--- Log opened Mon Mar 24 00:00:52 2014 | ||
-!- PirosB3 [~pirosb3@cpc4-newt31-2-0-cust631.newt.cable.virginm.net] has joined #shogun | 00:30 | |
bumblebee | HeikoS: there? | 00:43 |
---|---|---|
HeikoS | bumblebee: yes, hi | 00:50 |
bumblebee | Hey, i am not really familiar with IPython, any help, any link, I have a rough idea though | 00:52 |
bumblebee | btw, for #1982 "Process small datasets for recommendation systems", just wanted to know whethe i am on right track for processing and readind data | 00:53 |
bumblebee | using shogun i.e. | 00:54 |
HeikoS | bumblebee: you should ask google, ipython is very well documented in every sense | 00:58 |
bumblebee | ohk, sure!! will follow up | 00:59 |
HeikoS | bumblebee: which OS are you using? | 01:00 |
bumblebee | ubuntu | 01:00 |
HeikoS | bumblebee: on linux its just sudo pip install ipython | 01:00 |
bumblebee | yea, got that part, i did the install and all | 01:01 |
bumblebee | just not clear with how i will be using ML on top of a ipython notebook | 01:01 |
bumblebee | so is it that I will process and read data using python through a notebook, and then use ML algos using shogun | 01:02 |
bumblebee | cause it will be in C++, or i will be using python for that too | 01:02 |
bumblebee | ? | 01:02 |
bumblebee | HeikoS: so it's basically use the sushi data in a ipython notebook for a given algo. in shogun | 01:09 |
HeikoS | bumblebee: nono, its all python interface to shogun | 01:10 |
HeikoS | bumblebee: so you use the same class to load the data than you did in your c++ code | 01:10 |
HeikoS | but from python | 01:10 |
HeikoS | and then you will call the GP algorithms in Shogun from python | 01:10 |
HeikoS | all from the notebook | 01:11 |
bumblebee | oh alright | 01:11 |
bumblebee | HeikoS: no theory, | 01:12 |
bumblebee | ? | 01:12 |
bumblebee | HeikoS:Also shogun is written in C++, right? and it's interfacing is primaraily /always in python or is it that only for this example and matters as per users convinience? | 01:14 |
-!- travis-ci [~travis-ci@ec2-54-81-190-108.compute-1.amazonaws.com] has joined #shogun | 01:16 | |
travis-ci | [travis-ci] it's khalednasr's turn to pay the next round of drinks for the massacre he caused in khalednasr/shogun: http://travis-ci.org/khalednasr/shogun/builds/21390140 | 01:16 |
-!- travis-ci [~travis-ci@ec2-54-81-190-108.compute-1.amazonaws.com] has left #shogun [] | 01:16 | |
HeikoS | bumblebee: exactly! | 01:48 |
HeikoS | bumblebee: no theory? | 01:48 |
-!- jiaolong [553a5634@gateway/web/freenode/ip.85.58.86.52] has quit [Quit: Page closed] | 01:51 | |
-!- HeikoS [~heiko@0546d087.skybroadband.com] has quit [Ping timeout: 264 seconds] | 02:02 | |
-!- andrew__ [~andrew@c-50-184-110-141.hsd1.ca.comcast.net] has joined #shogun | 02:04 | |
-!- andrew__ [~andrew@c-50-184-110-141.hsd1.ca.comcast.net] has quit [Ping timeout: 264 seconds] | 02:21 | |
-!- khalednasr [~k.nasr92@41.69.156.170] has quit [Quit: Leaving] | 02:38 | |
-!- andrew__ [~andrew@adsl-99-96-246-54.dsl.pltn13.sbcglobal.net] has joined #shogun | 02:39 | |
-!- pl8787 [9fe22b60@gateway/web/freenode/ip.159.226.43.96] has quit [Ping timeout: 245 seconds] | 02:54 | |
-!- PirosB3 [~pirosb3@cpc4-newt31-2-0-cust631.newt.cable.virginm.net] has quit [Quit: PirosB3] | 02:59 | |
-!- pl8787 [9fe22b60@gateway/web/freenode/ip.159.226.43.96] has joined #shogun | 03:13 | |
andrew__ | I'm trying to run some examples to make sure everything built successfully- set PYTHONPATH to /usr/local/lib/python2.7/dist-packages/ as described by doc/INSTALL.md | 03:14 |
pl8787 | Linking CXX shared library libshogun.so /usr/bin/ld: /usr/local/lib/libnlopt.a(optimize.o): relocation R_X86_64_32S against `.text' can not be used when making a shared object; recompile with -fPIC /usr/local/lib/libnlopt.a: could not read symbols: Bad value | 03:14 |
pl8787 | How to solve this problem? | 03:14 |
andrew__ | but dist-packages/ is empty. Does this mean I failed to turn on the modular python interface? | 03:14 |
pl8787 | andrew__: what's your cmake command like? | 03:16 |
andrew__ | pretty sure I ran cmake -DPythonModular=ON | 03:20 |
-!- abinash_panda [01261427@gateway/web/freenode/ip.1.38.20.39] has joined #shogun | 03:31 | |
pl8787 | andrew__: do you run make install? | 03:32 |
pl8787 | Linking CXX shared library libshogun.so /usr/bin/ld: /usr/local/lib/libnlopt.a(optimize.o): relocation R_X86_64_32S against `.text' can not be used when making a shared object; recompile with -fPIC /usr/local/lib/libnlopt.a: could not read symbols: Bad value | 03:33 |
