05:01 < zoq> sourabhvarshney1: Sorry for the slow response, somehow I missed the message, you are right, it's forwarding a method given a type.

05:03 < zoq> sourabhvarshney1: Do you think instead of implementing a new visitor, we could reuse the existing, e.g. for the backward pass instead of providing a vistitor that takes two deltas we could call the existing backward visitor twice, once for each layer.

05:17 < rcurtin> wenhao: sorry for the slow response, I spent all day trying to fix a pinball machine :)

05:17 < rcurtin> there is no transform() or for_each() for sparse matrices; however, you could use sp_mat::iterator to loop over nonzero values

05:18 < rcurtin> Sumedh and I also wrote row_col_iterator for both arma::mat and arma::sp_mat, so if you are writing a function that could support both

05:18 < rcurtin> or to clarify I mean if you are writing a function that will support both dense and sparse matrices via a template parameter or something

05:19 < rcurtin> then you could use row_col_iterator, which for dense matrices will iterate over all elements

05:19 < rcurtin> and for sparse matrices will iterate over nonzero elements

05:19 < rcurtin> manish7294: hmm, maybe there is a bug or something. it could also be worth comparing with a reference implementation

05:20 < rcurtin> if you don't have success debugging, I can try and dig in also, but unfortunately it will have to be Tuesday at the earliest---tomorrow I am out all day, and Monday is a long all-day meeting :(

10:02 < jenkins-mlpack> Yippee, build fixed!

10:02 < jenkins-mlpack> Project docker mlpack nightly build build #324: FIXED in 2 hr 48 min: http://masterblaster.mlpack.org/job/docker%20mlpack%20nightly%20build/324/

11:03 < wenhao> rcurtin: Thanks for clarification! I am writing a function to loop over all non-zero values in a sp_mat so I think sp_mat::iterator will be good

15:23 < zoq> wenhao: Really interesting blog post, btw. it looks like that the current implementation expects two words for the author tag, we can just add a single character to solve the issue.

15:25 < zoq> ShikharJ: Really interesting blog post as well, really really productive week :)

15:31 < sourabhvarshney1> zoq: Wouldn't that mean that I have to do forward pass and then backward pass for the front direction and then for the back direction. But that will be against the ideas of the algorithm as the backward pass uses output parameter which is dependent on both front and back nodes.

15:34 < zoq> haritha1313: All this nice and interesting updates, thanks, really nice progress :)

15:37 < zoq> sourabhvarshney1: Good point, I guess in this case it would be great to write another visitor.

15:37 < sourabhvarshney1> zoq: Thanks! I would do that.

18:46 < travis-ci> manish7294/mlpack#1 (lmnn - fe59fe6 : Manish): The build has errored.

18:46 < travis-ci> Change view : https://github.com/manish7294/mlpack/commit/fe59fe6f37cd

18:46 < travis-ci> Build details : https://travis-ci.org/manish7294/mlpack/builds/381382558

18:49 < travis-ci> manish7294/mlpack#4 (lmnn - fe59fe6 : Manish): The build has errored.

18:49 < travis-ci> Change view : https://github.com/manish7294/mlpack/commit/fe59fe6f37cd

18:49 < travis-ci> Build details : https://travis-ci.com/manish7294/mlpack/builds/73922560