02:12 < rcurtin[m]> unfortunately I didn't see it either; by the time I made it outside, Jupiter and Saturn had already set :(

03:49 < AyushSingh[m]> zoq are you referring to adding seq2seq model to the models repository?

04:06 < GauravGhati[m]> Hey ryan I opened a pr few

04:08 < GauravGhati[m]> Hey #2769 few days back. please have a look whenever you have time.

12:31 < zoq> AyushSingh[m]: Yes, that's what I had in mind.

12:42 < Samyak> Hi, I am working on parallelization of algorithms implemented in mlpack using MPI(https://en.wikipedia.org/wiki/Message_Passing_Interface) It would give improvement in runtime performance. Has anyone else worked on parallelization before?

12:46 < tarunjarvis5Gitt> I am new to open source I was suggested to go through the goodfirstissue but as I am new I can hardly understand it can someone walk me through

12:46 < rcurtin[m]> Hi Samyak, I would recommend against using MPI directly. It will make the implementations too complex. Plus, mlpack is not really designed for clusters or situations where there are multiple distinct systems. I would suggest focusing on OpenMP because even after OpenMP support is added, the code is still understandable to people who do not know OpenMP well (that is definitely not the case with MPI)

12:59 < Samyak> MPI gave huge performance improvement on my machine. MPI would use all cores of a machine whereas OpenMP uses multithreading in a shared memory setting(single core). I would provide proper documentation wherever MPI is used.

13:01 < rcurtin[m]> A shared memory setting is the setting that the vast majority of people will run mlpack in: single node, multicore. mlpack is not really built for the multinode setting, and even if you did adapt some of the algorithms to use MPI, often the algorithmic strategy for machine learning algorithms in a multinode context needs to be vastly different due to communication overhead. If you are looking to implement MPI-based

13:01 < rcurtin[m]> machine learning algorithms, perhaps finding an explicitly distributed machine learning library might be a better choice?

14:11 < tarunjarvis5> I am new to open source I was suggested to go through the goodfirstissue but as I am new I am having difficulty understanding it can someone walk me through

14:55 < zoq[m]> Connection lost...

14:55 < rcurtin[m]> oops :(

14:57 < _slack_mlpack_U0> What happened 😂

15:02 < AyushSingh[m]> tarunjarvis5 - You can select the issue of your choice, read more about it(it's conversation thread and also through external sources), look in the thread if any PR related to it is merged, get a basic understanding of what is being done over there, look how people have contributed to that PR in the past and what all is left to be done and how can you contribute in it.

15:05 < AyushSingh[m]> zoq , okay, I will look into how to implement it.

15:09 < zoq> rcurtin[m]: About the NF monthly update, maybe it makes sense to put that on git so we can add things, and just take what's there once Walter asked.

15:18 < rcurtin[m]> zoq: sure, that seems like a great idea; where should we put it?

15:18 < rcurtin[m]> we could just take HISTORY.md perhaps, but that won't catch any PRs that aren't yet merged or things that don't have to do with the codebase

15:19 < zoq> HISTORY.md also only covers mlpack/mlpack

15:19 < zoq> I guess a wiki page or some new repo works?

15:20 < rcurtin[m]> yeah, I suppose either would be just fine

19:09 < zoq[m]> Hello Abhishek, there are two pages that should help you get started - <https://www.mlpack.org/community.html> and <https://www.mlpack.org/gsoc.html>.

19:09 < zoq[m]> Let us know if there is anything we can clarify.

22:26 < rcurtin[m]> @dkipke interesting. DBSCAN will sometimes not assign points to clusters---and as a result the returned assignment will be SIZE_MAX (or its equivalent in Go); do you think that is what is happening with your data?

22:49 < rcurtin[m]> hmm... what happens if you try to scale k-means or mean shift in the same way?