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/
shujon has joined #mlpack
Tanishq has quit [Ping timeout: 240 seconds]
Laymont has quit [Quit: Page closed]
< wasiq>
Hello
< rcurtin>
wasiq: hello there
< wasiq>
rcurtin, i had spoken to you about the Hogwild! approach to parellized SGD. I went through the ideas again and i found the Frank-Wolfe algorithm a much more interesting read.
< wasiq>
So,I'vebeen reading up on all the material thats been provided.One specific block that ive been having is this projects implementations. Is it along the lines test module or a module which can run all the algorithms based off of Frank-Wolfe?
< rcurtin>
sorry for the slow response, I am doing many things at once :)
< rcurtin>
multitasking is inefficient though...
< rcurtin>
anyway, all of the current mlpack optimizers have tests that are pretty simple--
< rcurtin>
they construct some function to be optimized (take a look at the test_functions.hpp files in the optimizer directories)
< rcurtin>
and then ensure that those functions converge to the proper result when they are optimized
< wasiq>
ok.
< wasiq>
i dint quite follow.You
< rcurtin>
okay, maybe I misunderstood your question then :(
< rcurtin>
I thought you were asking how the mlpack optimizer tests worked
< wasiq>
But isn't the FW algo a foundation over which the other algos are built?
< rcurtin>
no, the FW algorithm is a different optimizer entirely than anything mlpack has implemented
< wasiq>
no my question was on the purpose on the FW algo. Its al right if you reply late. you must be super busy.
< rcurtin>
it's okay, I am here now, I want to help out :)
< rcurtin>
so I'm still not sure I understand the question
< rcurtin>
if you want to implement FW as a project, then the deliverable should basically be:
< rcurtin>
- an optimizer like the other mlpack optimizers, in src/mlpack/core/optimizers/frank_wolfe/ (or whatever you choose to call it)
< rcurtin>
- a suite of tests for the FW optimizer, which will be useful for debugging when the code changes later
< rcurtin>
- optionally, support for the FW optimizer as an option in the various mlpack programs, like mlpack_logistic_regression
< rcurtin>
and of course good documentation for the code that is written :)
< wasiq>
Oh good. Crystal clear.
< wasiq>
Yeah i asked the question wrong. I didn't understand the purpose of FW at all. Now it all makes sense.
< wasiq>
I'm currently reading the paper written by Jaggi:http://m8j.net/math/revisited-FW.pdf
< wasiq>
Is there any other material that i can go through?
< rcurtin>
that paper is the most important one, but you could try reading the papers it references, like the original paper from 1956
< wasiq>
ill work on it then. ok my last question for now. How exactly is the FW optimizer different from the ones that have already been implemented?
< rcurtin>
it's a different algorithm, so it will behave differently for different problems
< rcurtin>
in some cases it may converge faster, in some cases it may converge slower, in some cases it may not converge at all
< wasiq>
hmmm i should go through a few other optimizers. Should help me understand FW better.
< wasiq>
Thanks a lot rcurtin. :) Is it compulsory to introduce myself on the mailing list?
< rcurtin>
no, not compulsory, don't feel obligated
< rcurtin>
there are no strict requirements for what needs to be done to apply (other than to submit an application)
< wasiq>
great :D ok.ill let you know if have any questions.thanks a lot and have a good day.
< rcurtin>
sure, you too :)
wasiq has quit [Ping timeout: 248 seconds]
Falit has joined #mlpack
ank_95_ has quit [Quit: Connection closed for inactivity]
Falit has quit [Remote host closed the connection]
sourad has joined #mlpack
sourad has quit [Read error: Connection reset by peer]
vtantia has quit [Ping timeout: 252 seconds]
aditya123 has quit [Quit: Connection closed for inactivity]
skon46 has joined #mlpack
sourad has joined #mlpack
palashahuja has joined #mlpack
sourad has quit [Ping timeout: 276 seconds]
sourad has joined #mlpack
tsathoggua has quit [Quit: Konversation terminated!]
