#mlpack on 2017-03-11 — irc logs at libera.irclog.whitequark.org

2015-01-15 23:05 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/

00:01 < AndChat493044> Yes zoq the ann is what i thought of ..

00:01 < AndChat493044> And u knw what .. when i installed mlpack through the ubuntu package install cmd

00:01 < AndChat493044> Mlpack gets installed without ann Library

00:02 < AndChat493044> Its strange .. ill install it through the regular download way

00:03 < zoq> AndChat4930: yeah, we are still working on the ann code, so it's not part of the latest release.

00:03 < AndChat493044> Oh ohkae :)

00:03 < AndChat493044> Yes .. ill open a PR

00:04 < zoq> AndChat4930: Looking forward to take a look :)

00:04 < AndChat493044> So now what to do about the Mario implementation?

00:05 < AndChat493044> And yes .. ive set again everything .. the nes code, fceux , etc and ran the mario implementation too.. cant wait to run mine..

00:06 < AndChat493044> The moment ill figure out the cov matrix thing .. ill be good to go ..

00:07 AndChat493044 has quit [Quit: Bye]

00:07 AndChat493044 has joined #mlpack

00:08 AndChat493044 has quit [Client Quit]

00:08 < zoq> AndChat493044: I would first test it on some easier tasks like the CartPole or DoubleCartPole task once we see some reasonable results, we can can test it on the Mario task.

00:08 < zoq> Also, we recently switch to OpenAI's Gym framework, that makes the setup a lot easier: https://github.com/zoq/gym_tcp_api

00:09 < zoq> And we can test it on a bunch of different tasks.

00:09 AndChat493044 has joined #mlpack

00:11 < AndChat493044> I'm writing the long one only .. using your ann .. sry cant wait :D

00:11 < zoq> AndChat4930: :)

00:12 < AndChat493044> Btw zoq . . Is the documentation present for it ?

00:12 < zoq> For the ann code?

00:13 < AndChat493044> Yes

00:14 < zoq> That's one part that isn't done yet .. .working on it. For now you can take a look at the tests, and I guess if you open the PR, together we can transform the code into something useful based on the ann code really fast.

00:15 < zoq> https://github.com/mlpack/mlpack/blob/master/src/mlpack/tests/feedforward_network_test.cpp

00:15 < zoq> https://github.com/mlpack/mlpack/blob/master/src/mlpack/tests/convolutional_network_test.cpp

00:15 < zoq> https://github.com/mlpack/mlpack/blob/master/src/mlpack/tests/recurrent_network_test.cpp

00:15 < AndChat493044> Ohkae zoq :)

00:19 < zoq> kris1: Here is the modified CMake file: https://gist.github.com/zoq/1810dcd8261078fe35dd118d454f229b also take a look at the comment.

00:25 AndChat493044 has quit [Ping timeout: 246 seconds]

00:39 AndChat493044 has joined #mlpack

00:43 AndChat493044 has quit [Ping timeout: 246 seconds]

01:48 emcenrue has quit [Quit: Page closed]

03:13 shihao has quit [Quit: Page closed]

03:32 trapz has joined #mlpack

03:39 trapz_ has joined #mlpack

03:40 trapz has quit [Ping timeout: 240 seconds]

03:40 trapz_ is now known as trapz

03:51 < rcurtin> today, March 11, is the 10-year anniversary of the first commit to the mlpack repository

03:53 < arunreddy> wow.. congratulations!!

03:54 < arunreddy> Happy 10th anniversary mlpack.!

03:57 < rcurtin> yeah, I was not here at the beginning, I first worked with mlpack in late 2009

03:57 < rcurtin> it has changed a lot from what it was in 2007 and 2008...

03:57 < arunreddy> where did it start? Was it from gatech?

03:58 < rcurtin> yeah, it was from Alex Gray's lab (he was one of my advisors)

03:58 < rcurtin> the focus was on dual-tree algorithms and fast implementations of machine learning algorithms

03:58 < rcurtin> so the "main selling point" of the library was the all-k-nearest-neighbors code and kernel density estimation code

03:59 < rcurtin> since then it's moved to be a more general library, but the emphasis on speed remains

04:01 < arunreddy> i think the fast implementations of ml algos was your contribution. I saw that on your cv..

04:01 < arunreddy> "Understand an algorithm, Make it faster."

04:03 < rcurtin> well, that is what I like to do, but many of the original fast implementations were not mine :)

05:02 vinayakvivek has joined #mlpack

05:03 shihao has joined #mlpack

05:04 < shihao> Suddenly this build error comes: /Users/shihaojing/Projects/mlpack/src/mlpack/../mlpack/prereqs.hpp:75:10: fatal error: 'boost/serialization/serialization.hpp' file not found.

05:05 < shihao> I can build it successfully before I pull some changes from master branch.

05:08 mikeling has joined #mlpack

05:10 < rcurtin> shihao: did you update any packages on your system or anything? maybe re-running CMake is needed (possibly even removing CMakeCache.txt from the build directory then reconfiguring)

05:10 < mikeling> zoq: oh, sorry! I somehow miss this message yesterday. Which message? I haven't solved the issue

05:10 < rcurtin> in either case I am headed to bed, so I won't be able to help further, sorry---I'll answer a question if there is one when I wake up

05:10 < mikeling> rcurtin: good night ;)

05:14 diehumblex has joined #mlpack

05:18 < shihao> rcurtin: Somehow I have to re-clone my remote base :(

05:34 trapz has quit [Quit: trapz]

05:49 shihao has quit [Quit: Page closed]

05:50 etotientz has joined #mlpack

05:51 < etotientz> Hello world !!

