--- Log opened Thu Jun 27 00:00:06 2013 | ||
shogun-buildbot | build #1271 of deb3 - modular_interfaces is complete: Success [build successful] Build details are at http://www.shogun-toolbox.org/buildbot/builders/deb3%20-%20modular_interfaces/builds/1271 | 00:09 |
---|---|---|
-!- travis-ci [~travis-ci@ec2-23-22-249-119.compute-1.amazonaws.com] has joined #shogun | 00:27 | |
travis-ci | [travis-ci] it's Heiko Strathmann's turn to pay the next round of drinks for the massacre he caused in shogun-toolbox/shogun: http://travis-ci.org/shogun-toolbox/shogun/builds/8481952 | 00:27 |
-!- travis-ci [~travis-ci@ec2-23-22-249-119.compute-1.amazonaws.com] has left #shogun [] | 00:27 | |
shogun-buildbot | build #1272 of deb3 - modular_interfaces is complete: Failure [failed test python_modular] Build details are at http://www.shogun-toolbox.org/buildbot/builders/deb3%20-%20modular_interfaces/builds/1272 blamelist: van51 <vangelis_51@hotmail.com> | 00:38 |
-!- iglesiasg [~nando@s83-179-44-135.cust.tele2.se] has quit [Quit: leaving] | 01:01 | |
-!- van51 [~van51@athedsl-408350.home.otenet.gr] has quit [Quit: Leaving.] | 01:51 | |
-!- FSCV [~FSCV@50.7.50.60] has quit [Quit: Leaving] | 02:14 | |
-!- shogun-notifier- [~irker@7nn.de] has quit [Quit: transmission timeout] | 02:23 | |
-!- nube [~rho@49.244.65.164] has quit [Quit: Leaving.] | 04:51 | |
-!- foulwall [~user@2001:da8:215:6100:bc02:b666:46de:6cba] has joined #shogun | 05:11 | |
-!- foulwall [~user@2001:da8:215:6100:bc02:b666:46de:6cba] has quit [Ping timeout: 264 seconds] | 05:24 | |
-!- nube [~rho@116.90.239.3] has joined #shogun | 05:41 | |
-!- foulwall [~user@2001:da8:215:6100:b8bb:f41d:16ab:399b] has joined #shogun | 06:17 | |
-!- nube1 [~rho@116.90.239.3] has joined #shogun | 06:17 | |
-!- nube [~rho@116.90.239.3] has quit [Read error: Connection reset by peer] | 06:17 | |
-!- foulwall [~user@2001:da8:215:6100:b8bb:f41d:16ab:399b] has quit [Ping timeout: 264 seconds] | 06:40 | |
-!- zxtx [~zv@cpe-66-68-190-37.austin.res.rr.com] has joined #shogun | 07:14 | |
-!- vgorbati [c3ee5cb1@gateway/web/freenode/ip.195.238.92.177] has joined #shogun | 09:05 | |
-!- Netsplit *.net <-> *.split quits: sanyam, nube1 | 09:35 | |
-!- Netsplit over, joins: nube1, sanyam | 09:37 | |
-!- foulwall [~user@2001:da8:215:6100:6c9c:8f7a:eb03:b755] has joined #shogun | 09:53 | |
foulwall | ping sonney2k | 09:57 |
-!- iglesiasg [~nando@s83-179-44-135.cust.tele2.se] has joined #shogun | 10:08 | |
-!- mode/#shogun [+o iglesiasg] by ChanServ | 10:08 | |
@iglesiasg | sonne|work: is the admin interface looking normal in your browser? | 10:10 |
@iglesiasg | the login and first page look without the common style of django admin | 10:11 |
-!- nube1 [~rho@116.90.239.3] has quit [Ping timeout: 246 seconds] | 10:16 | |
sonne|work | foulwall: hey there sec | 10:25 |
sonne|work | iglesiasg: yes all good | 10:25 |
@iglesiasg | program updated http://shogun-toolbox.org/page/Events/workshop2013_program | 10:25 |
@wiking | doh :) | 10:27 |
@wiking | sonne|work: i did not get a slot! | 10:27 |
foulwall | hey sonne|work, I rewrote a lot of code, I almost finished the js refractor, and have made a new toy data generator, and changed a bit of demo style... so I spent a bit more time on debugging... | 10:28 |
-!- nube [~rho@116.90.239.3] has joined #shogun | 10:29 | |
@iglesiasg | wiking: maybe we can squeeze a couple of slots and get one more with half an hour | 10:30 |
@wiking | i'd be great | 10:32 |
@iglesiasg | I think half an hour for my talk could be ok too | 10:33 |
@iglesiasg | let's ask sergey too about it | 10:34 |
-!- nube [~rho@116.90.239.3] has quit [Client Quit] | 10:34 | |
-!- hushell [~hushell@c-67-160-139-116.hsd1.or.comcast.net] has joined #shogun | 10:44 | |
@iglesiasg | hey hushell | 10:44 |
@iglesiasg | how are you? | 10:44 |
hushell | hey iglesiasg | 10:44 |
hushell | I am good | 10:44 |
hushell | How are you | 10:44 |
hushell | Do you have any idea: ImportError: cannot import name SerializableJsonFile | 10:45 |
@iglesiasg | I am fine too | 10:45 |
@iglesiasg | I don't know why it happens | 10:45 |
hushell | Failed at that example: error in serialization_complex_example.py | 10:45 |
@iglesiasg | lisitsyn mentioned that sonne|work had modified something about json | 10:45 |
@iglesiasg | maybe it is something related to that | 10:46 |
hushell | C interfaces are okay | 10:46 |
@iglesiasg | hushell: it is not your PR's fault I think | 10:46 |
hushell | okay. I always leave python check for Travis :) | 10:46 |
hushell | C# modular got problem too... | 10:48 |
-!- lisitsyn [~lisitsin@mxs.kg.ru] has joined #shogun | 11:12 | |
@iglesiasg | hushell: what is gram matrix? I have quite limited knowledge regarding kernels :) | 11:12 |
@iglesiasg | hushell: is it the same askernel matrix? | 11:12 |
@iglesiasg | hi lisitsyn, so what do you think about making our presentations in the workshop shorter and squeezing one more for wiking in the schedule? | 11:14 |
