--- Log opened Mon Jul 01 00:00:12 2013 | ||
-!- van51 [~van51@athedsl-408350.home.otenet.gr] has joined #shogun | 00:01 | |
shogun-buildbot | build #1164 of bsd1 - libshogun is complete: Failure [failed test_1] Build details are at http://www.shogun-toolbox.org/buildbot/builders/bsd1%20-%20libshogun/builds/1164 blamelist: Sergey Lisitsyn <lisitsyn.s.o@gmail.com> | 00:04 |
---|---|---|
shogun-buildbot | build #1287 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/1287 blamelist: Roman Votyakov <votjakovr@gmail.com> | 00:39 |
shogun-buildbot | build #1288 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/1288 blamelist: Sergey Lisitsyn <lisitsyn.s.o@gmail.com> | 00:45 |
-!- van51 [~van51@athedsl-408350.home.otenet.gr] has quit [Quit: Leaving.] | 01:25 | |
-!- pickle27 [~Kevin@d67-193-243-174.home3.cgocable.net] has quit [Quit: Leaving] | 01:57 | |
-!- nube [~rho@49.244.55.95] has joined #shogun | 02:28 | |
-!- shogun-notifier- [~irker@7nn.de] has quit [Quit: transmission timeout] | 02:38 | |
-!- foulwall|mobile [~foulwall@117.136.0.220] has joined #shogun | 02:57 | |
-!- foulwall|mobile [~foulwall@117.136.0.220] has quit [Read error: Connection reset by peer] | 03:33 | |
-!- foulwall|mobile [~foulwall@117.136.0.220] has joined #shogun | 03:33 | |
-!- foulwall|mobile [~foulwall@117.136.0.220] has quit [Read error: Connection reset by peer] | 03:40 | |
-!- foulwall|mobile [~foulwall@117.136.0.220] has joined #shogun | 03:41 | |
-!- nube [~rho@49.244.55.95] has quit [Quit: Leaving.] | 03:50 | |
-!- foulwall|mobile [~foulwall@117.136.0.220] has quit [Ping timeout: 268 seconds] | 03:53 | |
-!- zxtx [~zv@cpe-66-68-190-37.austin.res.rr.com] has quit [Ping timeout: 256 seconds] | 04:27 | |
shogun-buildbot | build #444 of nightly_default is complete: Failure [failed test] Build details are at http://www.shogun-toolbox.org/buildbot/builders/nightly_default/builds/444 | 04:35 |
-!- foulwall|mobile [~foulwall@117.136.0.221] has joined #shogun | 04:52 | |
-!- fw2k [~foulwall@117.136.0.221] has joined #shogun | 05:37 | |
-!- fw2k [~foulwall@117.136.0.221] has quit [Ping timeout: 246 seconds] | 06:03 | |
-!- foulwall|mobile [~foulwall@117.136.0.221] has quit [Ping timeout: 246 seconds] | 06:03 | |
-!- zxtx [~zv@cpe-66-68-190-37.austin.res.rr.com] has joined #shogun | 06:29 | |
-!- nube [~rho@116.90.239.3] has joined #shogun | 06:35 | |
-!- lisitsyn [~lisitsyn@5-28-21-24.clients.tlt.100megabit.ru] has left #shogun [] | 06:46 | |
-!- votjakovr [~votjakovr@host-46-241-3-209.bbcustomer.zsttk.net] has joined #shogun | 08:01 | |
-!- sonne|work [~sonnenbu@91-64-72-127-dynip.superkabel.de] has joined #shogun | 08:46 | |
-!- nube [~rho@116.90.239.3] has quit [Ping timeout: 260 seconds] | 09:06 | |
-!- iglesiasg [~iglesias@2001:6b0:1:1da0:c55f:fe6c:b69e:1325] has joined #shogun | 09:28 | |
-!- mode/#shogun [+o iglesiasg] by ChanServ | 09:29 | |
-!- shogun-notifier- [~irker@7nn.de] has joined #shogun | 09:33 | |
shogun-notifier- | shogun: Thoralf Klein :develop * fad0b0d / src/shogun/machine/OnlineLinearMachine.h: https://github.com/shogun-toolbox/shogun/commit/fad0b0dd172745fe8504b3e361d78cf024762618 | 09:33 |
shogun-notifier- | shogun: OnlineLinearMachine::get_w(): Copy weight before passing them to SGVector to prevent double-free on destruction. | 09:33 |
shogun-notifier- | shogun: Fernando Iglesias :develop * c6e3416 / src/shogun/machine/OnlineLinearMachine.h: https://github.com/shogun-toolbox/shogun/commit/c6e3416ac3cca42514762c974747a47ec2fad158 | 09:33 |
shogun-notifier- | shogun: Merge pull request #1196 from tklein23/double_free_hotfix | 09:33 |
shogun-notifier- | shogun: | 09:33 |
shogun-notifier- | shogun: Fixed double-free when using "SGVector<float32_t> OnlineLinearMachine::get_w()" | 09:33 |
-!- nube [~rho@116.90.239.13] has joined #shogun | 09:35 | |
-!- nube1 [~rho@116.90.239.3] has joined #shogun | 09:36 | |
-!- nube [~rho@116.90.239.13] has quit [Ping timeout: 276 seconds] | 09:41 | |
shogun-buildbot | build #1165 of bsd1 - libshogun is complete: Failure [failed test_1] Build details are at http://www.shogun-toolbox.org/buildbot/builders/bsd1%20-%20libshogun/builds/1165 blamelist: Thoralf Klein <thoralf.klein@zib.de> | 09:47 |
-!- travis-ci [~travis-ci@ec2-75-101-240-235.compute-1.amazonaws.com] has joined #shogun | 09:54 | |
travis-ci | [travis-ci] it's Fernando Iglesias'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/8612862 | 09:54 |
-!- travis-ci [~travis-ci@ec2-75-101-240-235.compute-1.amazonaws.com] has left #shogun [] | 09:54 | |
-!- zxtx [~zv@cpe-66-68-190-37.austin.res.rr.com] has quit [Ping timeout: 248 seconds] | 09:58 | |
@iglesiasg | wiking, ping | 10:00 |
-!- nube1 [~rho@116.90.239.3] has quit [Ping timeout: 256 seconds] | 10:06 | |
-!- nube [~rho@36.252.26.1] has joined #shogun | 10:14 | |
shogun-buildbot | build #1166 of bsd1 - libshogun is complete: Failure [failed test_1] Build details are at http://www.shogun-toolbox.org/buildbot/builders/bsd1%20-%20libshogun/builds/1166 blamelist: Fernando Iglesias <fernando.iglesiasg@gmail.com> | 10:23 |
shogun-buildbot | build #1289 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/1289 blamelist: Thoralf Klein <thoralf.klein@zib.de> | 10:27 |
@wiking | iglesiasg: pong | 10:32 |
@iglesiasg | wiking, how can we fix the annoying json serializable error in travis and bsd? | 10:36 |
@wiking | iglesiasg: travis is quite easy | 10:36 |
@iglesiasg | wiking, how then? :) | 10:37 |
@wiking | iglesiasg: remove libjson0-dev from before_install | 10:37 |
@wiking | and add a new line there after wget | 10:37 |
@iglesiasg | wiking, an empty line? | 10:38 |
@wiking | iglesi | 10:38 |
@wiking | no | 10:38 |
@wiking | wget --no-check-certificate https://github.com/json-c/json-c/archive/json-c-0.11-20130402.zip && unzip json-c-0.11-20130402.zip | 10:39 |
@wiking | and then add the path to the ./configure | 10:39 |
@wiking | iglesiasg: although you have to change the ./configure file | 10:39 |
@wiking | as there's no way to set json path | 10:39 |
@wiking | or better | 10:39 |
@wiking | wget --no-check-certificate https://github.com/json-c/json-c/archive/json-c-0.11-20130402.zip && unzip json-c-0.11-20130402.zip && cd <json dir> && sh autogen.sh && ./configure && make && sudo make install | 10:40 |
@wiking | something like this | 10:41 |