pl8787 | how to solve this problem? | 03:33 |
-!- andrew__ [~andrew@adsl-99-96-246-54.dsl.pltn13.sbcglobal.net] has quit [Ping timeout: 268 seconds] | 03:36 | |
shogun-buildbot | build #756 of nightly_default is complete: Failure [failed test] Build details are at http://buildbot.shogun-toolbox.org/builders/nightly_default/builds/756 | 03:43 |
-!- lisitsyn [~lisitsyn@109-226-71-150.clients.tlt.100megabit.ru] has left #shogun [] | 04:25 | |
-!- soumyaC_ [uid15286@gateway/web/irccloud.com/x-ywfpriynusavedzn] has quit [Quit: Connection closed for inactivity] | 04:25 | |
-!- abinash_panda [01261427@gateway/web/freenode/ip.1.38.20.39] has quit [Ping timeout: 245 seconds] | 04:25 | |
-!- zxtx [~zv@cpe-98-148-41-3.socal.res.rr.com] has quit [Ping timeout: 240 seconds] | 04:50 | |
-!- andrew___ [~andrew@c-50-184-110-141.hsd1.ca.comcast.net] has joined #shogun | 05:26 | |
-!- bumblebee [~quassel@202.78.172.162] has quit [Quit: No Ping reply in 180 seconds.] | 05:33 | |
-!- bumblebee [~quassel@202.78.172.162] has joined #shogun | 05:33 | |
-!- andrew___ [~andrew@c-50-184-110-141.hsd1.ca.comcast.net] has quit [Ping timeout: 252 seconds] | 05:43 | |
-!- andrew___ [~andrew@c-50-184-110-141.hsd1.ca.comcast.net] has joined #shogun | 05:54 | |
-!- bumblebee [~quassel@202.78.172.162] has quit [Ping timeout: 264 seconds] | 06:33 | |
-!- kprah [67157f55@gateway/web/freenode/ip.103.21.127.85] has joined #shogun | 07:11 | |
-!- zxtx [~zv@23-114-213-191.lightspeed.irvnca.sbcglobal.net] has joined #shogun | 07:12 | |
kprah | wiking: why do we have multiple model proximal bm solver but not multiple cp bmrm? | 07:12 |
kprah | wiking: please reply later if you get time. i have to discuss multiple cutting plane | 07:16 |
-!- kprah [67157f55@gateway/web/freenode/ip.103.21.127.85] has quit [Ping timeout: 245 seconds] | 07:20 | |
-!- bumblebee [~quassel@202.78.172.162] has joined #shogun | 07:51 | |
-!- lisitsyn [~lisitsin@mxs.kg.ru] has joined #shogun | 07:57 | |
-!- bumblebee [~quassel@202.78.172.162] has quit [Ping timeout: 240 seconds] | 08:03 | |
-!- bumblebee [~quassel@202.78.172.162] has joined #shogun | 08:05 | |
-!- kislay [~kislay@14.139.41.150] has quit [Remote host closed the connection] | 08:29 | |
-!- kislay [~kislay@117.239.94.112] has joined #shogun | 08:32 | |
-!- kislay [~kislay@117.239.94.112] has quit [Remote host closed the connection] | 09:24 | |
-!- jiaolong [9e6d09e4@gateway/web/freenode/ip.158.109.9.228] has joined #shogun | 09:53 | |
-!- bumblebee [~quassel@202.78.172.162] has quit [Ping timeout: 240 seconds] | 10:19 | |
-!- abinash_panda [01261427@gateway/web/freenode/ip.1.38.20.39] has joined #shogun | 10:28 | |
-!- nikolis [~nikolis@213.16.163.237.dsl.dyn.forthnet.gr] has joined #shogun | 10:35 | |
-!- PirosB3 [~pirosb3@cpc4-newt31-2-0-cust631.newt.cable.virginm.net] has joined #shogun | 10:37 | |
-!- rho [~quassel@193.227.20.2] has joined #shogun | 10:38 | |
-!- abinash_panda [01261427@gateway/web/freenode/ip.1.38.20.39] has quit [Ping timeout: 245 seconds] | 10:47 | |
-!- rho [~quassel@193.227.20.2] has quit [Quit: No Ping reply in 180 seconds.] | 10:48 | |
-!- rho [~quassel@193.227.20.2] has joined #shogun | 10:49 | |
-!- Rho__ [~rho@193.227.20.2] has joined #shogun | 11:00 | |
-!- rho [~quassel@193.227.20.2] has quit [Remote host closed the connection] | 11:01 | |
-!- Rho__ [~rho@193.227.20.2] has quit [Read error: Connection reset by peer] | 11:01 | |
-!- nikolis [~nikolis@213.16.163.237.dsl.dyn.forthnet.gr] has quit [Remote host closed the connection] | 11:02 | |
-!- DarkRoom [~darkroom@14.139.82.6] has joined #shogun | 11:38 | |
-!- nikolis [~nikolis@213.16.163.237.dsl.dyn.forthnet.gr] has joined #shogun | 11:39 | |
-!- PirosB3 [~pirosb3@cpc4-newt31-2-0-cust631.newt.cable.virginm.net] has quit [Quit: PirosB3] | 11:40 | |
-!- nikolis [~nikolis@213.16.163.237.dsl.dyn.forthnet.gr] has quit [Quit: Leaving] | 11:49 | |
-!- lambday [67157e54@gateway/web/freenode/ip.103.21.126.84] has joined #shogun | 11:49 | |
@wiking | lambday: yoyoyoooo | 11:59 |
@wiking | lambday: here? | 11:59 |
lambday | wiking: hey! | 12:01 |
@wiking | lambday: oh great u r here | 12:02 |
@wiking | so i'm almost done with the refactoring of eigen solver | 12:02 |
@wiking | but i have a small problem now | 12:02 |
@wiking | so currently in LanczosSolver | 12:03 |
lambday | wiking: oh great! what is it? | 12:03 |
@wiking | you basically calculate the extremal eigen values | 12:03 |
@wiking | while the new api currently doesn't support that | 12:03 |
