exists. Am I missing something?

11:40 < namratamukhija> Sorry, some copy paste error on my end.

11:42 < namratamukhija> The issue I am facing is this: I wrote some binding test cases for approx KFN. I am trying to compile them with the command: g++ approx_kfn_test.cpp -o object -std=c++11 -lmlpack -larmadillo -lboost_serialization -lboost_program_options -lboost_unit_test_framework. I get an error: fatal error: mlpack/core/util/mlpack_main.hpp: No such file or directory compilation terminated.

11:42 < namratamukhija> I checked the core/util folder and the file exists. Am I missing something?

11:54 < zoq> namratamukh: It's a lot easier to use cmake + make to build mlpack.

12:01 < bhuwnesh> zoq: For the decision tree section are the implementations of bagged decision trees, random forest and boosted decision trees required?

12:08 < zoq> bhuwnesh: Decision tree section? not sure what you mean, I guess I missed something.

12:13 < bhuwnesh> I mean that for the decision trees can I implement the above mentioned algorithms along with CART and C 4.5 with few changes.

12:19 < daivik> zoq: Are desai-aditya and Manthan-R-Sheth still working on the ANN input/output size issue? If not, then I'd like to take it up please.

13:10 < zoq> bhuwnesh: I see, if you like to work on that please feel free, I think the existing random forest is a really good basis.

13:11 < zoq> daivik: It might be a good idea to ask on the issue itself.

13:56 < manish7294> rcurtin: If you get any time to spare then please assess my proposal.

14:10 < rcurtin> manish7294: I actually glanced at it yesterday because I was interested to see how many proposals had already been submitted; I think it is clear and put together well

14:11 < rcurtin> my only thought was, it's not guaranteed that LMNN+LRSDP will work, so probably ensuring that LMNN can work with the other SDP solvers we have would be a good first test

14:11 < rcurtin> but I am not sure I looked closely enough to see if that detail was already there

14:16 < manish7294> rcurtin: Thanks for reviewing it. Your concern is up to the point as currently there is no work in this regard. Currently, I have put together a standard LRSDP form for the same and hoping that it will work.

14:17 < manish7294> I will try checking it with other SDP solvers in the meantime.

14:18 < rcurtin> keep in mind we do have a primal-dual solver, so I think it would not be too hard to implement the SDP problem itself for LMNN and then use any of the solvers

14:41 < Atharva> rcurtin: Does it matter how early the proposal is submitted before 27th?

14:53 < zoq> Atharva: No, but don't expect feedback on a proposal submitted days before the deadline; the review process comes on top of the things we are trying to manage right now.

14:55 < Atharva> zoq: Okay, so can I get feedback even after I have uploaded the proposal on the portal?

14:56 < rcurtin> Atharva: we can try and give feedback if we have time, but I don't think we can make any guarantees

14:56 < zoq> Right, also we will only review proposals that are accessible via the dashboard, I think you have to mark your proposal as final at the end.

14:56 < rcurtin> typically the most important things for a strong proposal is to make sure that the timeline is clear, to make sure the work you are proposing to do is clear, and to make sure that whatever you are proposing will be a useful and user-accessible part of the library when it is done

14:58 < Atharva> Okay, thank you, I will upload my proposal by this weekend, about 17th March, I hope it isn't very late.

14:59 < Atharva> Also, will only the mentors of that particular project review it?

15:05 < ShikharJ> zoq: I submitted the final pdf, in case I forgot to upload it at the end of the submission period. Does that mean that the draft won't be available for feedback?

15:07 < zoq> ShikharJ: let me check.

15:09 < zoq> ShikharJ: I can access the draft.

15:12 < manish7294> ShikharJ: You can try updating the final pdf, If it's working then it's all good otherwise you may not be able to update your proposal any further.

15:18 < zoq> Atharva: No, but I think each reviewer will be somewhat familiar with the topic.

16:57 < thepro> hello

16:58 < zoq> thepro: Hello there!

16:58 < thepro> the implementation of the decision tree says it is the generic tree classifier. Is generic referring to ID3 algorithm?

17:00 < zoq> thepro: C4.5, https://github.com/mlpack/mlpack/pull/858

17:01 < thepro> thanks

17:16 < thepro> do mlpack have an existing implementation for ID3?

17:20 < rcurtin> thepro: ID3 was originally written for only categorical attributes in the paper, so really what I would say is that mlpack's implementation closely resembles the later C4.5 tree building strategy with a few optimizations and minor changes

17:20 < rcurtin> note also that C4.5 is a newer version of ID3, so the two are closely related

18:51 < Prabhat-IIT> zoq: Please can you look at the doc I forwarded you today because I can then start framing my proposal as I would like to finalize exactly what we can use in the implementation of PSO

18:52 < Prabhat-IIT> As soon as you get time please take a look :)