@iglesiasg | right now we have 2 of 45 min, let's make 3 of 30min? | 11:14 |
-!- lambday [67157c4c@gateway/web/freenode/ip.103.21.124.76] has joined #shogun | 11:15 | |
hushell | iglesiasg: just the kernel matrix :) | 11:15 |
lambday | sonne|work: morning :) | 11:15 |
hushell | iglesiasg: some people use that name, so I learned from them | 11:16 |
@iglesiasg | hushell: good to know! | 11:17 |
sonne|work | wiking: what do you want to talk about? | 11:17 |
hushell | iglesiasg: Are you going to present something in the workshop? | 11:17 |
sonne|work | lambday: moin moin! | 11:18 |
lambday | :) | 11:18 |
@iglesiasg | hushell: yeah! I will talk about structured learning in Shogun | 11:18 |
sonne|work | foulwall: I am not too happy about doing the toy data generation in the .js | 11:18 |
-!- nube [~rho@116.90.239.3] has joined #shogun | 11:18 | |
lisitsyn | iglesiasg: hmm I don't mind | 11:18 |
sonne|work | foulwall: I know it is a lot faster to run on the client but we have some datagenerator class that we might extend to have the data generation in one place | 11:19 |
hushell | sonne|work: Are you going to record the talks? | 11:19 |
@iglesiasg | hushell: let me know if you have any suggestion ;) | 11:19 |
@iglesiasg | that would be very nice indeed, to record them | 11:19 |
hushell | Yeah, I'd like to watch you guys online :) | 11:19 |
sonne|work | foulwall: and then can use it also for examples | 11:20 |
sonne|work | hushell: I wish we had but we don't have equipment | 11:21 |
hushell | sonne|work: any webcam works :) | 11:22 |
hushell | maybe we can have some unofficial videos | 11:22 |
lisitsyn | sonne|work: what equipment do we need? | 11:24 |
sonne|work | some camera to do video recording of the whole workshop | 11:25 |
sonne|work | as in talks / | 11:25 |
sonne|work | discussions | 11:25 |
sonne|work | foulwall: do you understand why it is better to have the data generators in python not .js ? | 11:26 |
sonne|work | lisitsyn: wiking also wants to give a talk | 11:26 |
foulwall | sonne|work: ok, I'll change that back to server, the generator is now modular, so it's easy to do that | 11:26 |
sonne|work | lisitsyn: luckily only us want to talk :D | 11:26 |
sonne|work | foulwall: very good! | 11:27 |
foulwall | sonne|work: why we need do that in python | 11:27 |
sonne|work | foulwall: for now we do it in python but later we do the data generation via shogun's DataGenerator class | 11:28 |
sonne|work | foulwall: the reason is that we can have stand-a-lone examples to run on the same data | 11:28 |
lisitsyn | sonne|work: well I will be there with some camera for sure but it won't last that long I guess | 11:28 |
foulwall | sonne|work: gotcha | 11:29 |
@iglesiasg | hushell: so what do you think about the features issue? | 11:34 |
@iglesiasg | hushell: apart from that, from my part the pull request is ready | 11:35 |
hushell | iglesiasg: I think we have had a m_models in machine, so now we can access m_features by m_models | 11:36 |
@iglesiasg | hushell: yes | 11:36 |
hushell | but you are right, we finally will move m_features to LinearStructMachine | 11:37 |
@iglesiasg | exactly | 11:37 |
hushell | Do you think we need keep m_features in LinearSOMachine right now? | 11:37 |
@iglesiasg | well not really need | 11:37 |
hushell | Because currently it will not be used anywhere | 11:38 |
@iglesiasg | but my point is, why to remove it now if we know we are going to get it back? | 11:38 |
hushell | okay | 11:38 |
@iglesiasg | ok, I understand that too | 11:38 |
hushell | I saw it was in LinearSOMachine but didn't registered | 11:38 |
@iglesiasg | so it is up to you, remove it if you prefer that for the moment :) I am fine with that too | 11:38 |
@iglesiasg | hushell: I probably forgot to do register it :( | 11:39 |
hushell | In shogun, for a NULL pointer, do we need to register it? | 11:39 |
hushell | I mean always NULL | 11:39 |
-!- foulwall [~user@2001:da8:215:6100:6c9c:8f7a:eb03:b755] has quit [Ping timeout: 264 seconds] | 11:39 | |
@iglesiasg | hushell: when you register a variable, you register the variable, not its value | 11:40 |
hushell | iglesiasg: That's ok. We have so many things here :) | 11:40 |
hushell | iglesiasg: okay. I was thinking this registration is kind of memory management | 11:40 |
@iglesiasg | hushell: I am not actually aware of everything the registration entails | 11:41 |
@iglesiasg | I know part of it is related to model selection and cross validation | 11:42 |
@iglesiasg | but most of the init methods (or register parameters) happen before the variables have even been given a value | 11:42 |
hushell | iglesiasg: maybe not keep this pointer in LinearSOMachine right now, we will refine this after the GSOC or in the end :) | 11:43 |
@iglesiasg | hushell: all right, as you prefer | 11:43 |
hushell | iglesiasg: Travis failed again :( | 11:43 |
hushell | iglesiasg: Good to know more about Shogun! | 11:44 |