@wiking | i.e. we have like | 12:03 |
@wiking | compute() | 12:03 |
@wiking | that computes all of them | 12:03 |
@wiking | and compute_few() | 12:03 |
@wiking | where the few means an index intervall | 12:03 |
@wiking | (just like lapack) | 12:04 |
@wiking | if we use compute_few for calculating the extremals | 12:04 |
@wiking | it's just as if we call compute | 12:04 |
@wiking | i'm wondering if u have some api suggestions here? :) | 12:04 |
@wiking | like changing compute_few() into | 12:05 |
@wiking | where you can pass an SGVector<index_t> idx | 12:05 |
@wiking | that identifies what eigenvalues u r interested in? :) | 12:05 |
@wiking | or maybe have both... compute_few (lower index, upper index) | 12:06 |
@wiking | compute_few(SGVector<index_t> idx) | 12:06 |
lambday | wiking: well, if indices are far apart then it makes sense to compute all instead... | 12:06 |
lambday | this will make a diference only when a range is provided | 12:06 |
lambday | right? | 12:06 |
-!- kislay [~kislay@117.239.94.112] has joined #shogun | 12:06 | |
lambday | I mean, passing individually indices is probably something we won't need.. rather just compute all would be the way to go | 12:06 |
@wiking | lambday: that's a huge overhead | 12:07 |
@wiking | i mean compared to what we do now | 12:07 |
@wiking | right? | 12:07 |
@wiking | currently with lanczos u calculate only 2 eigen values | 12:07 |
lambday | yeah | 12:07 |
lambday | 2 calls | 12:07 |
@wiking | yeah | 12:08 |
@wiking | computing all | 12:08 |
@wiking | will take much longer | 12:08 |
lambday | yeah true | 12:08 |
@wiking | (of course it depends on the linear operator size) | 12:08 |
-!- bumblebee [~quassel@202.78.172.162] has joined #shogun | 12:08 | |
@wiking | but if they are not dummy tests then i'd say it'll be a huge runtime difference | 12:08 |
lambday | wiking: so I was thinking compute_few should do something like this - it computes eigenvalue indices from 0 to k... and j to N | 12:09 |
lambday | N is #evs | 12:09 |
lambday | and k and j are to be passed as params | 12:09 |
@wiking | ah ok | 12:09 |
@wiking | so instead of like what lapack | 12:09 |
@wiking | where you give the interval you want | 12:09 |
@wiking | you give lower upper and higher lower boundaries | 12:09 |
lambday | wiking: I guess that would fit nicely into the existing implementation | 12:10 |
@wiking | [0,k], [j,N] | 12:10 |
lambday | yo | 12:10 |
@wiking | yeah we can change into that | 12:10 |
@wiking | and then you know that in the SGVector<float64_t> the first k values are the for the lower interval and the remaining N-j values are fot the upper interval | 12:11 |
lambday | wiking: oh and one more thing - did you happen to check the CPU/GPU dot thread? using a LinearAlgebra wrapper for setting different backends for all dots, eigensolvers etc all at once? | 12:11 |
lambday | something like sg_linalg->set_backend() | 12:11 |
lambday | wiking: yeah! | 12:11 |
@wiking | lambday: yeah i'm into that | 12:11 |
@wiking | lambday: i mean you'll see my new commit | 12:11 |
@wiking | that doesn't have this yet | 12:11 |
@wiking | but i'll introduce later some sort of globally settable flag | 12:12 |
@wiking | where u can switch backends | 12:12 |
@wiking | (lapack, eigen etc.) | 12:12 |
lambday | wiking: absolutely - that would be great! | 12:12 |
lambday | wiking: and our linear algreba routines would rely on global functors or sort - so that one interface works for all backends | 12:13 |
@wiking | well that's the plan | 12:14 |
@wiking | but it'll be quite a pain | 12:14 |
lambday | massive pita | 12:14 |
lambday | wiking: plus heiko likes that vector-dot-operators should be brought under linear operator interface - that would involve lots of refactoring | 12:14 |
lambday | templating a lot of things | 12:15 |
lambday | I hope its worth it | 12:15 |
lambday | :) | 12:15 |
@wiking | well yeah that'll need a new feature branch i reckon | 12:15 |
lambday | wiking: by the way I am not sure about this [0,k], [j,N] thing - I mean, what if someone is only interested in N/2-th ev? | 12:17 |
lambday | :-/ | 12:17 |
@wiking | he is fucked :) | 12:17 |
lambday | wiking: lol so we gotta take care that he's not :D | 12:18 |
lambday | umm... | 12:18 |
@wiking | btw somebody fucked up src/shogun/mathematics/linalg/ratapprox/logdet/LogDetEstimator.cpp | 12:18 |
@wiking | this should not be present | 12:18 |
@wiking | CLogDetEstimator::CLogDetEstimator(SGSparseMatrix<float64_t> sparse_mat) | 12:18 |
@wiking | unless one has both EIGEN3 and LAPACK | 12:19 |