06:10 etotientz has quit [Quit: Page closed]

06:57 tash has joined #mlpack

06:59 < tash> hello, I was trying to build mlpack, I have some directory not found errors, causing compilation termination.

07:00 < tash> I'm getting this-

07:00 tash has quit [Client Quit]

07:01 tash has joined #mlpack

07:06 zack_ has joined #mlpack

07:07 zack_ has quit [Client Quit]

07:21 akasher has joined #mlpack

07:30 akasher has quit [Ping timeout: 260 seconds]

07:38 tanisha has joined #mlpack

07:38 vivekp has quit [Ping timeout: 260 seconds]

07:47 mikeling has quit [Quit: Connection closed for inactivity]

07:50 tash has quit [Ping timeout: 260 seconds]

08:05 akasher has joined #mlpack

08:07 vivekp has joined #mlpack

08:21 akasher has quit [Ping timeout: 260 seconds]

08:52 kris1 has joined #mlpack

09:05 vinayakvivek has quit [Quit: Connection closed for inactivity]

09:29 deepanshu_ has joined #mlpack

09:38 < kris1> zoq: look at my comment https://gist.github.com/zoq/1810dcd8261078fe35dd118d454f229b

10:05 thyrix has joined #mlpack

10:39 Nax has joined #mlpack

11:24 trapz has joined #mlpack

11:24 s1998 has joined #mlpack

11:25 < s1998> Hi, can anyone please take a look at this, I am unable to figure out where the bug is ?

11:25 < s1998> http://pastebin.com/1Ke1gRLe

11:46 thyrix has quit [Ping timeout: 260 seconds]

12:05 mikeling has joined #mlpack

12:25 trapz has quit [Quit: trapz]

12:42 s1998 has quit [Ping timeout: 260 seconds]

12:58 < zoq> rcurtin: 10 years ... wow

12:59 < zoq> mikeling: It's at 16:09 : http://www.mlpack.org/irc/mlpack.20170309.html

13:02 deepanshu_ has quit [Quit: Connection closed for inactivity]

13:06 trapz has joined #mlpack

13:09 < mikeling> zoq: Oh, I found it. But why I should use 'FitnessFunction::template Evaluate<UseWeights>(...)' instead?

13:09 trapz has quit [Client Quit]

13:10 trapz has joined #mlpack

13:18 < mikeling> zoq: Ah, it works! No error been reported :) But why? I don't know the reason I should add 'template' after FitnessFunction...

13:18 < mikeling> could you please tell me that?! Thank you!

13:19 < zoq> mikeling: It's because the compiler can't figure out whether the name is a template or not when parsing the code, for more information take a look at 14.2/4 http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2013/n3690.pdf

13:21 s1998 has joined #mlpack

13:21 < zoq> kris1: Strange can you remove the build folder an try again, even if the path is invalid it should still configure the project without any complaints.

13:30 < mikeling> zoq: ah, I see! I will read that pdf you give to me. It's awesome!

13:30 < mikeling> Thank you very much

13:31 < mikeling> that's the reason why I love open-source ;)

13:31 shikhar has joined #mlpack

14:07 Nax has quit [Ping timeout: 260 seconds]

14:10 trapz has quit [Quit: trapz]

14:14 Nax has joined #mlpack

14:19 trapz has joined #mlpack

14:24 trapz has quit [Quit: trapz]

14:33 diehumblex has quit [Quit: Connection closed for inactivity]

14:47 shr7 has joined #mlpack

15:06 s1998 has quit [Ping timeout: 260 seconds]

15:19 shr7 has quit [Ping timeout: 260 seconds]

15:24 Nax has quit [Ping timeout: 260 seconds]

15:28 sicko has quit [Ping timeout: 260 seconds]

15:41 sicko has joined #mlpack

16:25 govg has quit [Ping timeout: 260 seconds]

16:28 govg has joined #mlpack

17:35 sicko has quit [Ping timeout: 260 seconds]

18:07 mikeling has quit [Quit: Connection closed for inactivity]

18:13 shikhar has quit [Quit: Page closed]

19:10 arunreddy has quit [Quit: WeeChat 1.4]

20:09 sdohare has joined #mlpack

20:10 sdohare has quit [Client Quit]

20:56 Guest4565 has joined #mlpack

21:10 Guest4565 has quit [Quit: ...]

21:14 arunreddy has joined #mlpack

21:43 < kris1> zoq: it works now i also added options so that it works with gdb now.

22:21 < kris1> zoq: model.predict always gives zeros values.

22:21 < kris1> i intialized the network parametes with random init stil its giving 0

22:22 < kris1> when i try with forward visitor on many layers incompatile matrix dimension 0x0 is coming.

22:23 < kris1> here is the gist

22:23 < kris1> https://gist.github.com/kris-singh/54cc569b984e3c3f3bf85f7f58fdfc87

22:51 < kris1> if i change the second last layer to from model.Add<Linear<> >(hiddenLayerSize, 1) to model.Add<Linear<> >(hiddenLayerSize, 2); i do get some values.

22:51 < kris1> but my last layer is model.Add<LogSoftMax<>>(); so all my values should be between 0-1 which is not the case

22:57 < kris1> okay now i see the problem lies with use of softmax. hiddenLayerSize,1 works but without the softmax layer.