hushell | iglesiasg: I am going to sleep now, let me know if there are other problems about the PR | 11:47 |
@iglesiasg | hushell: in python modular it is the SerializableJsonFile | 11:47 |
hushell | yes | 11:47 |
@iglesiasg | and in C# I have no idea | 11:47 |
@iglesiasg | python modular is not your PR's fault | 11:48 |
hushell | so every merge need to make sure no errors in Travis? | 11:48 |
@iglesiasg | hushell: ideally yes | 11:48 |
@iglesiasg | but your PR is ready now | 11:48 |
@iglesiasg | I am going to merge it soon | 11:48 |
@iglesiasg | I just want to know if the C# fail is due to your PR or not | 11:48 |
hushell | maybe wait other PR merges | 11:49 |
hushell | one thing about git | 11:49 |
@iglesiasg | yes? | 11:49 |
hushell | once I fetch from upstream, I just need to run git rebase upstream/develop, right? | 11:50 |
@iglesiasg | yes | 11:50 |
hushell | okay, then I think I did right thing | 11:50 |
@iglesiasg | that is preferable to doing pull as we avoid a commit | 11:50 |
@iglesiasg | hushell: so after this PR you are ready to be coding the new stuff right? | 11:51 |
hushell | as you suggested, I use --amend | 11:51 |
hushell | iglesiasg: Actually I am coding new stuffs | 11:51 |
@iglesiasg | the factor graph I think it was the first thing to do | 11:51 |
hushell | they are in parallel | 11:51 |
@iglesiasg | awesome! | 11:52 |
hushell | but I want to know more about the SSVM in Shogun, so I choose to do this first | 11:52 |
@iglesiasg | I understand | 11:52 |
hushell | yes, factor graph should be new stuff to Shogun | 11:53 |
@iglesiasg | I am afraid tha the C# fail might actually be due to your PR | 11:55 |
hushell | The OpenGM guys got a CVPR paper about their graphical model lib, I asked them about the factor graph implementation | 11:56 |
hushell | let me check | 11:56 |
hushell | Is it because I didn't update the Swig for C#? | 11:59 |
@iglesiasg | I have no idea | 11:59 |
hushell | everything about C# in src/interfaces/csharp_modular? | 12:01 |
@iglesiasg | I think that is mostly autogenerated | 12:01 |
@iglesiasg | the actual SWIG code is in src/interfaces/modular | 12:01 |
@iglesiasg | but you did not create any new class so there should be no need to update SWIG | 12:02 |
hushell | that's a bit strange | 12:03 |
@iglesiasg | the only thing I come up with is that maybe any of the new includes has interfered | 12:03 |
hushell | let me test C# tomorrow locally then we think about the merge again | 12:04 |
@iglesiasg | hushell: why did you change the position of the includes here btw? | 12:04 |
@iglesiasg | https://github.com/shogun-toolbox/shogun/pull/1185/files#L29R17 | 12:04 |
@iglesiasg | hushell: ok, have a rest now :) | 12:04 |
hushell | the define __H__ should be before include files, isn't it? | 12:05 |
hushell | otherwise the macro trick would not make sense | 12:06 |
@iglesiasg | I don't know, how is it done in the other files? :) | 12:06 |
@iglesiasg | it makes sense what you are saying though | 12:06 |
@iglesiasg | yeah ok that should be right | 12:06 |
hushell | #ifndef _STRUCTURED_MODEL__H__ || | 12:06 |
hushell | #define _STRUCTURED_MODEL__H__ ||- struct | 12:06 |
hushell | ||| TMultipleCPinfo [shogun | 12:06 |
hushell | #include <shogun/base/SGObject.h> ||| CResultSet [shogun] | 12:07 |
hushell | #include <shogun/features/Features.h> || | 12:07 |
hushell | #include <shogun/labels/StructuredLabels.h> | 12:07 |
hushell | sorry, I copy from vi directly | 12:07 |
hushell | this is an example from StructuredModel | 12:07 |
@iglesiasg | ok | 12:07 |
hushell | okay, I'll test the csharp thing tomorrow and see what's wrong over there | 12:08 |
@iglesiasg | sounds good | 12:08 |
hushell | I only test cmd and python_modular currently on my computer | 12:08 |
@iglesiasg | yeah, me too | 12:08 |
@iglesiasg | I check with travis and the buildbot the others hehe | 12:09 |
hushell | this is the best part of travis :D | 12:09 |
hushell | okay see you later | 12:09 |
@iglesiasg | bye, and very nice job! | 12:09 |
@iglesiasg | I like how the StructuredModel is getting | 12:09 |
@iglesiasg | rather clean | 12:09 |
hushell | very limited contribution indeed :( but I am happy to join you guys! Have a nice day | 12:10 |
-!- foulwall [~user@2001:da8:215:6100:d93a:3d1:5210:fbf] has joined #shogun | 12:17 | |
-!- lambday [67157c4c@gateway/web/freenode/ip.103.21.124.76] has quit [Ping timeout: 250 seconds] | 12:23 | |
-!- HeikoS [~heiko@nat-180-39.internal.eduroam.ucl.ac.uk] has joined #shogun | 12:35 | |
-!- mode/#shogun [+o HeikoS] by ChanServ | 12:35 | |
-!- lambday [67157f4c@gateway/web/freenode/ip.103.21.127.76] has joined #shogun | 12:38 | |
lambday | HeikoS: hi | 12:38 |
@HeikoS | lambday: hi! | 12:39 |
@HeikoS | saw your mail, sounds reasonable | 12:39 |
lambday | HeikoS: yes, I guess makes things more easy | 12:39 |
@HeikoS | yep totally | 12:39 |
lambday | we can use lanczos later then | 12:39 |
lambday | alright | 12:39 |
lambday | sending the PR | 12:39 |
@HeikoS | cool! | 12:39 |