lambday | wiking: oh shit! | 12:19 |
@wiking | as currently LanczosSolver depends on both | 12:19 |
lambday | yeah! | 12:19 |
lambday | wiking: let's ask sunil (who changed this) to fix this | 12:19 |
lambday | wiking: should I open an issue or just fix it? | 12:24 |
@wiking | lambday: as u wish | 12:27 |
lambday | wiking: I guess we should provide different interfaces for compute_few... I was just working on a problem that requires to compute last k eigenvalues... | 12:27 |
@wiking | hehe yeah | 12:28 |
@wiking | there's unforetunetely no one for all api for this problem | 12:28 |
lambday | wiking: it can be set using j, k and a boolean flag I guess... when the flag is set, we compute [j, k]... if unset, we compute [0,j],[k,N] | 12:29 |
@wiking | hahah | 12:29 |
lambday | :D | 12:30 |
@wiking | that's a really aweful design | 12:30 |
lambday | crazy - works though! :D | 12:30 |
@wiking | we can have several compute_few functions | 12:30 |
@wiking | and then one calls the right one | 12:30 |
@wiking | and that's all | 12:30 |
@wiking | lambday: no it's not crazy it really bad design | 12:30 |
@wiking | a flag for doing things | 12:31 |
@wiking | it's like 1970s | 12:31 |
@wiking | that's like exactly how lapack is working | 12:31 |
@wiking | V if you want the eigenvalues as well | 12:31 |
@wiking | U if not | 12:31 |
lambday | :D | 12:31 |
@wiking | or some shit like that... please it's 2014 now | 12:31 |
@wiking | let's be a bit more sophisticated | 12:31 |
@wiking | if possible | 12:31 |
-!- HeikoS1 [~heiko@0546d087.skybroadband.com] has joined #shogun | 12:33 | |
lambday | HeikoS1: hi! | 12:33 |
lambday | wiking: the cg iterations is common for all compute methods.. so I think that can be pushed under a private method - say, compute_Lanczos_T_matrix().. all compute methods will require that | 12:37 |
lisitsyn | AbstractGeneralFactorySingletonManagerRAIICascadeOrchestrationDatabaseEngineWithSFINAEMembersCheck | 12:42 |
lisitsyn | lambday: ^ please use that technique | 12:43 |
lisitsyn | sometimes it is refered as AGFSMRAIICODEWSFINAEMC | 12:43 |
-!- bumblebee [~quassel@202.78.172.162] has quit [Read error: Operation timed out] | 12:43 | |
* lambday crashes on the floor! | 12:44 | |
lambday | lisitsyn: is that even a thing? :D | 12:44 |
-!- bumblebee [~quassel@202.78.172.162] has joined #shogun | 12:44 | |
lisitsyn | lambday: go figure! :D | 12:44 |
DarkRoom | lisitsyn: Hi! | 12:45 |
lisitsyn | DarkRoom: hey | 12:45 |
DarkRoom | I've applied to GSoC, the deep learning project. | 12:45 |
DarkRoom | I was looking at the convolution task mentioned by you. | 12:45 |
lisitsyn | DarkRoom: sure | 12:46 |
DarkRoom | so I planned on using OpenMP to parallize the code (to be run on the CPU) | 12:46 |
lisitsyn | hmm yeah why not | 12:46 |
lisitsyn | it makes sense | 12:46 |
DarkRoom | I got a runtime of 3.4 seconds for a 2000 x 2000 matrix and a 50 x 50 kernel. | 12:47 |
lisitsyn | DarkRoom: cool | 12:48 |
lisitsyn | DarkRoom: actually it is worth to post it to the issue | 12:48 |
DarkRoom | Sure. | 12:49 |
HeikoS1 | lambday: hi | 12:49 |
DarkRoom | I think I might be able to do better, though. | 12:49 |
lambday | HeikoS1: hey... me and wiking were just discussing about the interface for compute_few() in eigensolvers | 12:49 |
HeikoS1 | lambday: I see, and? | 12:50 |
HeikoS1 | what about having a get_eigenvectors method which returns a matrix | 12:50 |
lambday | HeikoS1: so, many cases we need to compute [0,j].. [k,N]... and [j,k] evs | 12:50 |
HeikoS1 | yeah | 12:50 |
lambday | HeikoS1: yeah that should also be there! | 12:51 |
HeikoS1 | lambday: ok tell me what you wanna do | 12:51 |
lisitsyn | DarkRoom: it would be better to submit some code than reach some performance | 12:51 |
lambday | HeikoS1: having separate interfaces for compute_few for these different cases is the solution I guess.. otherwise one interface would require flags and all.. makes it ugly | 12:52 |
HeikoS1 | lambday: ok, mmmmh | 12:53 |
HeikoS1 | why cant we just tell the thing in the beginning how many eigenvectors to compute | 12:53 |
HeikoS1 | and then compute does it | 12:53 |
DarkRoom | lisitsyn: Sure. Will do. But is it okay if I submit a standalone version (not integrated with the toolbox) for now, as I would like to first get better performances. | 12:53 |
lambday | HeikoS1: how many is fine.. but which ones... | 12:53 |
HeikoS1 | and then there are methods to get all computed eigenvectors (a matrix) and eigenvectors with a certain index | 12:53 |