@wiking | and that way ./configure should be able to detect the json-c lib | 10:41 |
@wiking | mmm no | 10:41 |
@iglesiasg | wiking, this wget should be in the line that starts with script? | 10:42 |
@wiking | you don't need the 'sh autogen.sh' part | 10:42 |
@iglesiasg | mmm | 10:42 |
@wiking | this should be right after the other wget | 10:42 |
@wiking | that handles gmock | 10:42 |
@iglesiasg | at the end it seems it is not 'quite easy' :P | 10:42 |
@wiking | well it's this eays | 10:42 |
@wiking | you just need to add this line | 10:42 |
@wiking | and remove libjson0-dev from the apt-get install line | 10:42 |
@wiking | only thing u need to figure out | 10:43 |
@iglesiasg | add it after before_install? | 10:43 |
@wiking | is what's the dir name | 10:43 |
@wiking | dude | 10:43 |
@wiking | https://github.com/shogun-toolbox/shogun/blob/develop/.travis.yml | 10:43 |
@wiking | https://github.com/shogun-toolbox/shogun/blob/develop/.travis.yml#L42 | 10:43 |
@wiking | do you see that line? | 10:43 |
@iglesiasg | my bad... I was reading the travis.yml in master. I just cloned the repo in the office computer | 10:44 |
@iglesiasg | wiking, yeah, I see that line | 10:44 |
@wiking | - wget --no-check-certificate https://github.com/json-c/json-c/archive/json-c-0.11-20130402.zip && unzip json-c-0.11-20130402.zip && cd json-c-json-c-0.11-20130402 && ./configure && make && sudo make install | 10:44 |
@wiking | this is the line you have to add after line 42 | 10:44 |
@wiking | and before line 43 which starts with before_script | 10:45 |
@iglesiasg | thank you | 10:45 |
@wiking | and remove libjson0-dev from line 41 | 10:45 |
@wiking | this way you'll have json 0.11 | 10:45 |
@wiking | and this should fix the travis build problems | 10:45 |
@iglesiasg | wiking, ok I am going to try right now then, thank you! | 10:46 |
@wiking | iglesiasg: just push it to develop | 10:46 |
@wiking | and check the logs | 10:46 |
@iglesiasg | sure | 10:46 |
@wiking | ok | 10:47 |
@wiking | wait | 10:47 |
@wiking | did u push? | 10:47 |
@iglesiasg | no | 10:47 |
@wiking | ok | 10:47 |
@wiking | dont | 10:47 |
@iglesiasg | it seems there is no make file | 10:47 |
shogun-notifier- | shogun: Viktor Gal :develop * b90b3e3 / .travis.yml: https://github.com/shogun-toolbox/shogun/commit/b90b3e34132352ddddc184851861a4545a5a3931 | 10:48 |
shogun-notifier- | shogun: Fix json version on travis | 10:48 |
@wiking | god bless webedit of github ;) | 10:48 |
@iglesiasg | hehe | 10:48 |
@wiking | iglesiasg: makefile is generated by ./configure | 10:49 |
@iglesiasg | no space left on device | 10:49 |
@iglesiasg | no Makefile generated | 10:49 |
@wiking | where? :) | 10:49 |
@iglesiasg | the laptop I am working on, nothing to do with travis | 10:50 |
@wiking | well what does df -h say? :) | 10:50 |
@iglesiasg | 100% used | 10:50 |
@wiking | heheh then you need to request for a bigger hdd ;) | 10:50 |
@wiking | ok i've made a mistake | 10:50 |
shogun-notifier- | shogun: Viktor Gal :develop * 77e3c59 / .travis.yml: https://github.com/shogun-toolbox/shogun/commit/77e3c595b7fb4d8be48b31f62eb1a1fbcafbbee6 | 10:52 |
shogun-notifier- | shogun: Travis: fix cwd before building | 10:52 |
@wiking | this should do it finally | 10:52 |
@iglesiasg | wiking, https://travis-ci.org/shogun-toolbox/shogun do you see some of the builds there gray? | 10:52 |
@wiking | iglesiasg: yeah | 10:52 |
@wiking | 10:50 <@wiking> ok i've made a mistake | 10:52 |
@wiking | did you read this? :) | 10:52 |
@iglesiasg | yes | 10:52 |
@wiking | 10:52 < shogun-notifier-> shogun: Travis: fix cwd before building | 10:52 |
@wiking | ^ | 10:52 |
@iglesiasg | I am used to see them red when a mistake is done rather than gray | 10:52 |
@wiking | should fix the grays | 10:53 |
@iglesiasg | don't really know what the gray is | 10:53 |
@wiking | error :) | 10:53 |
@iglesiasg | red is error as well | 10:54 |
@wiking | Checking for libjson (json-c) ... yes | 10:54 |
@wiking | ok | 10:54 |
@wiking | this should do it | 10:54 |
-!- travis-ci [~travis-ci@ec2-107-22-45-184.compute-1.amazonaws.com] has joined #shogun | 10:55 | |
travis-ci | [travis-ci] it's Viktor Gal'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/8614446 | 10:55 |
-!- travis-ci [~travis-ci@ec2-107-22-45-184.compute-1.amazonaws.com] has left #shogun [] | 10:55 | |
@iglesiasg | uhm ok | 10:56 |
@wiking | ok let see how build 1185 ends | 10:57 |
sonne|work | wiking: can you install the same thing on the bsd machine? | 11:01 |
@wiking | sonne|work: noup sorry | 11:01 |
@wiking | sonne|work: cannot do the same for bsd machine as i intend to keep packages there and not some hand compiled stuff :( | 11:02 |
@wiking | as it makes things much easier to maintain | 11:02 |
sonne|work | wiking: ok then let us disable python_modular on that machine | 11:02 |
@wiking | sonne|work: by all means | 11:03 |
@wiking | sonne|work: i'll let u know when ports gets libjson updated to 0.11 | 11:04 |
sonne|work | wiking: could you get the ubu1 buildot back online? | 11:07 |
@wiking | sonne|work: that one is out of my hands | 11:07 |
sonne|work | what happened? | 11:07 |
@wiking | sonne|work: it's a box at a uni | 11:07 |
sonne|work | it is down... | 11:07 |
@wiking | sonne|work: and none of us are actually at the uni | 11:07 |
@wiking | and whenever it goes down | 11:07 |
@wiking | we have to ask for favours | 11:07 |
@wiking | to bring it back | 11:07 |
sonne|work | so now what? | 11:07 |
@wiking | so it's not instant | 11:07 |
@wiking | :( | 11:07 |
shogun-buildbot | build #1290 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/1290 blamelist: Fernando Iglesias <fernando.iglesiasg@gmail.com> | 11:08 |
sonne|work | can you ask or is it 'lost'? | 11:08 |
@wiking | sometimes it takes 2 weeks to get it back | 11:08 |
@wiking | i've asked already | 11:08 |
sonne|work | ok | 11:08 |
@wiking | so it'll eventually come back | 11:08 |
@wiking | but i cannot tell exactly when | 11:08 |
sonne|work | hmmhh where is foulwall... | 11:09 |
@wiking | ok i think travis is at least fix | 11:11 |
@wiking | since python_modular is green now | 11:11 |
sonne|work | https://travis-ci.org/shogun-toolbox/shogun | 11:12 |
sonne|work | wiking: all error'd | 11:12 |
@wiking | sonne|work: that's 1184 | 11:12 |
@wiking | https://travis-ci.org/shogun-toolbox/shogun/builds/8614528 | 11:12 |
@wiking | check ^ that one | 11:12 |