lisitsyn | I heard someone is solving eigenproblems here | 12:40 |
lambday | lisitsyn: we we! :D | 12:42 |
@HeikoS | lambday: always ask the master on those :) | 12:43 |
@HeikoS | aka lisitsyn :) | 12:43 |
lisitsyn | I don't know anything but I worked on that a bit | 12:43 |
lisitsyn | HeikoS: I am thinking how can we re-use the things I have in tapkee | 12:44 |
@HeikoS | lisitsyn: we will need to solve gigantic eigenvalue problems at some point | 12:44 |
@HeikoS | but only largest/smallest | 12:44 |
@HeikoS | sparse matrices | 12:44 |
@HeikoS | with horrible horrible condition numbers | 12:44 |
@HeikoS | so we need something stable (priority over speed) | 12:45 |
lisitsyn | HeikoS: not arpack then | 12:45 |
lisitsyn | ;D | 12:45 |
lisitsyn | but you still need something with reverse interface | 12:45 |
lisitsyn | or specialized on sparse | 12:45 |
lisitsyn | HeikoS: http://www.ecs.umass.edu/~polizzi/feast/ this one could be more numericall stable | 12:46 |
lisitsyn | HeikoS: I am thinking if I can extract eigen things from tapkee to separate lib.. | 12:47 |
lisitsyn | then we can work together on that | 12:47 |
@HeikoS | lisitsyn: ok would be good | 12:47 |
@HeikoS | I can provide a matrix to test it | 12:47 |
lisitsyn | HeikoS: I mean I already did some things so this should not be lost | 12:48 |
@HeikoS | lisitsyn, lambday its very important that all the eigensolvers are based on linear operators that are applied to vectors rather than storing matrices btw | 12:48 |
lisitsyn | but using dimension reduction library from some other code is wrong | 12:48 |
lambday | HeikoS: yes | 12:48 |
lisitsyn | HeikoS: well that's arpack and feast | 12:48 |
lisitsyn | I am not sure I know other solvers with reverse interface | 12:49 |
lisitsyn | http://www.comp-phys.org/software/ietl/ | 12:49 |
lisitsyn | HeikoS: MKL uses FEAST btw | 12:50 |
lisitsyn | starting from this year february | 12:50 |
@HeikoS | no idea about these things :) | 12:51 |
@HeikoS | matlab fails on the matrices I have here btw | 12:51 |
@HeikoS | I want things up to machine precision >:-| | 12:51 |
lisitsyn | HeikoS: but you know what MKL is, right? | 12:51 |
@HeikoS | lisitsyn: not really, some sparsity stuff right? :D | 12:52 |
lisitsyn | HeikoS: no, huge intel librar | 12:52 |
lisitsyn | that does everything | 12:52 |
@HeikoS | ah this one | 12:52 |
lisitsyn | yes | 12:52 |
@HeikoS | sorry | 12:52 |
@HeikoS | no I don tknow that | 12:52 |
lambday | I should checkout FEAST then.. :-/ | 12:53 |
lambday | lisitsyn: what do you use in tapkee? | 12:53 |
lambday | HeikoS: I sent the PR, please have a look | 12:53 |
lisitsyn | lambday: arpack, randomized and dense method from eigen3 | 12:54 |
lambday | okay | 12:54 |
lisitsyn | randomized is out :D | 12:55 |
lisitsyn | dense is too | 12:55 |
lambday | yeah :( | 12:55 |
lambday | we want sparse :( | 12:55 |
lambday | FEAST has support for that? | 12:55 |
lisitsyn | lambday: arpack supports sparse | 12:56 |
lisitsyn | feast too | 12:56 |
@HeikoS | lambday: why do you add all those debug messages to the constructors? do you have memory leaks often=? | 13:00 |
@HeikoS | lambday: you can compile with --enable-trace-mallocs then shogun will print all allocated objects at the end of a program so its easy to track which ones are still in mem | 13:01 |
lambday | umm.. initially I had when I was trying to get SG_REF/UNREF.. now I just copy paste :D | 13:01 |
@HeikoS | olk | 13:01 |
lambday | I thought of removing all of them when the whole code is ready | 13:01 |
lambday | :-/ | 13:01 |
@HeikoS | they dont hurt us just wondering :) | 13:01 |
@HeikoS | nono leave them its fine | 13:01 |
@HeikoS | just wonderiung | 13:01 |
lambday | alright, I'll try enable-trace-mallocs | 13:02 |
@HeikoS | lambday: hey I just read about this indian mathematician that went to england before WW2, do you know about that guy? | 13:02 |
lambday | umm... no! :( | 13:02 |
lambday | I mean | 13:02 |
@HeikoS | https://en.wikipedia.org/wiki/Srinivasa_Ramanujan | 13:02 |
lambday | Ramanujan | 13:02 |
lambday | ohh | 13:02 |
@HeikoS | awesome guy! | 13:02 |
lambday | yeah !! | 13:02 |
lambday | he's a genius! | 13:02 |
lambday | absolute genius! I heard stories that when he was on his deathbed, someone came by a car to meet him... he looked up the window.. when that guy entered, he told him that your car plate number is a multiple of two huge prime numbers | 13:03 |
lambday | and then he told what those prime numbers are | 13:04 |
lambday | lol | 13:04 |
lambday | (not sure how far these stories are true anyway) | 13:04 |
@iglesiasg | it is a pity he passed away so young :( | 13:05 |
lambday | yes :( | 13:05 |
@HeikoS | lambday: yeah | 13:05 |
@HeikoS | lambday: currently reading a book on prime numbers where he had a chapter | 13:06 |
lambday | ahan! | 13:06 |
@HeikoS | lambday: I think storing the eigenvalues in the class is a bit strange if you provide a matrix in the compute method | 13:07 |