HeikoS1 | all computed eigenvectors are either a square matrix (all) or a matrix containing say only 3. | 12:54 |
HeikoS1 | or one can do get_eigenvector which returns a vector | 12:54 |
HeikoS1 | get_eigenvectors(idx) | 12:54 |
HeikoS1 | and then we can also have the get_largest and get_smallest, that is just wrapper stuff | 12:54 |
HeikoS1 | so everythign is handled in the constructor | 12:54 |
HeikoS1 | lambday: but I dont insist on that, if you want to have a seperate class for few eigenvectors and all eigenvectors its fine | 12:55 |
HeikoS1 | lambday: after all, few eigenvectors will only be nececssary for large sparse matrices | 12:55 |
lambday | HeikoS1: no no.. not separate class.. a separate overloaded method | 12:55 |
HeikoS1 | lambday: so the backends *will* definitely differ | 12:55 |
HeikoS1 | lambday: ah yeah, please go ahead | 12:55 |
lisitsyn | DarkRoom: having it integrated would be important too | 13:02 |
DarkRoom | lisitsyn: Yes, I'll do that too. | 13:08 |
lambday | HeikoS1: I added unit-tests for different number of samples and verified against this implementation - https://github.com/lambday/shogun-hypothesis-testing/blob/master/mmd.py | 13:16 |
lambday | HeikoS1: sending a PR :) | 13:16 |
HeikoS1 | lambday: nice :) Put the link also in the PR pls | 13:16 |
lambday | HeikoS1: yeah - put that in comment as well :) | 13:17 |
HeikoS1 | lambday: I will re-check this with dino today but I think it was already fine | 13:17 |
lambday | HeikoS1: okay :) | 13:17 |
lambday | HeikoS1: oh and should I again make a quick B-test implementation in python as well? then we can check if its fine and later use that for unit-tests | 13:17 |
HeikoS1 | lambday: totally! this is always usefu.l | 13:18 |
lambday | HeikoS1: alright :) | 13:18 |
-!- kislay [~kislay@117.239.94.112] has quit [Read error: Connection reset by peer] | 13:27 | |
-!- kislay [~kislay@117.239.94.112] has joined #shogun | 13:28 | |
-!- Saurabh7 [~Saurabh7@117.235.67.203] has joined #shogun | 13:45 | |
-!- HeikoS1 [~heiko@0546d087.skybroadband.com] has quit [Ping timeout: 252 seconds] | 13:46 | |
-!- eglif [8184de63@gateway/web/freenode/ip.129.132.222.99] has quit [Quit: Page closed] | 13:53 | |
-!- Saurabh7_ [~Saurabh7@115.248.45.78] has joined #shogun | 14:01 | |
-!- Saurabh7 [~Saurabh7@117.235.67.203] has quit [Ping timeout: 240 seconds] | 14:03 | |
-!- HeikoS1 [~heiko@pat-191-250.internal.eduroam.ucl.ac.uk] has joined #shogun | 14:19 | |
kislay | HeikoS1, hi :) | 14:21 |
HeikoS1 | kislay: hey there | 14:22 |
kislay | HeikoS1, I have sent a PR regarding #2064 | 14:23 |
kislay | HeikoS1, tried to keep their structure similar. | 14:23 |
HeikoS1 | kislay: nice, thanks! | 14:24 |
HeikoS1 | kislay: I will have a look later today | 14:24 |
kislay | HeikoS1, sure. | 14:24 |
kislay | HeikoS1, I have worked a lot of with these issues but I am kinda unable to start with the core shogun code. Any Issue you can tell where I can commit myself for now. | 14:27 |
kislay | HeikoS1, It will be really helpful. | 14:27 |
HeikoS1 | kislay: what do you mean exactly? | 14:27 |
HeikoS1 | kislay: maybe write a unit test for some existing methods, | 14:27 |
HeikoS1 | kislay: or wh ynot extend a web-demo with model-selection for GPs | 14:27 |
HeikoS1 | add a button ("learn parameters with ML2") | 14:28 |
kislay | HeikoS1, yup. thats what I was asking about :D | 14:28 |
HeikoS1 | for that you need some internals of the GP framework, see my notebook and other examples | 14:28 |
kislay | HeikoS1, Okay. Now I have a target. Thanks :) | 14:29 |
HeikoS1 | kislay: you are welcome, let me know if you have any quetsion :) | 14:29 |
HeikoS1 | questions | 14:29 |
kislay | HeikoS1, I will work on adding a button thing for now. Then Unit tests for later. | 14:30 |
HeikoS1 | kislay: yeah, for unit tests have a look into some issues | 14:31 |
HeikoS1 | kislay: in general, look at the issue and just pick something | 14:31 |
HeikoS1 | kislay: but in fact, all your notebook updates are appreciated even more :) | 14:32 |
HeikoS1 | kislay: but demo stuff would be cool, you might find bugs, so that will get you involved into Shoguns internals | 14:32 |
kislay | HeikoS1, oh! never knew this :D | 14:32 |
-!- sonne|work [~sonnenbu@24-134-74-216-dynip.superkabel.de] has quit [Quit: Leaving.] | 14:41 | |
-!- sonne|work [~sonnenbu@24-134-74-216-dynip.superkabel.de] has joined #shogun | 14:41 | |
-!- bumblebee [~quassel@202.78.172.162] has quit [Ping timeout: 265 seconds] | 14:47 | |
-!- lisitsyn [~lisitsin@mxs.kg.ru] has quit [Quit: Leaving.] | 15:17 | |