@wiking | that's 1185 | 11:12 |
@wiking | 3 are still building the rest is green | 11:12 |
@iglesiasg | it seems to be fixed now, nice! | 11:13 |
sonne|work | so current != current | 11:13 |
@wiking | seems like | 11:13 |
shogun-buildbot | build #1167 of bsd1 - libshogun is complete: Failure [failed test_1] Build details are at http://www.shogun-toolbox.org/buildbot/builders/bsd1%20-%20libshogun/builds/1167 blamelist: Viktor Gal <vigsterkr@gmail.com> | 11:13 |
-!- travis-ci [~travis-ci@ec2-54-226-17-233.compute-1.amazonaws.com] has joined #shogun | 11:18 | |
travis-ci | [travis-ci] it's Viktor Gal'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/8614528 | 11:18 |
-!- travis-ci [~travis-ci@ec2-54-226-17-233.compute-1.amazonaws.com] has left #shogun [] | 11:18 | |
@wiking | voila' ! | 11:19 |
@iglesiasg | it's weird to get that message even if all the builds are right | 11:19 |
@wiking | iglesiasg: it does it because of | 11:20 |
@wiking | on_success: change | 11:20 |
@wiking | from .travis.yml | 11:20 |
@wiking | i.e. when it goes from fail -> success it'll give a notification | 11:20 |
@iglesiasg | aham | 11:20 |
@wiking | but from now on until it's in success state it won't give any notification | 11:21 |
-!- iglesiasg [~iglesias@2001:6b0:1:1da0:c55f:fe6c:b69e:1325] has quit [Remote host closed the connection] | 11:25 | |
shogun-buildbot | build #1168 of bsd1 - libshogun is complete: Failure [failed test_1] Build details are at http://www.shogun-toolbox.org/buildbot/builders/bsd1%20-%20libshogun/builds/1168 blamelist: Viktor Gal <vigsterkr@gmail.com> | 11:26 |
-!- HeikoS [~heiko@nat-191-96.internal.eduroam.ucl.ac.uk] has joined #shogun | 11:27 | |
-!- mode/#shogun [+o HeikoS] by ChanServ | 11:27 | |
sonne|work | wiking: current is still not green but the *85 is | 11:32 |
sonne|work | does anyone have a clue about videorecording? I am thinking about renting gear http://www.dv-kameraverleih.de/ for the workshop | 11:32 |
@wiking | sonne|work: in best case u need just a decent camera | 11:33 |
@wiking | but do you know how's the light setup @ c-base? | 11:34 |
@wiking | i think i can bring with me an hd camera | 11:34 |
@wiking | sonne|work: does c-base have a good inet uplink? in that case we could even stream the conference? | 11:35 |
@wiking | sonne|work: i can offer my server as a proxy | 11:36 |
@wiking | sonne|work: so we basically stream the ogg to my server with icecast | 11:36 |
@wiking | and then from there anybody can watch it | 11:36 |
@wiking | so that the bandwidth is not sooo much of an issue | 11:36 |
@wiking | and from the dv cam i can create an ogg/theora stream and put that into icecast | 11:37 |
sonne|work | wiking: it is rather dark usually but I am totally clueless wrt all that. you would have to take care of the streaming solution. I was thinking of ustream or so | 11:38 |
@wiking | as most of the current cams are only streaming mp4 or similar... | 11:38 |
@wiking | sonne|work: ustream clinet is quite problematic if we have a custom recorder | 11:38 |
@wiking | i.e. not an android/iphone phone | 11:39 |
@wiking | icecast is pretty simple | 11:39 |
-!- nube [~rho@36.252.26.1] has quit [Ping timeout: 248 seconds] | 11:39 | |
@wiking | and then after the live stream using the original stream we can convert them to webm etc and then put out to the site the html5 videos | 11:40 |
@wiking | and that's all | 11:40 |
@wiking | pertty simple thing | 11:40 |
@wiking | done that couple of times | 11:40 |
@wiking | sonne|work: what would be a good thing to do is to mux the beamer and the camera stream into one video | 11:41 |
@wiking | so that the people watching the videos can see what's on the presentation stream | 11:41 |
-!- nube [~rho@36.252.26.1] has joined #shogun | 11:41 | |
@wiking | again that shouldn't be that hard to do | 11:42 |
@wiking | just need a video stream from the beamer | 11:42 |
@wiking | and then render that with the video stream of the camera | 11:42 |
sonne|work | the uplink might be okish. but IMHO it would be best if we have some low quality live stream and then HQ recordings (on youtube) later | 11:42 |
@wiking | sonne|work: we can use our own machine for publishing the videos i think | 11:43 |
@wiking | html5 is supported by all the major browsers | 11:43 |
@wiking | anyhow we can play with icecast of the quality of the stream | 11:43 |
shogun-buildbot | build #1291 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/1291 blamelist: Viktor Gal <vigsterkr@gmail.com> | 11:45 |
-!- nube [~rho@36.252.26.1] has quit [Quit: Leaving.] | 11:48 | |
sonne|work | wiking: get some setup working and we will use it! but please test it before :D | 11:51 |
@wiking | sonne|work: i'm just trying to create a live stream from that camcoder | 11:55 |
@wiking | sonne|work: we would need a good mic for sure | 11:57 |
@wiking | sonne|work: possibly an RF mic that you can hook up to somebody's t-shirt and the receiver can be connected to the camera | 11:57 |
sonne|work | hmmhh no idea where to get that though | 11:59 |
@wiking | sonne|work: at that site that you sent | 12:00 |
@wiking | sonne|work: u know these type of mic they use @ universities | 12:00 |
-!- lambday [67157f4c@gateway/web/freenode/ip.103.21.127.76] has joined #shogun | 12:04 | |
lambday | HeikoS: hi | 12:04 |
@HeikoS | lambday: hi! | 12:04 |
lambday | I implemented the get/set diagonal stuffs | 12:04 |
lambday | RationalApprroximationIndidivualJob is also implemented and unit-tested | 12:04 |
@HeikoS | lambday: nice man | 12:05 |
@HeikoS | so you agreed with the diagonals for matrices? | 12:05 |
lambday | as soon as I implement the LogRationalApproximation implementation, I'll be able to unit-test the shifts and all | 12:05 |
@HeikoS | rather than shifts for operators? | 12:05 |
@HeikoS | nice | 12:05 |
lambday | HeikoS: yes.. although we need to create a SGVector with the diagonal element separately, but that doesn't last long in the memory | 12:06 |
@HeikoS | lambday: no thats fine and not much | 12:06 |
lambday | yes | 12:06 |
lambday | 15MB extra | 12:06 |
@HeikoS | also keep in mind that for sequential we only have one of them | 12:06 |
lambday | that's affordable | 12:06 |
@HeikoS | for parallel, thats not the bottleneck | 12:06 |
@HeikoS | and for distributed, it doesnt matter | 12:06 |
lambday | yes | 12:06 |
