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/
trapz has quit [Quit: trapz]
trapz has joined #mlpack
trapz_ has joined #mlpack
trapz has quit [Ping timeout: 240 seconds]
trapz_ is now known as trapz
chenzhe has quit [Ping timeout: 246 seconds]
trapz has quit [Quit: trapz]
nu11p7r has joined #mlpack
chenzhe has joined #mlpack
chenzhe has quit [Ping timeout: 240 seconds]
chenzhe has joined #mlpack
chenzhe has quit [Ping timeout: 240 seconds]
chenzhe has joined #mlpack
chenzhe has quit [Ping timeout: 240 seconds]
chenzhe has joined #mlpack
chenzhe has quit [Ping timeout: 240 seconds]
chenzhe has joined #mlpack
chenzhe has quit [Remote host closed the connection]
chenzhe has joined #mlpack
msohcw has joined #mlpack
< msohcw>
Hi. This might be a little (or very) late but I was wondering if you guys think you might still be looking for someone to work on the RL project for GSoC.
< msohcw>
A little short intro about myself, I got interested in ML at the end of 2015/beginning of 2016, particularly interested in deep RL. I implemented the deepmind papers from scratch (to play flappy bird!) specifically, DQN, DDQN, prioritised experience replay and deep-duelling Q networks.
< msohcw>
I'm also familiar with current state of the art like A3C, DDPG etc.
< msohcw>
I'm currently with Machine Learning @ Berkeley, a student organisation that focuses on machine learning projects, and I'm on a team working on a deep RL library right now.
< msohcw>
I'm fluent with C++, albeit maybe not to the extent that you guys require, but I'm definitely willing to pick it up.
< msohcw>
Oh and if it's a bonus, I've also worked on OpenAI problems, and I think you guys are interested in integrating those in?
< msohcw>
I've read through most of the ML-related GSoC projects and mlpack has the most interesting one by far. Is the RL project going to be a focus for the team? Are there any particular areas which you are looking specifically for?
msohcw has quit [Ping timeout: 260 seconds]
chenzhe1 has joined #mlpack
chenzhe has quit [Ping timeout: 246 seconds]
chenzhe1 is now known as chenzhe
chenzhe has quit [Ping timeout: 240 seconds]
chenzhe has joined #mlpack
chenzhe has quit [Read error: Connection reset by peer]
nishanth has quit [Read error: Connection reset by peer]
Trion has joined #mlpack
nish21 has quit [Quit: Leaving]
whosc has quit [Ping timeout: 240 seconds]
whosc has joined #mlpack
benchmark has joined #mlpack
benchmark has quit [Client Quit]
govg has joined #mlpack
< whosc>
Hello, How do you know iff a system can be solved using linear or non linear ways?
< Trion>
If data can be represented as a line or can be separated by a line then use linear ways otherwise non-linear ways
shikhar has quit [Quit: Page closed]
mich has joined #mlpack
< mich>
hello
apa_na has joined #mlpack
< zoq>
mich: Hello there!
< Trion>
:P Thanks for the comments zoq! The Agent finally worked!
< zoq>
Trion: No problem, let me know if you have any further questions.
< Trion>
zoq: :P I still broke something, the iex server we have is giving "No module named monitoring" even though I installed it and verified in python REPL. I am connecting to it via 127.0.0.1:4040
< whosc>
Trion, and how do i know if the data can be separated by a line?
< whosc>
The basic problem is one that can be solved by parameterization.. I have an object , i just need to keep watch on real time data to make sure environmental variables do not cross a certain threshold.
< whosc>
Its a yes or a no type. Like the object is safe, or the object is unsafe.
< Trion>
I mostly set up a data visualization using matplotlib in python, makes understanding data easier
< zoq>
Trion: Sounds like you use an old version of gym, in the newer versions "env = wrappers.Monitor(env, directory)" isn't used anymore in favor of nv = wrappers.Monitor(env, directory). If you use the git version: https://github.com/openai/gym#installation you should be fine.
< zoq>
I have to step out for a while, back later.
