01:10 < jenkins-mlpack2> Project mlpack - git commit test build #584: STILL FAILING in 1 hr 35 min: http://ci.mlpack.org/job/mlpack%20-%20git%20commit%20test/584/

03:31 < taqd> https://github.com/mlpack/mlpack/pull/2718 :) Not sure if it's a bug, or I am doing something wrong, but I put in a pull request nonetheless

05:43 < taqd> I was wrong, hah!

11:14 < yashwants19[m]> I guess scheduled cron workflow is working fine.

15:29 < RishabhGarg108Gi> Hii, I am new to C++ design patterns. Can someone clarify that the "visitors" in `methods/ann/visitor` corresponds to "visitor design patterrn" or is it something else ?

15:37 < rcurtin> RishabhGarg108Gi: yep, that's correct; we are using the boost::visitor implementation

15:37 < rcurtin> taqd: glad you got it worked out! thanks for opening the PR even though that turned out to not be a bug

15:38 < rcurtin> could we improve our documentation or error messages to make it easier for you to have found out that the labels needed to be transposed?

15:43 < RishabhGarg108Gi> Okay. Thanks @rcurtin !

16:03 < rcurtin> so inside of my company I made bindings in our database product to mlpack, and I just got some really nice feedback yesterday in our public "#appreciation" channel:

16:03 < rcurtin> > Big appreciation to @Ryan Curtin for a very comprehensive set of ML algorithms and a very easy interface to use in Delve through his bindings to mlpack

16:04 < rcurtin> but really that appreciation should go to everyone here who makes all those algorithms possible :)

16:04 < zoq> Ohh that is great to hear.

16:05 < rcurtin> at the current moment, mlpack has a lot of exposure inside the company since it is the only machine learning toolkit accessible from our database language

16:09 < jeffin143[m]> rcurtin: not sure if I have asked you aldeardy ?

16:09 < jeffin143[m]> So you are making a db which has some intelligence ??

16:12 < rcurtin> jeffin143[m]: sort of... the quick "pitch" is this... most data scientists currently extract their data from the database into CSVs or something, then load it into their machine learning environment (could be Python, C++, R, whatever) and do their machine learning

16:12 < rcurtin> but many times, just the cost of getting the data out of the database could be a lot; it could be a complex join of some sort that is really expensive

16:13 < rcurtin> so there are some cases in which, it turns out, you could learn the machine learning algorithm directly on the underlying database tables *without* performing that expensive join, and this can give orders-of-magnitude speedups

16:13 < jeffin143[m]> Woohhhh that was nice

16:14 < rcurtin> so the idea is that we will build a database that does this, and we chose to make a new query language (not SQL, more like Datalog if you have ever heard of it---I hadn't :))

16:14 < rcurtin> the hope is that in this query language, it's very easy to express machine learning algorithms

16:14 < rcurtin> but for now, we only have mlpack bindings :)

16:17 < abernauer[m]> rcurtin: Is it going to be a DSL? I like the idea though seems like a great way to avoid that common workflow setback.

16:19 < rcurtin> abernauer[m]: not a DSL, actually a completely new query language

16:22 < abernauer[m]> Well sounds awesome and congrats on the appreciative feedback. Sounds like it addresses a real need for data scientists that's not currently being addressed.

16:22 < rcurtin> let's hope :) it's a very hard product to build since it's so interdisciplinary

16:23 < rcurtin> so I have to work with lots of database and compiler people, and we all speak completely different languages

16:23 < rcurtin> not only that but even our implementation language is foreign to many of us... we are writing the product entirely in Julia, but very few of us are "Julia native"

16:27 < abernauer[m]> <rcurtin "not only that but even our imple"> Yeah that sounds like one of the more difficult challenges. Personally, whenever I learn or work in a language try to write code that's idiomatic to that language.

16:29 < rcurtin> yeah, so interestingly, we get to define what "idiomatic" means, and that is one of the big discussion points in the company

16:29 < rcurtin> if someone writes a query, what is the "right" way to have written that query that we want to encourage people to use?

16:36 < abernauer[m]> Yeah I would think that get's complicated or opinionated as you're all coming from different backgrounds as mentioned. Sounds like a fun challenge though :')

16:40 < rcurtin> opinionated for sure :-D

16:56 < jeffin143[m]> > so the idea is that we will build a database that does this, and we chose to make a new query language (not SQL, more like Datalog if you have ever heard of it---I hadn't :))

16:56 < jeffin143[m]> @rcurtin:matrix.org since I some what work in cloud space so I have seen datadog and have worked with it

16:57 < rcurtin> oh no I mean 'datalog' not 'datadog' :)

16:57 < rcurtin> https://en.wikipedia.org/wiki/Datalog

16:57 < jeffin143[m]> Oh then I haven't

16:57 < jeffin143[m]> 😂

16:57 < jeffin143[m]> Haha

16:57 < rcurtin> :)

17:40 < AnushKiniGitter[> Hey guys, is there any way to restart the pipeline checks on a PR without making a commit?

17:53 < rcurtin> I think you can close and open the PR and that will re-trigger them

18:04 < AnushKiniGitter[> Thanks! Will try it out.

23:17 < taqd> Hi again, in mlpack_perceptron (and elsewhere) the code references "labels(i,j)", "predictedLabels(i,j)" and "instanceWeights(i,j)" and for the life of me I can't find where they are defined. Any suggestions?

23:19 < taqd> frig, nevermind