lambday | by the way, I just out of curiosity made the RationalApproximation class non-abstract (temporarily) to check whether shifts/weights/const-multiplier are being calculated correctly | 12:08 |
lambday | and it matches the results with krylstat | 12:08 |
-!- van51 [~van51@athedsl-408350.home.otenet.gr] has joined #shogun | 12:08 | |
lambday | just that I need a concrete implementation to unit-test that stuff | 12:08 |
lambday | oh and yes, using arprec for computing sn, cn, dn will surely increase accuracy | 12:09 |
lambday | cn is disturbing using std:: lib | 12:09 |
@HeikoS | lambday: wow thats awesome | 12:10 |
@HeikoS | really getting there! :) | 12:10 |
lambday | yeah! getting excited to see this work! but unfortunately my thesis adviser NEEDS to see me everyday in his lab :'( | 12:11 |
lambday | so, sitting in his lab, chatting with you :D | 12:11 |
lambday | my thesis is also interesting.. they are trying to build online classifier for malware | 12:11 |
@HeikoS | lambday: cool | 12:12 |
@HeikoS | lambday: please be careful about uni stuff :) | 12:12 |
@HeikoS | so what methods are you using? | 12:12 |
lambday | HeikoS: yes trying not to get my ass kicked! I'm really irregular :'( | 12:12 |
@HeikoS | and classifying what exactly? | 12:13 |
lambday | HeikoS: umm, till now they were using logistic regressions | 12:13 |
lambday | umm... its a project from quickheal (antivirus) | 12:13 |
@HeikoS | cool stuff | 12:13 |
@HeikoS | send me a copy when you are done | 12:13 |
lambday | they regularly update their database with new info... | 12:13 |
@HeikoS | I see | 12:13 |
lambday | HeikoS: sure :) this will go on for next 1 year | 12:13 |
lambday | yes, so they need some online classifier sort of thing... now they are moving towards transducer based kernels... | 12:14 |
lambday | sequence kernels :-/ (I am yet to read) | 12:14 |
@HeikoS | I dont know that stuff :) | 12:15 |
@HeikoS | online learning is interesting though | 12:15 |
@HeikoS | maybe some potential for shogun stuff there :) but thats later | 12:16 |
lambday | I so wish that I'd use shogun for experiments | 12:16 |
lambday | but my prof sticks to weka and java | 12:16 |
lambday | :( | 12:16 |
@HeikoS | I am currently working on MCMC for shogun, and we will use your gsoc project once things work | 12:16 |
lambday | HeikoS: awesome!!! :D | 12:16 |
@HeikoS | lambday: thats fine, shogun is sometimes a bit unstable ;) | 12:16 |
shogun-buildbot | build #1292 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/1292 blamelist: Viktor Gal <vigsterkr@gmail.com> | 12:17 |
lambday | btw why SGMatrix::clone doesn't use memcpy but rather copies elementwise? just wondering | 12:18 |
@HeikoS | lambday: there might be reasons for that | 12:18 |
@HeikoS | but I dont see any now :) | 12:18 |
shogun-notifier- | shogun: Roman Votyakov :develop * 7103b5c / src/shogun/machine/gp/ (8 files): https://github.com/shogun-toolbox/shogun/commit/7103b5c9f8650186c6cbe0c8ddbca9f31f0950e1 | 12:21 |
shogun-notifier- | shogun: refactor likelihood models | 12:21 |
shogun-notifier- | shogun: Heiko Strathmann :develop * 667f2f3 / src/shogun/machine/gp/ (8 files): https://github.com/shogun-toolbox/shogun/commit/667f2f32640e8348b3f404d9027725b5e653ddf0 | 12:21 |
shogun-notifier- | shogun: Merge pull request #1200 from votjakovr/feature/gp_refactoring | 12:21 |
shogun-notifier- | shogun: | 12:21 |
shogun-notifier- | shogun: refactor likelihood models | 12:21 |
@HeikoS | votjakovr: thanks for the patch :) | 12:21 |
votjakovr | HeikoS: hey, yep :) | 12:22 |
@HeikoS | votjakovr: hi! :) | 12:22 |
@HeikoS | votjakovr: whats the state of the logit stuff? what is missing for it to work? | 12:22 |
votjakovr | HeikoS: unit tests + that integral evaluation | 12:23 |
votjakovr | HeikoS: btw i'll also implement probit likelihood | 12:23 |
@HeikoS | votjakovr: cool! | 12:25 |
@HeikoS | votjakovr: that should be straight, right? | 12:25 |
votjakovr | HeikoS: yep | 12:26 |
-!- lambday [67157f4c@gateway/web/freenode/ip.103.21.127.76] has quit [] | 12:27 | |
@HeikoS | votjakovr: cool then, I will include an example on the workshop | 12:27 |
votjakovr | HeikoS: btw GPstuff toolbox uses Gauss-Kronrod quadrature formula for integral aprroximation | 12:27 |
@HeikoS | votjakovr: really? I thought they use this mixture stuff | 12:28 |
@HeikoS | votjakovr: I think it doesnt matter too much, maybe choose the one that you can do fastest | 12:28 |
@HeikoS | but I think its good to stay in synch with GPML box | 12:29 |
votjakovr | HeikoS: ok, i'll use mixture | 12:29 |
@HeikoS | votjakovr: is there a comparison of methods? | 12:30 |
@HeikoS | I mean if the Kronrod quadrature is better, and there is code, why not use it | 12:31 |
@HeikoS | or maybe add both (but later, just choose one for now) | 12:32 |
votjakovr | HeikoS: i choose approximation with mixture of cdf, because it's simpler to implement it :) | 12:34 |
@HeikoS | votjakovr: ok then! :) | 12:34 |
votjakovr | HeikoS: which example do you whant to include on the workshop? | 12:35 |
@HeikoS | votjakovr: just a toy binary classification | 12:36 |
@HeikoS | but with learning hyperparameters if possible | 12:37 |
shogun-buildbot | build #1169 of bsd1 - libshogun is complete: Failure [failed test_1] Build details are at http://www.shogun-toolbox.org/buildbot/builders/bsd1%20-%20libshogun/builds/1169 blamelist: Heiko Strathmann <heiko.strathmann@gmail.com> | 12:37 |
votjakovr | HeikoS: Ok :) But will include an example for regression also? | 12:40 |
@HeikoS | votjakovr: yes of course, best thing was to have one big example where one can click together points (like the existing svr/svm ones) and then have buttons to learn hyperparameters, plot predictive distributions, and use different inference methods (sparse, exact) | 12:41 |
@HeikoS | votjakovr: so if you enjoy coding such a thing, feel free! otherwise I will probably do some hacking on the weekend on this | 12:42 |
votjakovr | HeikoS: Ok, i'll try to do it :) | 12:44 |
@HeikoS | votjakovr: nice :) but please focus on the binary stuff first since thats more important for GSoC for now | 12:44 |
votjakovr | HeikoS: yeah, i agree | 12:48 |
shogun-buildbot | build #1170 of bsd1 - libshogun is complete: Failure [failed test_1] Build details are at http://www.shogun-toolbox.org/buildbot/builders/bsd1%20-%20libshogun/builds/1170 blamelist: Roman Votyakov <votjakovr@gmail.com> | 12:48 |