sourad has quit [Read error: Connection reset by peer]
yvtheja has joined #mlpack
sourad97 has quit [Read error: Connection reset by peer]
sourad97 has joined #mlpack
yvtheja has quit [Ping timeout: 240 seconds]
cache-nez has joined #mlpack
brawlyzebra has joined #mlpack
brawlyzebra has quit [Client Quit]
STW has joined #mlpack
STW has quit [Client Quit]
anveshi has joined #mlpack
mark42 has joined #mlpack
mark42 has quit [Ping timeout: 252 seconds]
cache-nez has quit [Ping timeout: 244 seconds]
yvtheja has joined #mlpack
anveshi has quit [Remote host closed the connection]
sk8r has quit [Ping timeout: 252 seconds]
palamgonz has joined #mlpack
csbqwerty_ has joined #mlpack
< csbqwerty_>
Hey, I'm very interested in this org and would like to contribute for GSOC. I went through this guide http://www.mlpack.org/docs/mlpack-2.0.1/doxygen.php?doc=build.html#build . But I dont know how to run it. I know C++ and learning machine learning on coursera right now. Hope someone can point me in the right direction. Thanks!
csbqwerty_ is now known as csbqwerty
Tanishq has joined #mlpack
csbqwerty is now known as Chinmaya
STW has joined #mlpack
vtantia has joined #mlpack
cache-nez has joined #mlpack
vineet has joined #mlpack
palamgonz has quit [Quit: Page closed]
anveshi has joined #mlpack
kirizaki has joined #mlpack
< zoq>
Chinmaya: Hello, I'm not sure I get your problem. 1. Configuring CMake (mkdir build && cmake ../) 2. make && make install
< Chinmaya>
Yeah, I did that. But i dont know how to use it .. as in what should i type into the shell to run it?
< kirizaki>
@zoq: You was saying something about remote desktop? :)
< zoq>
kirizaki: FreeBSD?
< kirizaki>
yes
cache-nez has quit [Ping timeout: 240 seconds]
< zoq>
It's more like a remote shell (ssh)
Chinmaya has quit [Remote host closed the connection]
vineet has quit [Quit: Konversation terminated!]
vineet has joined #mlpack
sourad97 has quit [Read error: Connection reset by peer]
vineet has quit [Quit: Konversation terminated!]
sourad97 has joined #mlpack
vtantia has quit [Ping timeout: 252 seconds]
Jenna has joined #mlpack
vtantia has joined #mlpack
< Jenna>
Hello! I am an interested GSOC applicant! I was wondering, do some project ideas have higher priorities over others? Also, I am fluent in Java, C, and Python but do not really have experience with C++. Will this disqualify me as an applicant, or could I go around this issue by getting up to speed with C++ on my own and focusing on projects that doesn't required "basic" to no C++ knowledge?
anveshi has quit [Remote host closed the connection]
cache-nez has joined #mlpack
Tanishq has left #mlpack []
NilayNigam has joined #mlpack
oguzhanunlu has joined #mlpack
skon46 has quit [Ping timeout: 240 seconds]
ranajn123 has joined #mlpack
< NilayNigam>
hello
< oguzhanunlu>
Hi guys, I am Oguzhan. I want to participate in GSOC within mlpack. I like to work with machine learning and C++, and mlpack is intersection of my passions. I have a very basic knowledge in ml but have a good knowledge of C++. To contribute to mlpack, I thought the idea with title 'Implement tree types' is suitable for me. Could someone help me with my questions?
< NilayNigam>
I am interested in project on decision trees and implementing tree types. But I dont know how to apply for it so can anyone help ?
< zoq>
Jenna: Hello, as long as you are willing to get familiar with C++ it shouldn't be a problem at all. Take a look at the https://github.com/mlpack/mlpack/wiki/DesignGuidelines to get an impression of the C++ features you should know before applying. We don't have priorities for the projects.
< zoq>
oguzhanunlu: There has been a lot of discussion on the mailing list already about
< ameya_>
Hello! I was trying to compile mlpack library. I encountered this rather bizarre build error. I installed armadillo 6.5 and the build fails giving the following error - http://pastebin.com/9kvxr26Y. Can you please confirm if it is a bug?