@HeikoS | lambday: either make the class store the matrix (and the eigenvalues) or return them and dont store them | 13:07 |
@HeikoS | lambday: you can do this via references | 13:07 |
@HeikoS | lisitsyn: do we have typemaps for std::pair? | 13:07 |
lambday | yes I thought of that | 13:07 |
lambday | okay then I'll change it | 13:08 |
@HeikoS | lambday: ok | 13:08 |
@HeikoS | and please do some unit tests for the non full rank systems | 13:08 |
lambday | okay | 13:08 |
@HeikoS | note that LLT can solve over and underdetermined systems but not those with negative eigenvalues | 13:08 |
lambday | yes.. | 13:09 |
lambday | I was thinking of adding a check in the linear solver I added (for eigenvalues), but then thought it will be too slow :-/ | 13:09 |
@HeikoS | lambday: man if you continue in that pace the project will be done in 2 weeks :) | 13:09 |
@HeikoS | lambday: LLT class should do this in fact | 13:09 |
@HeikoS | try it with a non-psd matrix | 13:10 |
@HeikoS | you can probably ask it whether the matrix is fine | 13:10 |
lambday | HeikoS: haha no don't worry my thesis adviser has threatened me to kick me out (because I haven't been doing any of his work since last 2 weeks), so I'll be a bit slow for sometime :( | 13:10 |
@HeikoS | lambday: careful then! uni is more important than GSoC in the long run | 13:11 |
@HeikoS | lambday: also you work very fast, so dont worry to slow down a bit | 13:11 |
lambday | HeikoS: LLT class :-/ | 13:11 |
lambday | HeikoS: haha yes.. | 13:11 |
@HeikoS | although I really like that things are going so well! | 13:11 |
@HeikoS | lambday: looks like it doesnt catch that | 13:12 |
lambday | HeikoS: you're the mastermind.. I just execute.. I too am really happy with how this is going too :) so, I didn't get much depressed with what thesis adviser told me (I was supposed to be more depressed) | 13:12 |
@HeikoS | lambday: then its fine, user sets the solver and if user provides wrong matrix its his problem | 13:13 |
lambday | HeikoS: yes that's what I thought | 13:13 |
@HeikoS | the std solver should be able to solve everything though, so probably QR/SVD | 13:13 |
lambday | user shouldn't call LLT for non psd matrices | 13:13 |
@HeikoS | lambday: yep LLT should be a conscious decision | 13:13 |
@HeikoS | lambday: there is ComputationInfo info() though | 13:14 |
@HeikoS | you could check this in any case | 13:14 |
lambday | in Eigen3? | 13:14 |
@HeikoS | lambday: its a method of LLT | 13:14 |
@HeikoS | LLT::info | 13:15 |
lambday | okay.. yes I'll check it | 13:15 |
@HeikoS | only check for Success | 13:15 |
@HeikoS | or maybe just print error type in case of no success | 13:15 |
lambday | ah, that would help me print an error msg if user mistakenly calls LLT on a nonpsd matrices | 13:15 |
@HeikoS | there are 3 cases | 13:15 |
lambday | yes yes! | 13:15 |
@HeikoS | lambday: ok gotta do some work now, will be back later in the day | 13:16 |
lambday | HeikoS: alright.. I'll check it | 13:16 |
lambday | see you :) | 13:16 |
@HeikoS | lambday: Ill keep an eye on PRs though | 13:16 |
@HeikoS | but I have to switch off irc, otherwise I wont work :D | 13:16 |
lambday | okay :) | 13:16 |
@HeikoS | so see you | 13:16 |
lambday | hehe | 13:16 |
-!- HeikoS [~heiko@nat-180-39.internal.eduroam.ucl.ac.uk] has quit [Quit: Leaving.] | 13:16 | |
-!- nube [~rho@116.90.239.3] has quit [Quit: Leaving.] | 13:17 | |
-!- nube [~rho@116.90.239.3] has joined #shogun | 13:17 | |
-!- gsomix [~Miranda@185.6.239.70] has joined #shogun | 13:45 | |
gsomix | *ba dum tss* | 13:45 |
gsomix | pain is gone | 13:47 |
gsomix | exams are passed :) | 13:47 |
@iglesiasg | congrats gsomix | 13:47 |
sonne|work | gsomix: hurray! | 13:51 |
-!- nube [~rho@116.90.239.3] has quit [Quit: Leaving.] | 13:56 | |
-!- sonne|work [~sonnenbu@sams-office-nat.tomtomgroup.com] has quit [Quit: Leaving.] | 14:11 | |
-!- sonne|work [~sonnenbu@sams-office-nat.tomtomgroup.com] has joined #shogun | 14:11 | |
-!- nube [~rho@49.244.14.83] has joined #shogun | 15:01 | |
-!- lambday [67157f4c@gateway/web/freenode/ip.103.21.127.76] has quit [] | 15:06 | |
-!- thoralf [~thoralf@enki.zib.de] has joined #shogun | 15:31 | |
thoralf | Hello everybody. | 15:31 |
sonne|work | hello hello | 15:33 |
-!- nube [~rho@49.244.14.83] has quit [Ping timeout: 264 seconds] | 15:47 | |
-!- zxtx [~zv@cpe-66-68-190-37.austin.res.rr.com] has quit [Ping timeout: 256 seconds] | 15:51 | |
-!- foulwall [~user@2001:da8:215:6100:d93a:3d1:5210:fbf] has quit [Ping timeout: 264 seconds] | 16:05 | |
-!- pickle27 [~Kevin@d67-193-243-174.home3.cgocable.net] has joined #shogun | 16:11 | |
-!- gsomix [~Miranda@185.6.239.70] has quit [Quit: Miranda IM! Smaller, Faster, Easier. http://miranda-im.org] | 16:25 | |
-!- FSCV [~FSCV@50.7.50.60] has joined #shogun | 16:26 | |
-!- nube [~rho@49.244.123.126] has joined #shogun | 16:38 | |
-!- FSCV_ [~FSCV@50.7.50.60] has joined #shogun | 16:39 | |
