02:32 < jjb[m]> rcurtin when you have a moment can you comment on the documentation site note: <https://github.com/mlpack/mlpack/pull/2633> ?

08:46 < AakashkaushikGi4> Hey everyone, the article got published [here](https://medium.com/syntechx/convolutional-neural-network-cnn-in-c-52c9ed47a6ea), maybe give it a look.

11:11 < say4n> rcurtin: yes, totally!

12:25 < rcurtin> jjb[m]: sorry for the slow response, I hadn't managed to get to it---answering now

13:51 < zoq> AakashkaushikGi4: Cool :)

14:18 < travis-ci> shrit/examples#89 (rl_cpp - 9778dfb : Omar Shrit): The build passed.

14:18 < travis-ci> Change view : https://github.com/shrit/examples/compare/45e2027519f2^...9778dfb743ff

14:18 < travis-ci> Build details : https://travis-ci.com/shrit/examples/builds/186882046

16:48 < shrit[m]> Aakash kaushik (Gitter): Sounds cool, very good article, it would be great if we have a similar tutorial either in`mlpack/doc` or in the example repository.

16:55 < AakashkaushikGi4> > `shrit` Aakash kaushik (Gitter): Sounds cool, very good article, it would be great if we have a similar tutorial either in`mlpack/doc` or in the example repository.

16:55 < AakashkaushikGi4> Hey, thanks a lot, I think i will be able to help with that if you we can have a guideline on how the examples should be, because i was free to write things in the article but i think we need to have some guidelines here so everyone can contribute.

16:55 < AakashkaushikGi4> (edited) ... if you we can have a guideline on ... => ... if we can have guidelines on ...

17:08 < AakashkaushikGi4> Also i have completed the mean_shift_test and the loss_functions_test.cpp so is it okay if i create the PRs on October 1st for the hacktoberfest ?

17:34 < AakashkaushikGi4> for the `BOOST_REQUIRE_CLOSE_FRACTION(x,y,z)` what should be the equivalent catch2 expression according to [this](https://stackoverflow.com/questions/1093453/difference-between-boost-check-close-and-boost-check-close-fraction) we can go with a rule of thumb such as if `z = 1e-4` in boost we can convert the expression to `REQUIRE(x == Approx(y).epsilon(1e-8))` basically doubling the decimal points.

17:34 < AakashkaushikGi4> Please let me know if i am wrong with this

17:36 < AakashkaushikGi4> Oh wait i think doubling it is wrong and increasing the decimal point by 2 will work such as if it was `1e-4` then `1e-6` may work ?

19:41 < shrit[m]> Aakash kaushik (Gitter): What guidelines are you thinking of ? Do you have specific ideas? Regarding the article you have wrote, you can provide more details in a tutorial to explain more specifically we are doing each lines, You can imagine the case of a user with only a theoretical background in machine learning and want to make a good use if this code.

21:59 < zoq> AakashkaushikGi4: Yes, 1e-6 should work.

22:13 < zoq> rcurtin: Working on another example for K-means, currently using the pima indians diabetes dataset, I plot the optimization steps - http://data.kurg.org/pima-indians-diabetes.gif. The cluster are almost fixed after 10 iterations, off the top of your head to you know any dataset that might be more interesting?

23:50 < rcurtin> ahh, unfortunately, in 2 dimensions, k-means usually converges really fast

23:51 < rcurtin> also, usually it is only the first ~5-10 iterations (even in high dimensions) that see most of the cluster movement

23:51 < rcurtin> personally I think that diabetes visualization is pretty cool---it's way cooler than, e.g., just k-means on two gaussians or something