-!- lisitsyn [~lisitsin@mxs.kg.ru] has joined #shogun | 13:02 | |
-!- lambday [67157c4f@gateway/web/freenode/ip.103.21.124.79] has joined #shogun | 13:06 | |
shogun-buildbot | build #1293 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/1293 blamelist: Heiko Strathmann <heiko.strathmann@gmail.com> | 13:18 |
-!- van51 [~van51@athedsl-408350.home.otenet.gr] has quit [Quit: Leaving.] | 13:22 | |
-!- foulwall [~user@2001:da8:215:6100:5063:d7c2:d2e0:7584] has joined #shogun | 13:24 | |
-!- iglesiasg [~nando@s83-179-44-135.cust.tele2.se] has joined #shogun | 13:26 | |
-!- mode/#shogun [+o iglesiasg] by ChanServ | 13:26 | |
sonne|work | foulwall: around? | 13:30 |
foulwall | Hi sonne|work , I would like to turn my working hour to night... sorry for your waiting. This week I have to go for college fieldwork from Monday to Friday, and after that I'll turn my working hour to normal | 13:33 |
sonne|work | foulwall: hmmh, please tell such things before you actually do them. | 13:35 |
sonne|work | foulwall: when do you have time to discuss? The latest commit state is rather b0rken | 13:35 |
foulwall | sonne|work: Maybe 11 utc this week? and after this week we can make it back to 9 utc | 13:36 |
sonne|work | not sure if I make it 11 utc maybe a little later (lunch break...) | 13:37 |
sonne|work | foulwall: currently the clustering example is broken | 13:38 |
sonne|work | when you click clear | 13:38 |
shogun-buildbot | build #1294 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/1294 blamelist: Roman Votyakov <votjakovr@gmail.com> | 13:38 |
sonne|work | foulwall: you still have the old points | 13:38 |
foulwall | sonne|work: ok, I found that bug today, and is the GaussianProcessRegression class updated? now gp-demo seems not accept the argument. | 13:40 |
sonne|work | foulwall: yes you have to fix the example to work with the new interface votjakovr implemented | 13:41 |
foulwall | and I'll rewrite kernel_matrix visualization today. and implement a simple data importer. | 13:41 |
foulwall | sonne|work: was there any docs? I don't know how to fix that? | 13:42 |
sonne|work | foulwall: well you took it from the examples/ dir and that one got updated so just do the same thing | 13:43 |
foulwall | sonne|work: gotcha! | 13:43 |
sonne|work | foulwall: and each example should have 'interactive data', 'load toy data', 'generate data' | 13:44 |
foulwall | sonne|work: ok | 13:45 |
foulwall | hey sonne|work, I found a little bug on shogun's zh-cn manual. http://www.shogun-toolbox.org/doc/cn/current/installation.html | 14:32 |
foulwall | interface section, interfaces=python_modular not python-modular | 14:33 |
foulwall | and python_static not python, etc... this manual looks old. | 14:38 |
sonne|work | foulwall: yes it is. I cannot update it - cannot parse kanjis :D | 14:38 |
shogun-notifier- | shogun: Roman Votyakov :develop * 550de70 / src/shogun/machine/gp/ (3 files): https://github.com/shogun-toolbox/shogun/commit/550de70e7cb7bd017e29267309e58cd6b0f910dc | 14:58 |
shogun-notifier- | shogun: change default value of the derivative wrt hypers | 14:58 |
shogun-notifier- | shogun: Heiko Strathmann :develop * f19478d / src/shogun/machine/gp/ (3 files): https://github.com/shogun-toolbox/shogun/commit/f19478d8e38e5191a72b717cc309fdadcd1f02fa | 14:58 |
shogun-notifier- | shogun: Merge pull request #1201 from votjakovr/feature/gp_refactoring | 14:58 |
shogun-notifier- | shogun: | 14:58 |
shogun-notifier- | shogun: change default value of the derivative wrt hypers | 14:58 |
@HeikoS | votjakovr: what are the cases when the derivatives cannot be computed, what happens then? | 14:58 |
votjakovr | HeikoS: for example we can't comute derivative wrt parameter, which is not used by likelihood model | 15:01 |
@HeikoS | then the method should throw an error | 15:01 |
@HeikoS | not quietly return an empty vector | 15:01 |
@HeikoS | votjakovr: are there other cases (where no error should be thrown?) | 15:02 |
votjakovr | HeikoS: yep, please look at CLaplacianInferenceMethod::get_marginal_likelihood_derivatives | 15:03 |
@HeikoS | votjakovr: I see | 15:04 |
@HeikoS | if (deriv.num_cols*deriv.num_rows > 0) | 15:04 |
@HeikoS | this one you mean right? | 15:04 |
@HeikoS | makes sense | 15:04 |
votjakovr | HeikoS: yep, it's not a perfect solution, but it works | 15:05 |
@HeikoS | votjakovr: yeah, something like try-catch would be nicer, but ok lets leave it as it is | 15:05 |
shogun-buildbot | build #1171 of bsd1 - libshogun is complete: Failure [failed test_1] Build details are at http://www.shogun-toolbox.org/buildbot/builders/bsd1%20-%20libshogun/builds/1171 blamelist: Heiko Strathmann <heiko.strathmann@gmail.com> | 15:11 |
-!- foulwall [~user@2001:da8:215:6100:5063:d7c2:d2e0:7584] has quit [Remote host closed the connection] | 15:16 | |
-!- FSCV [~FSCV@65.19.131.247] has joined #shogun | 15:17 | |
-!- lisitsyn [~lisitsin@mxs.kg.ru] has quit [Quit: Leaving.] | 15:20 | |
shogun-buildbot | build #1172 of bsd1 - libshogun is complete: Failure [failed test_1] Build details are at http://www.shogun-toolbox.org/buildbot/builders/bsd1%20-%20libshogun/builds/1172 blamelist: Roman Votyakov <votjakovr@gmail.com> | 15:23 |
-!- iglesiasg [~nando@s83-179-44-135.cust.tele2.se] has quit [Quit: Leaving] | 15:35 | |
shogun-buildbot | build #1295 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/1295 blamelist: Heiko Strathmann <heiko.strathmann@gmail.com> | 15:40 |
-!- lambday [67157c4f@gateway/web/freenode/ip.103.21.124.79] has quit [Ping timeout: 250 seconds] | 15:45 | |
shogun-notifier- | shogun: foulwall :develop * 27be102 / examples/undocumented/python_modular/graphical/interactive_gp_demo.py: https://github.com/shogun-toolbox/shogun/commit/27be102a5d223497d23ff5acec21b299582b1b7a | 16:13 |
shogun-notifier- | shogun: fixed interactive gp demo, under votjakovr's new GaussianProcessRegression class | 16:13 |
shogun-notifier- | shogun: Soeren Sonnenburg :develop * 38a4003 / examples/undocumented/python_modular/graphical/interactive_gp_demo.py: https://github.com/shogun-toolbox/shogun/commit/38a4003ae4c98a7e0eddeaaccea4d3d3a2973507 | 16:13 |
shogun-notifier- | shogun: Merge pull request #1202 from foulwall/develop | 16:13 |
shogun-notifier- | shogun: | 16:13 |
shogun-notifier- | shogun: fix graphical interactive gp demo, under votjakovr's new GaussianProcessRegression class | 16:13 |