-!- iglesiasg [~nando@s83-179-44-135.cust.tele2.se] has quit [Quit: leaving] | 16:39 | |
-!- FSCV [~FSCV@50.7.50.60] has quit [Ping timeout: 252 seconds] | 16:41 | |
-!- vgorbati [c3ee5cb1@gateway/web/freenode/ip.195.238.92.177] has quit [Quit: Page closed] | 16:57 | |
@wiking | lisitsyn: ping | 17:06 |
@wiking | sonne|work: pong | 17:06 |
@wiking | sonne|work: latent (so) solvers | 17:10 |
@wiking | sonne|work: and maybe about pending design issues of shogun ;P | 17:11 |
-!- nube [~rho@49.244.123.126] has quit [Ping timeout: 248 seconds] | 17:13 | |
-!- lisitsyn [~lisitsin@mxs.kg.ru] has quit [Quit: Leaving.] | 17:13 | |
-!- shogun-notifier- [~irker@7nn.de] has joined #shogun | 17:17 | |
shogun-notifier- | shogun: lambday :develop * a7eccd9 / / (5 files): https://github.com/shogun-toolbox/shogun/commit/a7eccd95928426bea68df6c784d351fa86cf465c | 17:17 |
shogun-notifier- | shogun: EigenSolver added, implementation using Eigen3 added | 17:17 |
shogun-notifier- | shogun: lambday :develop * 8333c76 / / (6 files): https://github.com/shogun-toolbox/shogun/commit/8333c76efcbfd9489c7107ea87befed9f889e6cf | 17:17 |
shogun-notifier- | shogun: minor changes in EigenSolver, LinearSolver | 17:17 |
shogun-notifier- | shogun: Heiko Strathmann :develop * faed169 / / (6 files): https://github.com/shogun-toolbox/shogun/commit/faed169e23c96274cb6b32c32dbd5a7f650784b6 | 17:17 |
shogun-notifier- | shogun: Merge pull request #1192 from lambday/feature/log_determinant | 17:17 |
shogun-notifier- | shogun: | 17:17 |
shogun-notifier- | shogun: EigenSolver added, implementation using Eigen3 added | 17:17 |
-!- lambday [67157d4c@gateway/web/freenode/ip.103.21.125.76] has joined #shogun | 17:22 | |
shogun-buildbot | build #983 of cyg1 - libshogun is complete: Failure [failed configure] Build details are at http://www.shogun-toolbox.org/buildbot/builders/cyg1%20-%20libshogun/builds/983 blamelist: Heiko Strathmann <heiko.strathmann@gmail.com>, lambday <heavensdevil6909@gmail.com> | 17:24 |
-!- foulwall [~user@2001:da8:215:c252:d10:e269:ac4b:bc2f] has joined #shogun | 17:30 | |
shogun-buildbot | build #1151 of bsd1 - libshogun is complete: Failure [failed test_1] Build details are at http://www.shogun-toolbox.org/buildbot/builders/bsd1%20-%20libshogun/builds/1151 blamelist: Heiko Strathmann <heiko.strathmann@gmail.com>, lambday <heavensdevil6909@gmail.com> | 17:31 |
shogun-buildbot | build #1152 of bsd1 - libshogun is complete: Failure [failed test_1] Build details are at http://www.shogun-toolbox.org/buildbot/builders/bsd1%20-%20libshogun/builds/1152 blamelist: lambday <heavensdevil6909@gmail.com> | 17:42 |
-!- lambday [67157d4c@gateway/web/freenode/ip.103.21.125.76] has quit [] | 17:43 | |
-!- travis-ci [~travis-ci@ec2-23-20-149-45.compute-1.amazonaws.com] has joined #shogun | 17:44 | |
travis-ci | [travis-ci] it's Heiko Strathmann's turn to pay the next round of drinks for the massacre he caused in shogun-toolbox/shogun: http://travis-ci.org/shogun-toolbox/shogun/builds/8508507 | 17:44 |
-!- travis-ci [~travis-ci@ec2-23-20-149-45.compute-1.amazonaws.com] has left #shogun [] | 17:44 | |
shogun-buildbot | build #984 of cyg1 - libshogun is complete: Failure [failed compile] Build details are at http://www.shogun-toolbox.org/buildbot/builders/cyg1%20-%20libshogun/builds/984 blamelist: lambday <heavensdevil6909@gmail.com> | 17:46 |
shogun-buildbot | build #1273 of deb3 - modular_interfaces is complete: Success [build successful] Build details are at http://www.shogun-toolbox.org/buildbot/builders/deb3%20-%20modular_interfaces/builds/1273 | 18:06 |
-!- pickle27 [~Kevin@d67-193-243-174.home3.cgocable.net] has quit [Remote host closed the connection] | 18:28 | |
-!- pickle27 [~Kevin@d67-193-243-174.home3.cgocable.net] has joined #shogun | 18:30 | |
shogun-buildbot | build #1274 of deb3 - modular_interfaces is complete: Failure [failed test python_modular] Build details are at http://www.shogun-toolbox.org/buildbot/builders/deb3%20-%20modular_interfaces/builds/1274 blamelist: lambday <heavensdevil6909@gmail.com> | 18:33 |
-!- foulwall [~user@2001:da8:215:c252:d10:e269:ac4b:bc2f] has quit [Ping timeout: 245 seconds] | 19:01 | |
-!- foulwall [~user@2001:da8:215:c252:9a3:d4b7:6d1e:6ecc] has joined #shogun | 19:58 | |
-!- foulwall [~user@2001:da8:215:c252:9a3:d4b7:6d1e:6ecc] has quit [Ping timeout: 264 seconds] | 20:04 | |
-!- hushell [~hushell@c-67-160-139-116.hsd1.or.comcast.net] has quit [Ping timeout: 248 seconds] | 20:05 | |
-!- lisitsyn [~lisitsyn@37.61.181.61] has joined #shogun | 20:12 | |
lisitsyn | wiking: pong | 20:12 |
-!- shogun-notifier- [~irker@7nn.de] has quit [Quit: transmission timeout] | 20:17 | |
-!- HeikoS [~heiko@nat-180-39.internal.eduroam.ucl.ac.uk] has joined #shogun | 20:18 | |
-!- mode/#shogun [+o HeikoS] by ChanServ | 20:18 | |
@HeikoS | sonney2k: around? | 20:18 |
@HeikoS | sonne|work: around? | 20:23 |
@HeikoS | lets do the PO tomorrow then, going home | 20:29 |
@HeikoS | bye all | 20:30 |
