verne.freenode.net changed the topic of #mlpack to: http://www.mlpack.org/ -- We don't respond instantly... but we will respond. Give it a few minutes. Or hours. -- Channel logs: http://www.mlpack.org/irc/
< rcurtin>
so technically, you could miss both meetings and still see what happened, but I think the IRC meeting is nice because it is as close as we can get to an in-person meeting :)
< rcurtin>
no need to thank---remember you are the ones who put together the good proposals, so really, we should be thanking you I suppose :)
< rcurtin>
anyway, as you've probably already seen, this year we got a record 10 slots
< rcurtin>
so this will certainly be the most busy GSoC mlpack has had yet; I'm looking forward to it
< rcurtin>
everyone should have gotten an email from me with some useful information about the project, some links, and some guidance; if you didn't get one of those, let me know and I'll send it your way
< rcurtin>
as far as schedule goes...
< rcurtin>
right now, we are in the 'Community Bonding' period, which goes from May 4 to May 30
< rcurtin>
ideally, during this time, you get to know your mentor a little bit, get to know the community a little bit, maybe have some fun, maybe learn a bit more about mlpack
< rcurtin>
some nice ways to do this are on the mailing list, or direct emails to your mentor, or here on IRC---don't feel restricted to talk only about mlpack, it is not disallowed to have some fun in the channel too :)
< rcurtin>
in the past we have had some discussions about music recommendations that have been very helpful to my work environment :)
< rcurtin>
once the community bonding period is over, the actual coding goes from may 30 to august 21
< rcurtin>
and during that time, there are two midterm evaluation periods; one at the end of June, and one at the end of July
< rcurtin>
then, at the end of the summer (end of August), there is a final evaluation
< rcurtin>
after that, there is no schedule imposed by Google, but we hope that you will stick around and continue to participate :)
< saurabh>
Ok! to begin with, I like the quotes in the end of your mails. They are really funny š
< rcurtin>
if there are any stipend issues or administrative issues, probably Google will be the right people to ask (I have limited control as an organization administrator), but I can definitely try and help, so definitely speak up
< rcurtin>
thanks!
< rcurtin>
and I realize now, I never introduced myself!
< saurabh>
That makes two of us
< rcurtin>
I'm Ryan Curtin, probably most of you have figured that out by now... I'm the mlpack organization administrator, which is why I am leading this meeting :)
< rcurtin>
but let's save the rest of the introductions until the end, if that's ok
< rcurtin>
I have a few more things to cover:
< kris1>
Can we in addition also start writing code in the Community Bonding Time. At least some broiler pate code
< rcurtin>
kris1: yes, of course, definitely
< rcurtin>
so, I guess, before I move on, I should ask if there are any other questions about the schedule or anything I've talked about so far so we can address them now
< shangtong_>
Just for curiosity, do you know how google decides slots for each org?
< rcurtin>
no, I don't have much insight into that... I would guess they look at the success ratio of previous summers (i.e. how many students passed vs. how many students failed)
< rcurtin>
and they probably factor in the number of mentors and the size of the organization too
< rcurtin>
for comparison, last summer we got 6 slots, and in 2014 we got 5, and only 3 in 2013 (the first year)
< shangtong_>
cool, it's a good trend
< rcurtin>
ok, I'll go ahead and move on now to student and mentor expectations
< rcurtin>
I should say, I don't expect these expectations to be too difficult; we think you are all great students, so there will probably be no problem here
< rcurtin>
but still, it is important to set the groundwork and be clear about these expectations, in case there is any need to revisit this later
< rcurtin>
during the GSoC coding period, students are expected to be doing the equivalent of a full-time job; so, a full work week
< rcurtin>
it's okay if some weeks you work more and some weeks you work less, but in the end the overall amount of time you put in should be about equivalent to what it would be if you had a full-time internship at some company
< rcurtin>
(but unfortunately, unlike a company you might intern at, we don't have any free snacks here :( )
< saurabh>
Do we have a office? If yes, where?