shogun-buildbot | build #1296 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/1296 blamelist: Roman Votyakov <votjakovr@gmail.com> | 16:15 |
shogun-buildbot | build #1173 of bsd1 - libshogun is complete: Failure [failed test_1] Build details are at http://www.shogun-toolbox.org/buildbot/builders/bsd1%20-%20libshogun/builds/1173 blamelist: Soeren Sonnenburg <sonne@debian.org> | 16:26 |
-!- van51 [~van51@athedsl-408350.home.otenet.gr] has joined #shogun | 16:33 | |
shogun-buildbot | build #1174 of bsd1 - libshogun is complete: Failure [failed test_1] Build details are at http://www.shogun-toolbox.org/buildbot/builders/bsd1%20-%20libshogun/builds/1174 blamelist: foulwall <liuzhengyang@outlook.com> | 16:38 |
-!- foulwall [~user@2001:da8:215:c252:6564:69ee:7ca1:6375] has joined #shogun | 16:43 | |
votjakovr | foulwall: hi, thanks for the fixes! | 16:45 |
foulwall | votjakovr: you're welcome | 16:46 |
shogun-buildbot | build #1297 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/1297 blamelist: Soeren Sonnenburg <sonne@debian.org> | 16:54 |
sonne|work | foulwall: now only the demo needs your fix :) | 17:01 |
-!- lisitsyn [~lisitsin@mxs.kg.ru] has joined #shogun | 17:12 | |
foulwall | sonne|work: bugs are already fixed, after data_importer I'll send a full pr. | 17:19 |
foulwall | sonne|work: and I added a menu on gp-demo, make user available to choose their custom kernel function | 17:21 |
@wiking | sonne|work: ok tomorrow hopefully i'll manage to get a firewire 800-to-400 converter and then i'll be able to test the live streaming... in my little town there's no such thing (i've tried 6 different shops :P) | 17:25 |
shogun-buildbot | build #1298 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/1298 blamelist: foulwall <liuzhengyang@outlook.com> | 17:29 |
lisitsyn | oh guys I didn't resist and bought iphone | 17:29 |
-!- pickle27 [~Kevin@d67-193-243-174.home3.cgocable.net] has joined #shogun | 17:33 | |
@HeikoS | lisitsyn: hahahha :) | 17:43 |
-!- lambday [67157e4f@gateway/web/freenode/ip.103.21.126.79] has joined #shogun | 17:44 | |
-!- nube [~rho@49.244.125.166] has joined #shogun | 17:44 | |
@HeikoS | lisitsyn: I hope you like it at least ;) | 17:47 |
@wiking | lisitsyn: why not nexus? | 17:47 |
lisitsyn | HeikoS: enjoy it a lot comparing to philips x513 I had before | 17:48 |
lisitsyn | wiking: don't know if it is even available here | 17:48 |
-!- lisitsyn [~lisitsin@mxs.kg.ru] has quit [Quit: Leaving.] | 17:54 | |
-!- nube [~rho@49.244.125.166] has quit [Ping timeout: 264 seconds] | 18:03 | |
-!- nube [~rho@36.253.62.98] has joined #shogun | 18:10 | |
pickle27 | hey everyone | 18:19 |
pickle27 | today is Canada Day! | 18:19 |
van51 | pickle27: so, what aboot it? :P | 18:34 |
pickle27 | van51: just a fun fact! | 18:34 |
pickle27 | van51: its Canada's Birthday so its a national holiday with fireworks etc | 18:35 |
van51 | pickle27: aah sounds nice! happy birthday then :) | 18:35 |
pickle27 | thanks! | 18:35 |
-!- foulwall [~user@2001:da8:215:c252:6564:69ee:7ca1:6375] has quit [Ping timeout: 264 seconds] | 18:36 | |
-!- nube1 [~rho@49.126.118.140] has joined #shogun | 18:40 | |
-!- nube [~rho@36.253.62.98] has quit [Read error: Connection reset by peer] | 18:40 | |
-!- foulwall [~user@2001:da8:215:c252:1c25:3b2:82f0:6f94] has joined #shogun | 18:41 | |
van51 | nai | 18:42 |
van51 | ops | 18:42 |
@wiking | sonne|work: ping? | 18:45 |
@wiking | HeikoS: here? | 18:48 |
@HeikoS | wiking: hi! | 18:48 |
@wiking | HeikoS: do u have mplayer at hand? | 18:48 |
@HeikoS | wiking: nope, why? | 18:49 |
@wiking | mmm can u play this stream: http://maeth.com:8000/test.ogg | 18:49 |
@HeikoS | wiking: btw whats with this unit test stuff, it feels like ages ago | 18:49 |
@HeikoS | yep can play | 18:49 |
@HeikoS | some books | 18:49 |
@wiking | ok cool | 18:49 |
@wiking | did u hear me? | 18:49 |
@HeikoS | noting moves though | 18:49 |
@wiking | :) | 18:49 |
@HeikoS | have no sound here | 18:50 |
@HeikoS | put your hand in | 18:50 |
@wiking | u seen my fingers? | 18:50 |
@wiking | :) | 18:50 |
@HeikoS | nope | 18:50 |
@HeikoS | image is stuck | 18:50 |
@HeikoS | time also doesnt move | 18:50 |
@wiking | mmm i think there's a lag | 18:50 |
@HeikoS | ah there is a finger | 18:50 |
@wiking | ok cool | 18:50 |
@HeikoS | but framerate is about 1 fps | 18:51 |
@wiking | yeah | 18:51 |
@HeikoS | ehm | 18:51 |
@HeikoS | 0.1 | 18:51 |
@HeikoS | or even less | 18:51 |
@HeikoS | no the finger is gone | 18:51 |
@wiking | the problem is that i'm generating the input with a raspberry pi | 18:52 |
@wiking | so the rendering time is not optimal ;) | 18:53 |
@wiking | anyhow the thing works | 18:53 |
-!- zxtx [~zv@rrcs-74-62-200-195.west.biz.rr.com] has joined #shogun | 18:53 | |
@wiking | HeikoS: u dont have audio on your machine? | 18:53 |
@wiking | or you dont get the audio strema? | 18:53 |
@wiking | *stream | 18:56 |
@wiking | sonney2k: ok the video/audio streaming works. | 18:56 |
@wiking | sonney2k: we just need a good (strong with cpu time) linux machine dedicated for rendering the video... | 18:57 |
-!- van51 [~van51@athedsl-408350.home.otenet.gr] has quit [Quit: Leaving.] | 18:57 | |
@HeikoS | wiking: sorry no not really currently | 19:01 |
@wiking | HeikoS: ok no worries | 19:01 |
@wiking | HeikoS: i see with mplayer that the audio is there as well | 19:01 |
@wiking | and i've heard it as well | 19:01 |
@wiking | the only thing is that we need a dedicated machine for this | 19:01 |
@wiking | and i still need to figure out how to mux it in a way | 19:02 |
@wiking | ah i think i know | 19:02 |
@wiking | tr will do it for me | 19:02 |
lambday | HeikoS: hi | 19:03 |
@HeikoS | lambday: hi! | 19:03 |
lambday | shifts, weights, etc tested.. works fine *but* | 19:03 |
@wiking | mmm what is the mux command in linux? | 19:04 |
lambday | HeikoS: I tried to add a log-det test as well.. something is wrong.. its coming out to be near the actual log-det value.. not even close | 19:04 |
lambday | :( | 19:04 |
lambday | till shifts I tested.. | 19:04 |
@HeikoS | lambday: what did you try exactly? | 19:05 |
lambday | similar like the one before.. estimate log-det using rational approximation this time | 19:05 |
lambday | HeikoS: well, could you please tell me if this is what we are supposed to do.. | 19:06 |