-!- HeikoS [~heiko@nat-180-39.internal.eduroam.ucl.ac.uk] has quit [Quit: Leaving.] | 20:30 | |
-!- gsomix [~Miranda@80.234.17.77] has joined #shogun | 20:41 | |
-!- hushell [~hushell@67.23.197.94] has joined #shogun | 20:46 | |
-!- hushell [~hushell@67.23.197.94] has quit [Ping timeout: 276 seconds] | 21:03 | |
-!- van51 [~van51@athedsl-408350.home.otenet.gr] has joined #shogun | 21:25 | |
-!- gsomix [~Miranda@80.234.17.77] has quit [Quit: Miranda IM! Smaller, Faster, Easier. http://miranda-im.org] | 21:53 | |
lisitsyn | pickle27: hey! | 22:30 |
pickle27 | lisitsyn: hey! | 22:30 |
lisitsyn | pickle27: how is it going? | 22:30 |
pickle27 | lisitsyn: good! I finished my code that tests jadiag vs the R version | 22:31 |
pickle27 | and they matched! | 22:31 |
pickle27 | today I have been working on ffdiag and I think I am getting close to being done | 22:31 |
lisitsyn | pickle27: cool! | 22:31 |
lisitsyn | more time to get some examples done | 22:32 |
pickle27 | lisitsyn: yeah! | 22:32 |
pickle27 | lisitsyn: I think I am going to try and get a BSS example going after ffdiag and before porting uwedge | 22:33 |
lisitsyn | pickle27: what example? | 22:33 |
pickle27 | lisitsyn: for now I just want it to work on the scikit-learn example data | 22:33 |
pickle27 | I haven't got to it yet but it looks like ADJ is just one of the step for BSS so I think theres more to do | 22:34 |
pickle27 | and I want to get started looking at that part | 22:35 |
pickle27 | lisitsyn: http://scikit-learn.org/stable/modules/decomposition.html#independent-component-analysis-ica | 22:37 |
lisitsyn | pickle27: I like that mixed signal thing | 22:38 |
pickle27 | lisitsyn: I have the example ported to C++ I just need to finish the methods and run them on it | 22:39 |
lisitsyn | pickle27: ohh you don't like easy ways ;) | 22:39 |
pickle27 | lisitsyn: oh like running the whole thing in python? | 22:40 |
lisitsyn | yeah would be easier for you | 22:40 |
pickle27 | lisitsyn: hah true that would have been easier | 22:40 |
pickle27 | lisitsyn: although it'll be easier to debug now that its all c++ | 22:40 |
lisitsyn | pickle27: same thing if you are using gdb | 22:41 |
lisitsyn | pickle27: ah btw try cgdb | 22:41 |
pickle27 | lisitsyn: yeah Im not that great with gdb I tend to just use printing and cin.get for debugginh | 22:42 |
lisitsyn | pickle27: oh that's mostly gdb that is not nice enough | 22:42 |
lisitsyn | I mean it is rather difficult to see a matrix there | 22:43 |
pickle27 | I'll have to check out cgdb | 22:43 |
lisitsyn | pickle27: it is a editor (or text viewer) combined with gdb | 22:44 |
lisitsyn | so you may put breakpoints | 22:45 |
@sonney2k | van51, how is it going? | 22:45 |
lisitsyn | using just spacebar | 22:45 |
pickle27 | cool yeah I've used gdb a few times | 22:45 |
van51 | sonney2k: it's been a slow day today tbh :$ | 22:46 |
van51 | sonney2k: now I'm converting the densematrix to a sparse representation like you commente | 22:46 |
van51 | d | 22:46 |
van51 | and after that I'm thinking of starting to see the streaming features | 22:46 |
van51 | btw, is there a specific way I should expect a document collection to be represented? | 22:47 |
-!- iglesiasg [~nando@s83-179-44-135.cust.tele2.se] has joined #shogun | 22:56 | |
-!- mode/#shogun [+o iglesiasg] by ChanServ | 22:56 | |
-!- FSCV_ [~FSCV@50.7.50.60] has quit [Quit: This computer has gone to sleep] | 22:58 | |
pickle27 | lisitsyn: | 23:00 |
lisitsyn | pickle27: yes | 23:00 |
pickle27 | lisitsyn: does memory from memcpy need to be free'd manually? | 23:00 |
lisitsyn | pickle27: what memory? | 23:01 |
lisitsyn | destination? | 23:01 |
pickle27 | I've got a seg fault that throws after my code runs and this memcpy I've got is the only thing I suspect | 23:01 |
pickle27 | its from one SGNDArray to another | 23:01 |
van51 | does a SGSparseMatrix expect its SGSparseVectors to have the same size? | 23:01 |
lisitsyn | pickle27: hmm still not clear | 23:02 |
van51 | because I see it can take a num_feat as a parameter, but it's not documented on its meaning | 23:02 |
pickle27 | this is the code I have: | 23:03 |
van51 | does it mean the size of the vectors or the size of the dimension they represent? | 23:03 |
pickle27 | SGNDArray<float64_t> C(C0.dims,3); | 23:04 |
pickle27 | memcpy(C.array, C0.array, C0.dims[0]*C0.dims[1]*C0.dims[2]*sizeof(float64_t)); | 23:04 |
pickle27 | lisitsyn: ^ | 23:04 |
pickle27 | C0 is passed by ref to the function | 23:04 |
lisitsyn | pickle27: I expect C0.dims may be deleted before C | 23:05 |
lisitsyn | memcpy part looks ok | 23:05 |
pickle27 | ahh so I should do a deep copy of that too | 23:05 |
pickle27 | okay I'll try that | 23:05 |
-!- FSCV_ [~FSCV@187.210.54.166] has joined #shogun | 23:06 | |
-!- FSCV_ [~FSCV@187.210.54.166] has quit [Client Quit] | 23:06 | |
pickle27 | lisitsyn: that fixed it thanks | 23:07 |
pickle27 | lisitsyn: I don't know if I would have caught that | 23:07 |