< rcurtin>
regular contact and communication with mentors is very important and expected; preferably, this would be via the public #mlpack channel in IRC, but email or whatever you work out with your mentor is okay too
< rcurtin>
saurabh: no office, I don't think any mlpack developers are in the same place :)
< rcurtin>
I guess both Shangtong and Chenzhe are in Calgary, maybe they can set up an office there or something :)
< rcurtin>
in the past I've generally found IRC to be good for debugging, because it's real-time chat, whereas with email there are typically longer delays for responses
< shangtong_>
no.. in Edmonton
< saurabh>
Thought of visiting in person and have a look. Guess! that is not going to happen :(
< rcurtin>
oh! sorry about that :(
< shangtong_>
hah
< rcurtin>
using the public #mlpack channel is also nice, because it means if your mentor happens to not be around, maybe someone else in the channel can help
< rcurtin>
and definitely there is no problem reaching out for help to someone other than your mentor. we are a community, after all :)
< rcurtin>
students are also expected to provide some kind of weekly status update to the community
< rcurtin>
this could be an email post, like these:
< rcurtin>
whichever format you would like to use is up to you, but weekly updates are important because more in the community than just your mentor may want to read about how the project is going
< rcurtin>
and we can help you get set up with the right permissions to post to there after the meeting
< rcurtin>
next, if you're going to go on vacation or will be unavailable, please inform your mentors in advance
< rcurtin>
disappearing students is a big issue with GSoC projects, so we want to avoid that type of situation :)
< kris1>
can't we just use something like github.io for the blogs
< rcurtin>
kris1: probably, this one is set up with Jenkins. ideally, it is nice to have all of the blogs in the same place for posterity, but you could definitely post the updates to a personal blog also
< rcurtin>
for notifying your mentor, though, there's no need to tell your mentor every time you get up to go to lunch or anything; that level of detail is probably not necessary :)
< rcurtin>
but in either case, you should be in regular (probably daily) contact with your mentor during the project
< rcurtin>
this is so that the mentor can help you overcome any roadblocks or sticking points, so that you can accomplish what you set out in your timeline
< rcurtin>
however, it's definitely true that not every project goes according to the proposal
< rcurtin>
sometimes, the reality of the project is more difficult than what was anticipated
< rcurtin>
but this is okay! don't worry
< rcurtin>
if this type of situation happens, the student and the mentor should work together to adjust the timeline or proposal in accordance with what is feasible
< rcurtin>
we really hope we don't have to fail any students this year; but, we will fail a student who disappears or doesn't appear to be working the expected amount of time, or who is otherwise seriously underperforming
< rcurtin>
if any situation like that occurs, the student will be made fully aware with warnings, so any failure will not be unexpected
< rcurtin>
personally, I really don't expect this to be an issue, and it hasn't been in the past, but I think it's really important to make that all clear before we get started
< rcurtin>
any questions about student expectations before I move on to the related mentor expectations? I know that was a lot of text :)
< kris1>
Just one small question. For contacting the mentor, can we use #mlpack channel or should we do it over email or pm to the mentor.
< rcurtin>
kris1: you can work that out with your mentor and decide what you like best
< rcurtin>
an advantage of #mlpack is that other people can also help with the questions
< rcurtin>
but there have definitely been students over the years who prefer email communication, and that's okay too
< kris1>
yeah. That's what i would prefer also
< pretorium>
If i have already started coding my poject, when should I present report - weekly or in the beginning of coding period, or otherwise?
< rcurtin>
kris1: ok, sounds good, you should talk with Mikhail and work out what you will do
< rcurtin>
pretorium: if you like you could send "early" status reports, but don't feel obligated. during the coding period, you should aim for one a week
< kris1>
yes, will do that
< rcurtin>
if you didn't want to send an "early" status report, then you could have a really big first week status report or something like that, where you can update on everything you have done so far
< lozhnikov>
kris1: It's ok if you prefer emails and so on
< ironstark>
can that first week report also include the work we may have already done even before submitting our proposals?
< rcurtin>
ironstark: sure, if you like, that's okay. the status update emails are for the benefit of the community, so explaining what you have been up to is always helpful :)
< ironstark>
okay :)
< kris1>
lozhnikov: Actually i would prefer the irc channel. But i am also okay with email.