lambday | Q is the matrix | 19:06 |
lambday | so, I move shifted ops like this --> (zI-Q) | 19:06 |
lambday | z=shfit | 19:06 |
@HeikoS | lambday: try the matrix log itself first maybe | 19:06 |
@HeikoS | so multiply with a fixed vector rather than random | 19:07 |
lambday | HeikoS: alright... | 19:07 |
@HeikoS | but that wont solve it I guess | 19:07 |
lambday | yes | 19:07 |
lambday | so, I solve system with (zI-Q)x=s, s=sample | 19:07 |
@HeikoS | so you are solving system individually currently? | 19:07 |
lambday | yes | 19:08 |
@HeikoS | so check the solutions for all these systems | 19:08 |
@HeikoS | and before check whether the shifts are correct ( I think you did that already right?) | 19:08 |
lambday | yes shifts are correct as per krylstat | 19:08 |
@HeikoS | lambday: okay | 19:08 |
-!- lisitsyn [~lisitsyn@83.234.54.206] has joined #shogun | 19:08 | |
@HeikoS | then it could be the systems | 19:08 |
@HeikoS | maybe wrong solution | 19:08 |
lambday | but can't check whether individual solutions are correct | 19:08 |
@HeikoS | so check them | 19:08 |
@HeikoS | why not? | 19:09 |
lambday | krylstat uses cocg_m | 19:09 |
@HeikoS | compare against matlab/octave | 19:09 |
lambday | okay | 19:09 |
@HeikoS | and if those are correct, there is probably something wrong with the formulae | 19:09 |
lambday | I think something is wrong with the formula I am using here.. | 19:10 |
lambday | after I solve the systems, I multiplied weight w*x, then sum it all up over all the shifts | 19:11 |
lambday | then took the imag, applied operator to that imag vector once again, then multiplied with the constant multiplier... | 19:11 |
lambday | I mean, pretty much what krylstat does | 19:11 |
lambday | :- | 19:11 |
lambday | :-/ | 19:11 |
@HeikoS | lambday: dont worry | 19:11 |
@HeikoS | just check step by step | 19:11 |
lambday | alright | 19:12 |
@HeikoS | maybe a little toy example in octave reveals things | 19:12 |
@HeikoS | and if its also wrong in octave, you know its the math | 19:12 |
-!- shogun-notifier- [~irker@7nn.de] has quit [Quit: transmission timeout] | 19:13 | |
lambday | alright, I am checking | 19:13 |
@HeikoS | lambday: also check the individual poarts of the imag computation | 19:15 |
@HeikoS | maybe we missed something there in the associative rules | 19:16 |
-!- shogun-notifier- [~irker@7nn.de] has joined #shogun | 19:20 | |
shogun-notifier- | shogun-demo: foulwall :master * d5fc7b8 / / (11 files): https://github.com/shogun-toolbox/shogun-demo/commit/d5fc7b86e515904ee2898eebee80f2084ade0f8b | 19:20 |
shogun-notifier- | shogun-demo: 1. fix bug in clustering. 2.adapt gp demo to new | 19:20 |
shogun-notifier- | shogun-demo: GaussianProcessRegression class. 3. user now can choose kernel_function | 19:20 |
shogun-notifier- | shogun-demo: in gp demo. 4. rewrite the kernel_matrix visualization demo in new | 19:20 |
shogun-notifier- | shogun-demo: framework, but got a defective result. I'll check if there's any better | 19:20 |
shogun-notifier- | shogun-demo: way than conrec.js for heatmap drawing, or else I'll fix the defect. | 19:20 |
shogun-notifier- | shogun-demo: Soeren Sonnenburg :master * 2a074ed / / (11 files): https://github.com/shogun-toolbox/shogun-demo/commit/2a074ed63d81db03258887b3ec5794ffee52f3fb | 19:20 |
shogun-notifier- | shogun-demo: Merge pull request #11 from foulwall/perceptron | 19:20 |
shogun-notifier- | shogun-demo: | 19:20 |
shogun-notifier- | shogun-demo: 1. fix bug in clustering. 2.adapt gp demo to new | 19:20 |
lambday | HeikoS: okay I am checking.. but as per Hale paper, it should be alright :( | 19:20 |
lambday | HeikoS: I'm checking log(Q) first... may be that will reveal the issue | 19:21 |
@HeikoS | lambday: yep good idea! | 19:21 |
-!- foulwall [~user@2001:da8:215:c252:1c25:3b2:82f0:6f94] has quit [Ping timeout: 264 seconds] | 19:21 | |
-!- foulwall [~user@2001:da8:215:c252:b8f1:eb8a:e017:c0a1] has joined #shogun | 19:32 | |
@sonney2k | foulwall, ok clustering works again! | 19:35 |
@wiking | sonney2k: ok so this camera is 720x576, ie dvd quality | 19:36 |
@sonney2k | foulwall, GPs do mostly (but creating a new train data set should clear) | 19:36 |
@wiking | sonney2k: if we want better quality we need to rent | 19:36 |
@sonney2k | wiking, IMHO good enough | 19:36 |
@wiking | sonney2k: other than that the icecast worked nicely | 19:36 |
@wiking | sonney2k: i'm just investigating how we can render the slides as well into the video stream nicely | 19:36 |
@wiking | sonney2k: i'll prepare the scripts | 19:36 |
@sonney2k | wiking, can we store to the recordings too for putting them on youtube later? | 19:37 |
@wiking | sonney2k: but can u bring a good laptop/pc which can handle the rendering | 19:37 |
@sonney2k | rendering? | 19:37 |
-!- van51 [~van51@athedsl-408350.home.otenet.gr] has joined #shogun | 19:37 | |
@sonney2k | what is needed? I only have a core2duo | 19:37 |
@wiking | sonney2k: hehehe well the streaming is totally command line: dvgrab -format raw -V -input /dev/video0 -| ffmpeg2theora -a 0 -v 5 -f dv -x 320 -y 240 -o /dev/stdout -|oggfwd -p -d "Shogun test stream" -n "Shogun workshop" maeth.com 8000 ... | 19:37 |
@sonney2k | van51, evening gentlement | 19:37 |
foulwall | sonney2k: ok, I'll fix them tomorrow!. too late now, cu | 19:38 |
@wiking | sonney2k: so i can insert there a | tee dvstream.dv | | 19:38 |
@sonney2k | wiking, from usb? then maybe someone with sth faster would work! | 19:38 |
@sonney2k | foulwall, don't forget your weekly report! | 19:38 |
@wiking | sonney2k: i have firewire output as well | 19:38 |
van51 | sonney2k: hello | 19:38 |
@sonney2k | wiking, a fine | 19:38 |
@wiking | sonney2k: just make sure that machine has a firewire input | 19:38 |
@wiking | sonney2k: but since it's a firewire 400 | 19:38 |
@wiking | sonney2k: it's as good as usb 2.0 high speed ;P | 19:38 |
@sonney2k | wiking, I guess we could later do the conversion | 19:38 |
foulwall | ok, I'll write on my cellphone | 19:39 |
@wiking | sonney2k: well yeah we shoudl do this small resolution streaming on the fly | 19:39 |
@wiking | sonney2k: with a core2duo it should be fine | 19:39 |
@wiking | sonney2k: i was testing with a raspberry pi :P | 19:39 |
@wiking | sonney2k: i mean the grabbing + encoding + streaming | 19:39 |
@wiking | sonney2k: and it was ok :P | 19:39 |
