cameron.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/
gtank has joined #mlpack
govg has quit [Quit: Lost terminal]
govg has joined #mlpack
Rishabh has joined #mlpack
< Rishabh>
Hello, i was looking around and saw this idea about automatic bindings
< Rishabh>
has anyone worked on this??? other than the matlab version
< Rishabh>
or if you could give me some idea upon how to implement this for say python
< Rishabh>
i would like to work on this as a project, maybe GSoC maybe sooner than that
< Rishabh>
thanks
AndChat-394809 has joined #mlpack
AndChat-394809 has quit [Read error: Connection reset by peer]
govg has quit [Read error: Connection reset by peer]
govg has joined #mlpack
Rishabh has quit [Quit: Page closed]
< jenkins-mlpack>
Starting build #2145 for job mlpack - svn checkin test (previous build: UNSTABLE -- last SUCCESS #2143 2 days 11 hr ago)
< jenkins-mlpack>
michaelfox99: Now regression_distribution.cpp
govg has joined #mlpack
govg has quit [Quit: leaving]
jbc_ has joined #mlpack
AndChat-394809 has joined #mlpack
< AndChat-394809>
Anyone
< AndChat-394809>
Whatsupp
< naywhayare>
hello there
< jenkins-mlpack>
Starting build #2146 for job mlpack - svn checkin test (previous build: STILL UNSTABLE -- last SUCCESS #2143 2 days 19 hr ago)
AndChat-394809 has quit [Ping timeout: 255 seconds]
Rishabh has joined #mlpack
< naywhayare>
Rishabh: I saw your messages yesterday; sorry that I was not able to respond
< naywhayare>
have you looked through the mailing list archives for emails about the automatic bindings project?
< Rishabh>
No I will look into that
< naywhayare>
yeah; in the past two years there has been a good amount of discussion on that project
< naywhayare>
unfortunately I have to step out now to get lunch, but I will be back in a few hours and maybe can give a better explanation, if the mailing list doesn't helpa
< naywhayare>
*help
< Rishabh>
That will be a good starting point
< Rishabh>
okay I will be online if I am still awake
< Rishabh>
its night here :P
< naywhayare>
:)
< naywhayare>
I'm usually in the channel so I get all messages eventually
< naywhayare>
so if you're not around tonight, we can catch up tomorrow or whenever
< naywhayare>
anyway, talk to you later... I'm outta here for now
< jenkins-mlpack>
* Ryan Curtin: The const gets ignored (-Wignored-qualifiers).
< jenkins-mlpack>
* Ryan Curtin: Minor spacing fix.
< jenkins-mlpack>
* Ryan Curtin: Minor formatting fixes: tabs->spaces, etc.
< Rishabh>
as far as i understand we need to develop a (script maybe??) which takes source file (??) and outputs a .cpp and a .m file similar to ones in /bingings/matlab
< Rishabh>
which on compilation (installation) will result in generation of mex files at root installation directory of matlab (or so i understand)
< Rishabh>
i am thinking about a matlab based automated bindings generator first then maybe i will expand to python and then R
< naywhayare>
Rishabh: yeah, just got back
< naywhayare>
so the main problem with maintaining separate bindings for matlab/python/R/etc. as well as command-line programs (i.e. kmeans_main.cpp in src/mlpack/methods/kmeans/) is that they each have to be maintained
< naywhayare>
and, supposing that I add an option to kmeans_main.cpp, I need to remember to add it to each of the bindings
< naywhayare>
this is kind of a maintenance nightmare...
< naywhayare>
what I noticed was that each _main.cpp file has three "parts": parsing input parameters, running the algorithm with those parameters, and saving the output
< naywhayare>
and also the hope for bindings is that we provide methods that look very similar to the command-line programs
< naywhayare>
so for instance, from the command line I might run
< naywhayare>
or python... kmeans(input_file='dataset.csv', output_file='output.csv', clusters=5)
< naywhayare>
so basically, like you wrote, the project will involve generating a script or program of some sort that can parse the _main.cpp files (maybe they will need to be modified) and output bindings for each language
< naywhayare>
the meat of the script will basically be a conversion layer between the input parameter section of the _main.cpp file and whatever is used for input in the language of choice
< naywhayare>
anyway, I hope I have given a reasonably coherent explanation of the problem and my ideas for a solution here... maybe I've typed up these ideas a little better in one of the emails to the mlpack list
< Rishabh>
yeah thanks for the explaination
< Rishabh>
another question is like if we have a script which processes these _main.cpp files, there is no issue running them everytime the file is updated right
< Rishabh>
i will install matlab tonight before sleeping and rebuild the src so that i can try out how implemented matlab bindings work out
jbc_ has quit [Quit: jbc_]
< Rishabh>
i have a few ideas i would like to try
< Rishabh>
:)
< naywhayare>
the implemented matlab bindings in trunk are not maintained so I doubt they will work right...
< naywhayare>
but maybe they are a good starting point :)
< Rishabh>
oh.. maybe getting them working will give me an insight to why they dont work
< Rishabh>
:P
< naywhayare>
the underlying APIs for the methods they call have probably changed completely
< Rishabh>
ouch...
< Rishabh>
no issues i can spend time to get them working
< Rishabh>
:)
< Rishabh>
part of the learning curve
< Rishabh>
thanks a lot, i really got some insight into this
< naywhayare>
well, they aren't maintained because the automatic bindings solution should eventually supersede them :)
< naywhayare>
one of the neat things about the automatic bindings project is that I think it could be eventually spun off into its own project, kind of like zoq's benchmarks project from two years ago
< jenkins-mlpack>
Starting build #2148 for job mlpack - svn checkin test (previous build: SUCCESS)