< rcurtin>
I'll go ahead and move on to mentor expectations now; this is a bit shorter :)
< rcurtin>
mentors should work with their students to determine times that they are both available regularly to work with each other, and they should be willing to help debug code problems and help the student understand theory as needed
< rcurtin>
similar to how students should be responsive to their mentors, mentors should also be responsive to their students, providing reasonably quick replies
< lozhnikov>
kris1: oh, I misunderstood, I was a little inattantive:)
< rcurtin>
the mentor shouldn't do the majority of the work of the project, of course, but, they should be there to help when needed
< rcurtin>
mentors should also give their students advance notice if they are going to be unavailable
< rcurtin>
in general I think it's a good idea for GSoC communication to be done in public; like I said earlier, this is so that many people can answer a student's questions
< rcurtin>
but that isn't necessarily applicable to every case, so that is not a strict requirement of course
< rcurtin>
the two student midterm evaluations and the final evaluation are done by the mentor (not the organization administrator)
< rcurtin>
so when that period happens, mentors will need to fill out the evaluations (don't worry, it is easy)
< rcurtin>
also, in order to get the Github accounts all set up right, I'll need to collect the github accounts everyone would like to use (I think I already have all the mentors, but not all the students)
< rcurtin>
so if you would like to each say it here or email me or something, I can get that set up right :)
< rcurtin>
I think that does it for the expectations; does anyone have any questions, or is there anything anyone would like some clarification on?
< rcurtin>
ok, added everybody to the 'GSoC Students' team; I think each of you will have to accept the invitation that Github sends
< rcurtin>
if there aren't any questions, I'll move on to some quick history of mlpack, and maybe some predictions about the future directions, and then I will be done :)
< rcurtin>
feel free to interrupt with any questions if I didn't wait long enough
< rcurtin>
so, the history (I'll keep it short)
< rcurtin>
mlpack was first created in 2007 in a lab at Georgia Tech
< rcurtin>
the lab had maybe ~10 people regularly contributing to the library over 2008-2009 (you can look at the github history, it's all there still), and then I joined the lab in late 2009
< rcurtin>
the lab, which was called the "FASTLab", was led by Alex Gray, who specialized in tree-based algorithms
< rcurtin>
I was brought on in part to make the library more consistent and then finally release it as an open source product
< rcurtin>
but... there was a lot of work to be done, because the library definitely was not ready for release
< rcurtin>
so myself and a team of a couple others in the lab worked on refactoring it through all of 2010 and 2011 and finally released 1.0.0 at the end of 2011 at a NIPS workshop
< rcurtin>
the FASTLab itself ended up shutting down when Alex Gray left to found a company called Skytree (which actually just got acquired by Infosys so I guess it no longer exists)
< rcurtin>
and it was about at this time that mlpack started getting involved in Google Summer of Code, in 2013
< rcurtin>
since then, the project has grown a lot, and now we have something like 80+ contributors
< rcurtin>
I think there is a contributor from every continent except Antarctica
< rcurtin>
and we even have some contributions that were done in North Korea at a university there
< rcurtin>
we also even have some contributions from a deep learning system intended to detect bugs (https://github.com/C0deAi)
< rcurtin>
watching the growth of the community over these years has been really exciting, and it is also really cool that when I travel to conferences now, people have heard of mlpack and sometimes have even used it
< shangtong_>
Is there any special relationship between mlpack and armadillo, I notice you are also a contributor of armadillo :)
< rcurtin>
shangtong_: yeah, we have contributed stuff back upstream to Armadillo for many years now, including the sparse matrix support
< rcurtin>
so Conrad is in regular contact, although somehow after all these years I still have yet to meet him at any conference or anything :)
< rcurtin>
in the future, there will be a GPU version of Armadillo, so I think that is an exciting thing that could also be used in mlpack (hopefully without too many code changes)
< rcurtin>
also in the future, the neural network code (which many of the GSoC projects are about) will become stable and be released as part of mlpack 3.0.0
< rcurtin>
and I'm very excited about that---I think there will be a lot of interest and adoption when it is announced
< rcurtin>
similarly, I'm currently working on Python bindings and bindings for other languages (nearly done!), and I think this too will have a big impact when those are released
< mentekid>
wow that video is cool and I have no idea what it shows :)
< rcurtin>
and I think each of the projects that have been accepted have a lot of potential to provide neat new support, and I look forward to talking about it to anyone who will listen :)
< rcurtin>
the benchmarking system that has been a part of GSoC projects since 2013 also has some great potential for nice visualizations: http://mlpack.org/benchmarks.html is an out of date example
< rcurtin>
which brings me to the next bit, the infrastructure---through my work at Symantec, I have obtained a pretty hefty amount of computing power that is currently being used as our build and test servers
< rcurtin>
the system at http://masterblaster.mlpack.org/ has 72 cores and I think 256GB of RAM, so it can build mlpack with lots of configurations
< rcurtin>
since we have lots of infrastructure like this, if anyone has resource limits locally or needs a larger system to run mlpack code on, just let me know
< kris1>
Wow 256GB of RAM!!!!!