lisitsyn | pickle27: two years of catching such errors ;) | 23:07 |
lisitsyn | slowed down though | 23:08 |
@wiking | lisitsyn: yo | 23:14 |
@wiking | brb | 23:14 |
@wiking | here | 23:15 |
@wiking | lisitsyn: can i borrow 15 minutes of your time? | 23:16 |
lisitsyn | wiking: yes sure | 23:16 |
@wiking | lisitsyn: thnx heaps | 23:16 |
lisitsyn | what's up? | 23:16 |
@wiking | i mean your workshop presentation time ;) | 23:16 |
lisitsyn | wiking: ahha :D | 23:17 |
lisitsyn | yes sure | 23:17 |
lisitsyn | wiking: what do you want to present? | 23:17 |
@wiking | well mostly the yet to be commited shit i was working between sept-dec | 23:17 |
@wiking | give a little demo | 23:17 |
@wiking | what i was doing with it | 23:17 |
lisitsyn | wiking: DVOR images please | 23:17 |
@wiking | DVOR? | 23:18 |
lisitsyn | I can't forget that category | 23:18 |
@wiking | ahhahahah | 23:18 |
lisitsyn | remember that? | 23:18 |
@wiking | indeed | 23:18 |
@wiking | i doo | 23:18 |
@wiking | it's watermarked in my eyes | 23:18 |
@wiking | vividly | 23:18 |
@sonney2k | wiking, so what did you want to talk about? | 23:18 |
lisitsyn | haha yes same too | 23:18 |
lisitsyn | I just wake up and see that | 23:18 |
@wiking | sonney2k: so if possible would like to give a little demo of the stuff i was doing with shogun | 23:19 |
@sonney2k | van51, I would rather want you to finish up the stuff and get it merged ASAP | 23:19 |
@iglesiasg | wiking: latent stuff? | 23:19 |
@wiking | iglesiasg: yep | 23:20 |
@iglesiasg | wiking: yeah, do a latent presentation! | 23:20 |
@sonney2k | van51, I was expecting that this takes you 2 days no more... | 23:20 |
@iglesiasg | that sounds mighty | 23:20 |
lisitsyn | wiking: iglesiasg: latent like with some gay allusions ;) | 23:20 |
@wiking | iglesiasg: yeah .. but unfortunately i cannot pythonize it :( | 23:21 |
@wiking | lisitsyn: :D | 23:21 |
lisitsyn | wiking: what is it in? | 23:21 |
@wiking | lisitsyn: parts detection in image | 23:21 |
lisitsyn | wiking: I mean written in | 23:21 |
lisitsyn | you said you cannot pythonize | 23:21 |
@wiking | ah yeah | 23:21 |
@wiking | remember the discussion back in october | 23:21 |
@wiking | about what can be and what cannot be directorized? :) | 23:21 |
@wiking | by swig | 23:22 |
@wiking | :) | 23:22 |
lisitsyn | wiking: so C++? | 23:22 |
@wiking | yep | 23:22 |
@wiking | i could create wrapper classes for it | 23:22 |
@wiking | but i doubt that anybody would ever use it again in the same way | 23:22 |
@sonney2k | van51, it would be cool to move on | 23:24 |
@sonney2k | wiking, sure but need to schedule it lisitsyn, iglesiasg any thoughts? | 23:25 |
lisitsyn | sonney2k: I am ok to give my 15 minutes | 23:25 |
@wiking | sonney2k: only thing is that maybe presenting after iglesiasg would make the most sense, if iglesiasg don't mind, as i'll be referring to SOSVM... | 23:26 |
-!- FSCV [~FSCV@50.7.50.60] has joined #shogun | 23:28 | |
@wiking | btw anybody realised that u can edit code on github page directly? :) | 23:29 |
@wiking | i mean actually edit HEAD and do commit there straight away | 23:29 |
@sonney2k | wiking, ok so then sergey should be last | 23:29 |
@sonney2k | and you inbetween | 23:29 |
@wiking | k | 23:29 |
@sonney2k | wiking, do you have a real demo? | 23:30 |
@sonney2k | or slides? | 23:30 |
lisitsyn | wiking: I use it | 23:30 |
@wiking | sonney2k: yeah i'm using it | 23:30 |
lisitsyn | sometimes | 23:30 |
@wiking | sonney2k: i mean i have a real application | 23:30 |
@wiking | sonney2k: that is actually being used | 23:30 |
@sonney2k | wiking, that you can show? | 23:30 |
@sonney2k | ok | 23:30 |
lisitsyn | sonney2k: hopefully not these images I have seen once | 23:30 |
@sonney2k | iglesiasg, would it be OK if you do just 30 minutes SO, followed by wiking with an SO application? | 23:31 |
@iglesiasg | sonney2k: sure | 23:31 |
@sonney2k | wiking, what is the title of the talk then? | 23:37 |
@sonney2k | iglesiasg, ok then please adjust the schedule on the website | 23:37 |
@iglesiasg | wiking: give me a title please :) | 23:38 |
@wiking | sonney2k: mmm | 23:42 |
@sonney2k | wiking, ? | 23:43 |
@wiking | iglesiasg: (latent)SOSVM in practice | 23:43 |
van51 | sonney2k: yeah I know.. I have to pick up some speed | 23:50 |
@sonney2k | van51, otherwise we cannot do the fun part | 23:52 |
@sonney2k | it would be stupid to get stuck in the easiest thing already | 23:52 |
@sonney2k | so van51 go go go! | 23:52 |
@sonney2k | alrighty | 23:53 |
* sonney2k dead | 23:53 | |
@sonney2k | cu | 23:53 |
@sonney2k | van51, btw regarding sparse vectors just alloc number of ngrams/words many elements | 23:53 |
@sonney2k | and fill them | 23:53 |
van51 | sonney2k: that would leave some spaced unused, wouldn't it? | 23:54 |
@iglesiasg | Spain to the final guys, we are there! | 23:59 |
@iglesiasg | well not me ... whatever | 23:59 |
--- Log closed Fri Jun 28 00:00:08 2013 |
Generated by irclog2html.py 2.10.0 by Marius Gedminas - find it at mg.pov.lt!