naywhayare | HeikoS1: it did not look like there were any submissions to shogun for the benchmarking project | 16:03 |
naywhayare | I took a quick look through and did not see any | 16:04 |
-!- nikolis [~nikolis@213.16.163.237.dsl.dyn.forthnet.gr] has joined #shogun | 16:17 | |
nikolis | Hello guys . I have installed shogun on my pc and then did the export and chmod for the library now I just run python and import shogun libraries or am I missing something ? | 16:19 |
pfm_ | if you open python and type import modshogun do you get any errors? | 16:19 |
-!- lambday [67157e54@gateway/web/freenode/ip.103.21.126.84] has quit [Ping timeout: 245 seconds] | 16:20 | |
nikolis | I did this from shogun.Kernel import | 16:20 |
nikolis | and recieved this : Traceback (most recent call last): | 16:20 |
nikolis | File "<stdin>", line 1, in <module> | 16:20 |
nikolis | ImportError: No module named shogun.Kernel | 16:20 |
-!- bumblebee [~quassel@202.78.172.162] has joined #shogun | 16:22 | |
-!- nikolis [~nikolis@213.16.163.237.dsl.dyn.forthnet.gr] has quit [Quit: Leaving] | 16:23 | |
-!- nikolis [~nikolis@213.16.163.237.dsl.dyn.forthnet.gr] has joined #shogun | 16:25 | |
sonne|work | nikolis: did you set the PYTHONPATH to the correct dir? did you use the same python version like you used for compiling, did you run ldconfig? | 16:53 |
sonne|work | ... | 16:53 |
nikolis | i ges i did not set the python path because i figured out that if I run python from within the directory modshogun.py exist the imports run smoothly | 16:55 |
nikolis | the directory that contains * modshogun.py | 16:55 |
sonne|work | well then it would work when you install it someplace too | 16:55 |
nikolis | I am not sure if i get your last point | 16:57 |
nikolis | I mean i have already installed it | 16:57 |
sonne|work | where to? | 16:57 |
nikolis | shogun-install directory in my home folder | 16:58 |
sonne|work | and what did you set PYTHONPATH too? | 16:58 |
sonne|work | and LD_LIBRARY_PATH? | 16:58 |
nikolis | no i just set LD_LIBRARY_PATH | 16:59 |
nikolis | I followed the instraction from quick start file | 16:59 |
sonne|work | well that is for C/C++ stuff only you also need to set PYTHONPATH | 17:00 |
sonne|work | to the directory where modshogun.py is in (in you shogun-install dir) | 17:00 |
sonne|work | export PYTHONPATH=/some/path/to/some/dir/where/modshogun/is/in | 17:02 |
nikolis | more that informative thanks | 17:02 |
-!- soumyaC_ [uid15286@gateway/web/irccloud.com/x-txzwqkotkaqemdvr] has joined #shogun | 17:19 | |
chakri | can anyone provide a link to "entrance task list on github", please? | 17:20 |
jiaolong | https://github.com/shogun-toolbox/shogun/issues?labels=entrance&page=1&state=open | 17:27 |
chakri | thank you, jiaolong!! | 17:29 |
HeikoS1 | naywhayare: yeah unfortunately, noone applied :( | 17:56 |
naywhayare | HeikoS1: even so, it is possible that we may have a student working on it, and perhaps we could work together with this student a little bit | 17:59 |
naywhayare | so that he focuses on features that both of us can use | 17:59 |
HeikoS1 | naywhayare: I would love to do that | 17:59 |
HeikoS1 | naywhayare: can I see his proposal ? | 17:59 |
naywhayare | HeikoS1: how can I make that happen in Melange? | 17:59 |
naywhayare | add you as an mlpack mentor? | 18:00 |
naywhayare | gonna grab some lunch... back in about an hour | 18:03 |
HeikoS1 | naywhayare: enjoy. Yeah if you add me maybe then. Or just share it via mail or so? | 18:06 |
-!- newlido [~quassel@41.237.151.196] has joined #shogun | 18:11 | |
-!- soumyaC_ is now known as soumyaC | 18:17 | |
-!- lisitsyn [~lisitsyn@80.252.20.67] has joined #shogun | 18:23 | |
-!- abinash_panda [75ef5e6f@gateway/web/freenode/ip.117.239.94.111] has joined #shogun | 18:26 | |
-!- bumblebee [~quassel@202.78.172.162] has quit [Read error: Connection reset by peer] | 18:28 | |
-!- kislay [~kislay@117.239.94.112] has quit [Ping timeout: 240 seconds] | 18:44 | |
-!- kislay [~kislay@117.239.94.112] has joined #shogun | 18:47 | |
-!- pickle27 [d8bfe7da@gateway/web/freenode/ip.216.191.231.218] has joined #shogun | 18:57 | |
kislay | pickle27, Hi Kevin :) | 19:26 |
pickle27 | hey | 19:38 |
kislay | pickle27, Any head start over the integration issue. I looked over the cmake file though. | 19:41 |
kislay | pickle27, what should be my plan of action for now? | 19:42 |
-!- abinash_panda [75ef5e6f@gateway/web/freenode/ip.117.239.94.111] has quit [Ping timeout: 245 seconds] | 19:45 | |