< rcurtin>
it's a little tedious to set up new accounts, but it can be done if needed
< rcurtin>
kris1: yeah, it surprised me very much when I joined the company that they have so many resources available for use
< rcurtin>
I can't think of anything else to say, so if there aren't any questions or clarifications I can give, I guess now is a good time for introductions so we can get to know each other a little bit :)
< rcurtin>
there is no framework in place for us to decide who introduces themselves next so I think it may be a little chaotic, but I will start :)
< shangtong_>
so is mlpack an official tool of Symantec?
< rcurtin>
like I already said, I'm Ryan; I went to Georgia Tech for about 10 years and got a PhD, and now I work at Symantec Corporation on applying machine learning to security problems (and also on getting them to use mlpack!)
< kris1>
Yeah i think big companies do have a lot resources available. Just some days in ICLR a presentation said used 800 gpu's for computation.
< rcurtin>
shangtong_: it's used internally a little bit (I hope to make that more soon!), and they fund the development
< rcurtin>
so I guess I should put a Symantec logo on the about page of the website
< shangtong_>
awesome
< rcurtin>
but they don't necessarily exert control over the direction of development or anything like this
< rcurtin>
in my free time, I like to pursue some hobbies; one of my favorites is to race go-karts. here's a picture:
< kris1>
Haha that an amazing t-shirt ...............
< saurabh>
haha great!
< shikhar>
:D
< rcurtin>
the version I got is actually a nice breathable material, so it's perfect to keep cool while racing :)
< zoq>
Okay my turn.
< zoq>
My name is Marcus and I live in Berlin, Germany; Iām a PhD student at the Free University of Berlin focusing on automated theorem proving using neural networks like the four color theorem.
< zoq>
When I'm not working, I like to fly not with a plane, although that would also be quite cool, but with a gliding parachute:
< ironstark>
would like to try that someday if i get a chance
< zoq>
if you wonder what this cable is all about; it's called winch launch, it's a simple way to make launches possible without having a mountain.
< saurabh>
Adventure sports š± amazing
< rcurtin>
ah finally, a picture of the paragliding :)
< kris1>
looks really cool
< zoq>
I guess that's from my side.
< saurabh>
I'll go next
< mentekid>
"I race cars" and "I paraglide" must be two of the coolest answers to "What do you do in your free time?"
< saurabh>
I am Saurabh (as the nick suggests) from India, a final year student of graduation in Information technology and mathematics. As zoq, I like adventure sports and wandering mountains a lot
< rcurtin>
does the river look that green in real life, or is that an artifact of the picture?
< saurabh>
I have much more, will save it for the next meetings :)
< saurabh>
Okay, that's a filter I guess :)
< rcurtin>
ah, ok :)
< pretorium>
I think I'll go next
< pretorium>
I'm Konstantin Sidorov from Astrakhan, Russia. I am a first-year student in Software Engineering. I have less adventurous hobby - in my free time, I play Go (the ancient Chinese game - not the programming language)
< shangtong_>
As a Chinese.. I don't know how to play GO ...
< pretorium>
shangtong_: If that matters, I also study Chinese language :)
< kris1>
Wao i Chinese like a type-5 language right
< rcurtin>
I have never played Go, maybe I should learn :)
< zoq>
I can "play" chess, but I guess go is way more complicated.
< kris1>
GO and Chinese wao!!!
< kris1>
Yup GO is exponentially harder than chess
< saurabh>
zoq, we can schedule a one to one sometime :)
< pretorium>
zoq: not quite, because in Go you have only one type of units
< kris1>
I will go next if that's okay.
< zoq>
saurabh: chess or go?
< mentekid>
I played the linux implementation of the game once and got my ass handed to me by the computer
< mentekid>
I only realized it's just that I suck at it when AlphaGo became a thing. Up until then I thought it was just another game computers are better at
< saurabh>
zoq: chess :)
< kris1>
I am kris from india. A masters student in ML. I do like to travel a lot and have been to almost 14 states in india. Sorry i didn't bring a picture maybe next time.
< zoq>
saurabh: Sure, great idea!
< saurabh>
zoq: haha (evil laugh) yes
< shikhar>
I guess I'll introduce myself now.
