04:13 < AakashkaushikGit> Hey, did you guys hear about this: https://github.com/flashlight/flashlight

04:16 < AakashkaushikGit> Just wanted to have your thoughts on this, seems interesting to me, i didn't completely explore it but seems like something that can match mlpack and the way they tout the small re compile times after the code change maybe it does compiles pretty efficiently.

15:43 < rcurtin[m]> I saw it---my opinion is that it's great to see more C++ machine learning libraries. looks like this one is specific to deep learning type applications, if I read right

15:44 < rcurtin[m]> very early on (like 2000-2008), C and C++ were probably the most popular languages for machine learning libraries, but with the rise of Python and how easy it was to prototype applications in Python, that changed; but personally, I've always thought that the speed benefits of C++ outweigh the prototyping benefits of Python, and when you need to take your prototype into production, it's a million times easier to do that if

15:44 < rcurtin[m]> your prototype was already written in C++

15:45 < rcurtin[m]> but, the ecosystem for C++ machine learning has been kind of limited for the past decade or so, so if flashlight is in C++ and draws people to C++, I think that's a good thing :)

15:50 < shrit[m]> When it comes to my colleagues that they do more theoretical machine learning about this discussion, they often tell me that it is really hard to find someone especially in research how has perfect knowledge in theoretical mathematics, its application to machine learning, programming languages such as C++ and algorithms, therefore they chose python since their objective is to prove the idea, and writing code in python is so

15:50 < shrit[m]> simple, that even if your knowledge in computer science is very limited you can create a great thing with.

15:53 < shrit[m]> I only learned to write code in C/C++ for a long time, I encountered python very late. Now, It is very hard for me to reason in Python, Even when I write in Python it feels like a modified C++.

16:43 < rcurtin[m]> ok... going to try to upgrade the gitter bridge now

16:43 < rcurtin[m]> sorry to any gitter users who get disconnected for a bit :)

16:48 < rcurtin[m]> actually, nevermind, it looks like I need to email gitter support to get them to upgrade this correctly

18:01 < AakashkaushikGit> @rcurtin yes it was built to be more comparable to torch and tensorflow.

18:01 < AakashkaushikGit> Also is the issue with gitter fixed, as this is my only source to irc.

18:05 < rcurtin[m]> Aakash kaushik (Gitter): yeah, the gitter bridge is still up---it turns out to transition to the new way Gitter wants to do things, I just had to email support... so now I'm waiting on their response; probably will hear back in a day or two

19:16 < AakashkaushikGit> Thank you for the update @rcurtin.