pickle27 | kislay: sorry whats your github handle | 19:49 |
kislay | pickle27, my user name is kislayabhi | 19:49 |
@sonney2k | HeikoS1, pickle27 ! | 19:49 |
@sonney2k | hey pickle27! | 19:49 |
pickle27 | okay cool | 19:57 |
pickle27 | kislay: for now I would just keep working on entrance tasks and more ipython-notebooks! | 19:58 |
kislay | pickle27, yeah. I would love too :D | 19:59 |
@sonney2k | wiking, https://hubic.com/en/ | 20:18 |
@sonney2k | wiking, ^ sounds like a nice way for backups | 20:18 |
@sonney2k | pickle27, do you have any idea how we could use the the *.md files on the website to replace the docu? | 20:19 |
@wiking | sonney2k: lol | 20:19 |
@sonney2k | wiking, ? | 20:19 |
pickle27 | sorry what exactly do you want the md files from where and to replace what | 20:20 |
@wiking | sonney2k: very nice | 20:20 |
@sonney2k | pickle27, we have all this documentation sometimes as doxygen and sometimes on the website and sometimes as .md files in shogun's src | 20:20 |
@sonney2k | pickle27, IMHO we should just write the .md files in the src | 20:20 |
@sonney2k | and update these to have them on the website where needed | 20:21 |
pickle27 | yeah it should be easy enough to pull md from github for the site | 20:21 |
@wiking | sonney2k: how much is google's space? | 20:21 |
@sonney2k | pickle27, I mean we have that somehow | 20:21 |
@sonney2k | pickle27, http://shogun-toolbox.org/page/documentation/md/README | 20:21 |
@sonney2k | wiking, $99 for 10TB/ month | 20:22 |
pickle27 | yeah I know we have that | 20:22 |
@wiking | sonney2k: ok so this is insanely cheap | 20:22 |
@sonney2k | pickle27, problem is to come up with some concept where this stuff is located | 20:23 |
pickle27 | oh you want to replace the rest of the docs content so all we have is md on github and docygen | 20:23 |
@sonney2k | pickle27, and then start replacing things like FAQ | 20:23 |
@wiking | sonney2k: do they use some standard api to do it... or one has to use their own application...? | 20:23 |
pickle27 | oh like the whole site even | 20:23 |
@sonney2k | and a number of rapidly changing pages on the website with the .md files | 20:23 |
@sonney2k | pickle27, maybe not whole | 20:23 |
@sonney2k | but stuff like install/faq/readme/licence/authors or so | 20:24 |
pickle27 | okay I know roughly how that was done, in the next bit I will add to the readme how to make a page load from github | 20:24 |
@sonney2k | wiking, no idea - just saw that offer - it is from the ovh people (where shogun-toolbox.org is hosted and fatbot) | 20:24 |
@wiking | sonney2k: yeah i know... https://api.hubic.com/ | 20:25 |
@wiking | shame is that they are not a real cdn ;) | 20:25 |
* sonney2k smells abuses ;) | 20:26 | |
@sonney2k | pickle27, if you go to http://shogun-toolbox.org/page/documentation/md/README you will see that the links are b0rken - e.g. shogun-toolbox.org/page/documentation/md/doc/md/QUICKSTART.md | 20:32 |
pickle27 | kk | 20:33 |
@sonney2k | pickle27, so one thing to fix is links... | 20:33 |
@sonney2k | I guess the logic here is follow what we have in shogun's src | 20:33 |
@sonney2k | where we have doc/md | 20:33 |
@sonney2k | so doc/md/QUICKSTART.md | 20:33 |
@sonney2k | etc | 20:33 |
HeikoS1 | sonney2k: ? | 20:40 |
-!- shogun-notifier- [~irker@7nn.de] has joined #shogun | 21:09 | |
shogun-notifier- | shogun: lambday :develop * c2dd6d1 / tests/unit/statistics/QuadraticTimeMMD_unittest.cc: https://github.com/shogun-toolbox/shogun/commit/c2dd6d177d771623c251c2ac53040d603905e1d8 | 21:09 |
shogun-notifier- | shogun: added unit-tests for quadratic mmd with different number of samples | 21:09 |
shogun-notifier- | shogun: Heiko Strathmann :develop * 02c08c8 / tests/unit/statistics/QuadraticTimeMMD_unittest.cc: https://github.com/shogun-toolbox/shogun/commit/02c08c8498c94176c8e7050ad80686506f5e21a2 | 21:09 |
shogun-notifier- | shogun: Merge pull request #2079 from lambday/develop | 21:09 |
shogun-notifier- | shogun: | 21:09 |
shogun-notifier- | shogun: added unit-tests for quadratic mmd with different number of samples | 21:09 |
-!- iglesiasg [~iglesiasg@524AE0A7.cm-4-3d.dynamic.ziggo.nl] has joined #shogun | 21:26 | |
-!- mode/#shogun [+o iglesiasg] by ChanServ | 21:26 | |
-!- newlido [~quassel@41.237.151.196] has quit [Ping timeout: 246 seconds] | 21:27 | |
-!- PirosB3 [~pirosb3@cpc4-newt31-2-0-cust631.newt.cable.virginm.net] has joined #shogun | 21:40 | |
-!- sonne|osx [~sonne@f052235130.adsl.alicedsl.de] has joined #shogun | 21:58 | |
-!- PirosB3 [~pirosb3@cpc4-newt31-2-0-cust631.newt.cable.virginm.net] has quit [Quit: PirosB3] | 21:59 | |