< Trion>
whosc: You can train Naive Bayes Classifier with labeled data for that object if the object is safe or unsafe
sumedhghaisas has joined #mlpack
mich has quit [Quit: Page closed]
Trion has quit [Quit: Have to go, see ya!]
chenzhe has joined #mlpack
shikhar has joined #mlpack
chenzhe has quit [Ping timeout: 256 seconds]
msohcw has joined #mlpack
msohcw has quit [Client Quit]
vss has quit [Quit: Page closed]
chenzhe has joined #mlpack
< vivekp>
rcurtin: Regarding the SMORMS3 PR, I don't understand how we'd use for_each() and a lambda to avoid allocating an extra matrix, stepSizeMat
< vivekp>
Could you elaborate a bit?
chenzhe has quit [Ping timeout: 246 seconds]
apa_na has quit [Ping timeout: 260 seconds]
chenzhe has joined #mlpack
chenzhe has quit [Ping timeout: 246 seconds]
chenzhe has joined #mlpack
shikhar has quit [Quit: Page closed]
chenzhe has quit [Ping timeout: 246 seconds]
< rcurtin>
vivekp: I am in meetings all day today but I can probably find a chance to write something on.that PR, don't worry :)
< rcurtin>
I'll comment when I have a chance
chenzhe has joined #mlpack
chenzhe has quit [Ping timeout: 246 seconds]
whosc has quit [Quit: Leaving]
mikeling has quit [Quit: Connection closed for inactivity]
< vivekp>
rcurtin: ah, okay thanks :)
chittaranjan has joined #mlpack
ironstark has joined #mlpack
< ironstark>
Can we add benchmark scripts for graphlab create library in python?
< rcurtin>
ironstark: sure, as long as graphlab create is free, I can install it on the benchmarking systems
< ironstark>
for that we need to update the dataset folder and add some .gl files because the library requires .gl extension datasets
< rcurtin>
could you write a shim that takes in .csv and outputs .gl?
< rcurtin>
the datasets are pretty big so it would be good to try and keep the size from exploding
< rcurtin>
so making a local copy would be preferable to adding lots to the repos
< ironstark>
I can try
chenzhe has joined #mlpack
< ironstark>
I'll first see whether there are options to directly read datasets in csv format or not
< rcurtin>
sure, I would imagine that there are
< rcurtin>
I remember working with graphlab from years ago and I seem to remember using CSVs
< rcurtin>
are they still developing Create or have they kind of disappeared since being bought by Apple?
< ironstark>
It is still being developed I guess but its not open source
chenzhe has quit [Ping timeout: 246 seconds]
< ironstark>
also to install graphlab for free we need to register as an academic user
< ironstark>
Is it possible for mlpack to register as an academic user?
< chittaranjan>
Are we inclined towards benchmarking more libraries, or more algorithms of libraries we have now?
< rcurtin>
ironstark: honestly if they aren't giving it away for free and we have to register, it's going to be a lot of maintenance to keep it installed and working, and I think it's not worth bothering
< rcurtin>
unless they come to us and contribute some patches to the benchmarking system or something, which I doubt they will
< rcurtin>
chittaranjan: either is fine
< ironstark>
rcurtin : ya it requires registration and regular product key renewals to maintain so I should drop this idea for now
< rcurtin>
agreed
chenzhe has joined #mlpack
chenzhe has quit [Ping timeout: 246 seconds]
Rebryk has joined #mlpack
Rebryk has quit [Client Quit]
chenzhe has joined #mlpack
DarkishPizza has joined #mlpack
chenzhe has quit [Ping timeout: 246 seconds]
< zoq>
ironstark: If you search for other interesting libraries, I guess https://github.com/teemupitkanen/mrpt might be interesting to look into, would be interesting to see if mlpack can compete.
< ironstark>
zoq: Thanks I'll look into this one and try to implement
< zoq>
Maybe rcurtin already tested the method ... not sure
< ironstark>
Also I have removed the if condition in the randomforesttest PR(#34). Please test it once.