< ironstark>
okay I'll go next :My name is Dewang Sultania and I am a third year undergrad student at Birla Institute of Technology,India pursuing Engineering in Computer Science.My hobbies include Dancing and trying out different adventure sports(whenever I get a chance).Here's a pic of me rafting with my cousins. https://ibb.co/c3VhY5
< saurabh>
I went for rafting and cliff jumping recently. It's fun š
< ironstark>
yaa it is
< rcurtin>
ironstark: are you steering the boat there? :)
< ironstark>
No
< shikhar>
I am a sophomore year undergraduate student from Delhi, India, studying Mathematics and Computing Engineering at DTU. In my leisure time, besides other things, I like to tinker with electronics. Everything from "blink an LED with a 555 timer" to "set up a microcontroller with few sensors".
< ironstark>
i am the one at the front seat
< shikhar>
Guess I'll bring a photo next time of something I built(if it exists as a single entity) :)
< shangtong_>
I guess I can go next :) I am Shangtong, a first year master student at Canada. My primary interest is reinforcement learning.
< ironstark>
shikar: That will be a cool pic to see
< shangtong_>
My hobby is marathon
< shangtong_>
I completed a half marathon one year ago and now preparing for a full marathon this summer
< shangtong_>
unfortunately ... it seems I don't have recent picture ..
< saurabh>
How many miles?
< shangtong_>
21 km for half marathon
< shangtong_>
43 for full
< shangtong_>
*42
< ironstark>
i dont think i could even walk that many miles
< ironstark>
š
< mentekid>
Wow that is great, good luck
< shangtong_>
:)
< mentekid>
Yeah I am against movement if it's not absolutely necessary too...
< ironstark>
haha
< shikhar>
Haha
< kris1>
mentekid: same here
< ironstark>
same here
< saurabh>
I used to run for 10 km. I can relate how much effort it takes . Kudos to you and good luck
< rcurtin>
I prefer motorized movement; I am not good at actual running :)
< mentekid>
who's left? Should I go next?
< zoq>
shangtong_: My next local half marathon is next month :)
< rcurtin>
I have to run to lunch now, it was great talking with all of you :)
< shangtong_>
zoq: awesome!
< rcurtin>
if you want to sit through this meeting again, come back at 0000 UTC :)
< shikhar>
Thanks a lot for the time!
< rcurtin>
welcome to the community everyone :)
< ironstark>
Thanks a lot :)
< rcurtin>
I'll be back in ~60-90mins
< saurabh>
Nice to meet you all š good bye for now
< kris1>
Thanks for a great intro to the community. :)
< mentekid>
I'll try to make the next one too, but I have a flight too catch
pretorium has quit [Quit: Page closed]
< mentekid>
Nice to meet you all for now, looking forward to a great summer :)
< ironstark>
Nice to meet everyone. :-)
< kris1>
Great to meet you all. This was really great session. Bye for now :)
< zoq>
Yeah, nice to meet all of you :)
< shangtong_>
Nice to all of you here
< zoq>
If anybody has some music recommendations, I'm here :)
mentekid has quit [Quit: Leaving.]
< zoq>
Right now I'm listening to The Glitch Mob - We can make the world stop
< kris1>
zoq: Gorillaz is great for coding late night.
< saurabh>
Rockabye has great lyrics
< zoq>
Cool, thanks queued :)
shangtong_ has quit [Quit: Page closed]
< shikhar>
zoq: Depth Charge by Flume
govg has joined #mlpack
< kris1>
shikhar: what about nucleya :)
kris1 has left #mlpack []
< zoq>
I like Flume's new EP
shikhar has quit [Quit: WeeChat 1.4]
mentekid has joined #mlpack
mikeling has quit [Quit: Connection closed for inactivity]
mentekid has quit [Quit: Leaving.]
govg has quit [Ping timeout: 240 seconds]
govg has joined #mlpack
< rcurtin>
ack, sorry everybody, I'll invite students to the 'GSoC 2017 Students' team on Github, so everyone who have me their github ID will get two invitations, sorry about that :)
chenzhe has joined #mlpack
< zoq>
rcurtin: If you like I can set up the jenkins job, should be easy.
< rcurtin>
sure, if you want to go ahead and do it, I have a few other things on my checklist first today :)
shikhar has joined #mlpack
< zoq>
okay, I guess once the code is in the jenkins-conf repo it should be ready