-!- pickle27 [d8bfe7da@gateway/web/freenode/ip.216.191.231.218] has quit [Quit: Page closed] | 22:19 | |
-!- khalednasr [~k.nasr92@41.69.143.205] has joined #shogun | 22:22 | |
-!- Saurabh7_ [~Saurabh7@115.248.45.78] has quit [Ping timeout: 246 seconds] | 22:26 | |
-!- thoralf [~thoralf@91-66-33-156-dynip.superkabel.de] has joined #shogun | 22:42 | |
thoralf | Good evening ladies and gentlemen. | 22:43 |
-!- sonne|osx [~sonne@f052235130.adsl.alicedsl.de] has quit [Quit: sonne|osx] | 22:58 | |
-!- khalednasr [~k.nasr92@41.69.143.205] has quit [Ping timeout: 268 seconds] | 23:08 | |
-!- PirosB3 [~pirosb3@cpc4-newt31-2-0-cust631.newt.cable.virginm.net] has joined #shogun | 23:09 | |
-!- khalednasr [~k.nasr92@41.69.227.30] has joined #shogun | 23:17 | |
-!- HeikoS1 [~heiko@pat-191-250.internal.eduroam.ucl.ac.uk] has quit [Ping timeout: 246 seconds] | 23:27 | |
-!- jiaolong_ [553a5634@gateway/web/freenode/ip.85.58.86.52] has joined #shogun | 23:30 | |
shogun-notifier- | shogun-data: Fernando Iglesias :revert_knn_nanoflann_data * 6bfb763 / testsuite/python3-tests/classifier_knn_modular0.txt: https://github.com/shogun-toolbox/shogun-data/commit/6bfb763921282b3c551b5b45dc1a8d8cd325720d | 23:38 |
shogun-notifier- | shogun-data: Revert "Merge pull request #39 from dhruv13J/KDTree" | 23:38 |
shogun-notifier- | shogun-data: | 23:38 |
shogun-notifier- | shogun-data: This reverts commit 42c9a54f5afadbb61d27a2da09c6c56c0d0a57de, reversing | 23:38 |
shogun-notifier- | shogun-data: changes made to ff34862b9871856a08e80ec6dc3a51983fec56d1. | 23:38 |
shogun-notifier- | shogun-data: Fernando Iglesias :revert_knn_nanoflann_data * 0337522 / testsuite/python2-tests/ (2 files): https://github.com/shogun-toolbox/shogun-data/commit/03375226998bd32e584fc01c11303e489c072360 | 23:38 |
shogun-notifier- | shogun-data: Revert "Merge pull request #37 from dhruv13J/KDTree" | 23:38 |
shogun-notifier- | shogun-data: | 23:38 |
shogun-notifier- | shogun-data: This reverts commit 082eeb56ea20fc55085950e6114ef4e7849d438d, reversing | 23:38 |
shogun-notifier- | shogun-data: changes made to 6615cf007634595d459853bf4dc6f1a227d2450c. | 23:38 |
shogun-notifier- | shogun-data: Fernando Iglesias :master * 6bfb763 / testsuite/python3-tests/classifier_knn_modular0.txt: https://github.com/shogun-toolbox/shogun-data/commit/6bfb763921282b3c551b5b45dc1a8d8cd325720d | 23:39 |
shogun-notifier- | shogun-data: Revert "Merge pull request #39 from dhruv13J/KDTree" | 23:39 |
shogun-notifier- | shogun-data: | 23:39 |
shogun-notifier- | shogun-data: This reverts commit 42c9a54f5afadbb61d27a2da09c6c56c0d0a57de, reversing | 23:39 |
shogun-notifier- | shogun-data: changes made to ff34862b9871856a08e80ec6dc3a51983fec56d1. | 23:39 |
shogun-notifier- | shogun-data: Fernando Iglesias :master * 0337522 / testsuite/python2-tests/ (2 files): https://github.com/shogun-toolbox/shogun-data/commit/03375226998bd32e584fc01c11303e489c072360 | 23:39 |
shogun-notifier- | shogun-data: Revert "Merge pull request #37 from dhruv13J/KDTree" | 23:39 |
shogun-notifier- | shogun-data: | 23:39 |
shogun-notifier- | shogun-data: This reverts commit 082eeb56ea20fc55085950e6114ef4e7849d438d, reversing | 23:39 |
shogun-notifier- | shogun-data: changes made to 6615cf007634595d459853bf4dc6f1a227d2450c. | 23:39 |
shogun-notifier- | shogun-data: Fernando Iglesias :master * 618d1f1 / testsuite/ (3 files): https://github.com/shogun-toolbox/shogun-data/commit/618d1f149896c000317e96a2682396fa3a97ba8c | 23:39 |
shogun-notifier- | shogun-data: Merge pull request #40 from shogun-toolbox/revert_knn_nanoflann_data | 23:39 |
shogun-notifier- | shogun-data: | 23:39 |
shogun-notifier- | shogun-data: Revert knn nanoflann data | 23:39 |
shogun-notifier- | shogun: Fernando Iglesias :develop * a68d792 / data: https://github.com/shogun-toolbox/shogun/commit/a68d79208b589461e6dc9f928e86b57eebcdf630 | 23:43 |
shogun-notifier- | shogun: Update data version after KNN reverts | 23:43 |
-!- khalednasr [~k.nasr92@41.69.227.30] has quit [Quit: Leaving] | 23:47 | |
@iglesiasg | shogun-buildbot, force build 'modular_interfaces' | 23:54 |
shogun-buildbot | no such builder 'modular_interfaces' | 23:54 |
@iglesiasg | shogun-buildbot, force build 'deb3 - modular_interfaces' | 23:54 |
shogun-buildbot | build forced [ETA 38m51s] | 23:54 |
shogun-buildbot | I'll give a shout when the build finishes | 23:54 |
--- Log closed Tue Mar 25 00:00:53 2014 |
Generated by irclog2html.py 2.10.0 by Marius Gedminas - find it at mg.pov.lt!