@wiking | but obviously we need something more decent | 19:40 |
@wiking | a c2d is going to be just fine | 19:40 |
@wiking | imho we should save the dv raw input | 19:40 |
@wiking | so we can do some postprocessing and encode the video in higher resolution | 19:40 |
@wiking | and put it up later... | 19:41 |
@wiking | sonney2k: only thing we need is a good microphone | 19:41 |
@sonney2k | wiking, don't you have a 'faster' notebook | 19:41 |
@sonney2k | yeah that remains a problem | 19:41 |
@wiking | sonney2k: i do... but it's currently osx | 19:42 |
@wiking | and dvgrab and shit won't work | 19:42 |
@sonney2k | wiking, usb mic? | 19:42 |
@wiking | sonney2k: naaah | 19:42 |
@sonney2k | wiking, what then? | 19:42 |
@wiking | sonney2k: i need something that i can plug in to the camcoder | 19:42 |
@wiking | which is an analog input | 19:42 |
@sonney2k | van51, so what is the status? | 19:42 |
@wiking | you know the usual headphone jack | 19:42 |
@sonney2k | wiking, yeah but then it wouldjust work right? | 19:42 |
@sonney2k | I mean how is this osx? | 19:43 |
van51 | sonney2k: well, I think it's ok now in general | 19:43 |
@wiking | sonney2k: yeah because then the camcoder reads the audio from that input instead of the built in mic | 19:43 |
@wiking | sonney2k: in osx i have the problem with libusb/libfirewire missing .... | 19:43 |
@wiking | sonney2k: and dvgrab is totally dependent on that | 19:43 |
@wiking | sonney2k: and i don't have an equivalent utility on osx like dvgrab | 19:44 |
van51 | sonney2k: I' m just thinking now that it may be redundant to handle a CStringFeatures object inside and that it could be switched to a plain SGStringList | 19:44 |
@wiking | that simply reads the dv stream from usb/firewire | 19:44 |
@wiking | sonney2k: but in worst case if we cannot find a spare machine | 19:44 |
@wiking | what i can do is simply | 19:44 |
@wiking | boot from a usb stick a linux | 19:44 |
@wiking | and that should do it | 19:44 |
@wiking | i'll bring an external storage | 19:45 |
@wiking | to save there the raw dv stream | 19:45 |
@wiking | i'll still have to figure out how to grab the video stream from the beamer | 19:45 |
@wiking | and mux it into the video stream | 19:45 |
@wiking | but i'll solve this in the next couple of days | 19:45 |
@wiking | sonney2k: i hope the machine that is hooked up to the beamer is some linux distro ;P | 19:46 |
@wiking | sonney2k: can we do a pilot test on 11th ? | 19:46 |
@wiking | sonney2k: with the beamer etc....? | 19:46 |
@sonney2k | wiking, in the morning of the 12th only | 19:46 |
@wiking | sonney2k: ok | 19:46 |
@wiking | sonney2k: then i'll come early and that's it | 19:46 |
@wiking | shoudln't take more than 20 mins max | 19:46 |
@wiking | sonney2k: can u take care of the mic? | 19:47 |
@sonney2k | wiking, but you could probably ask the c-base guys on the 11th if you just go there for a drink with sergey lisitsyn | 19:47 |
@wiking | sonney2k: rent it from somewhere in berlin? | 19:47 |
@wiking | i mean the microphone | 19:47 |
@sonney2k | wiking, no idea I will try but no promises :/ | 19:47 |
@wiking | sonney2k: or at least can u give me some urls of multimedia rent places in berlin | 19:48 |
@wiking | and i'll pintpoint u the mic we would need | 19:48 |
@wiking | sonney2k: if we can setup the live stream nicely | 19:48 |
@sonney2k | van51, I only ``know'' the one I posted | 19:48 |
@sonney2k | wiking, ^ | 19:48 |
@sonney2k | van51, it is ok with using StringFeatures | 19:49 |
@wiking | sonney2k: we should be advertise that people who are watching the stream can ask questions here on the irc channel | 19:49 |
@sonney2k | wiking, during discussion | 19:49 |
@sonney2k | makes sense :D | 19:49 |
@wiking | sonney2k: i:P | 19:49 |
van51 | sonney2k: It's just a bit messy when you initialize it, but that's all | 19:49 |
van51 | sonney2k: I beliebe it's good to be merged then :) | 19:49 |
van51 | believe* | 19:49 |
@sonney2k | van51, ok then I will have another look | 19:50 |
@wiking | sonney2k: what was the link? | 19:50 |
@wiking | ok found it | 19:51 |
@wiking | sonney2k: c-base should have some basic a/v equipment no? | 19:51 |
@sonney2k | yeah they have everything. which doesn't mean they lend everything to us | 19:52 |
@wiking | sonney2k: can u ask them if they have an rf mic? | 19:52 |
@wiking | sonney2k: that's the only thing we would need | 19:52 |
@wiking | and inet uplink :P | 19:53 |
@wiking | sonney2k: http://www.amazon.com/Pyle-Pro-PDWM96-Lavalier-Wireless-Microphone/dp/B0007L8BQW/ref=sr_1_37?ie=UTF8&qid=1372701248&sr=8-37&keywords=microphone | 19:54 |
@wiking | sonney2k: something similar | 19:54 |
@wiking | sonney2k: as u can see there's one part with the jack that would go for the camcoder and the other part is the mic and the rf transmitter | 19:54 |
@wiking | nothing fancy | 19:54 |
@wiking | sonney2k: i'll even bring a tripod so they dont even have to lend us that :P | 19:55 |
@sonney2k | http://www.amazon.de/Wireless-Funkmikrofon-Mikrofon-Tranmitter-Empf%C3%A4nger/dp/B00ALNI436/ref=pd_sim_sbs_ce_4 | 19:57 |
@wiking | yeah that's the stuff | 19:57 |
@sonney2k | wiking, alright then let me buy it | 20:02 |
@sonney2k | van51, could you do some benchmarks then? | 20:04 |
@sonney2k | van51, do you have a decent machine? | 20:04 |
van51 | sonney2k: yeah I think so | 20:04 |
@sonney2k | van51, a long time ago I organized this challenge http://largescale.ml.tu-berlin.de/submission/evaluation/1/2/ | 20:05 |
@sonney2k | so why not take this data set | 20:05 |
@sonney2k | van51, data is here ftp://largescale.ml.tu-berlin.de/largescale/webspam/ | 20:06 |
@sonney2k | you need train data and labels | 20:06 |
@sonney2k | van51, ftp://largescale.ml.tu-berlin.de/largescale/README | 20:07 |
@sonney2k | that is the description | 20:07 |
@sonney2k | van51, take say a subset of a few thousand examples first and try to predict on some held out set of say 50k examples | 20:08 |
@sonney2k | van51, ask if you need more details :D | 20:08 |
@sonney2k | wiking, ok ordered | 20:08 |
van51 | sonney2k: ok! | 20:09 |
@sonney2k | van51, I still see some issues | 20:10 |
van51 | sonney2k: ok, tell me | 20:11 |
@sonney2k | van51, one is that dim is not used - maybe we drop dim altogether and just use num_bits | 20:11 |
@sonney2k | then get_nnz_features_for_vector should return the string length not dim | 20:11 |
van51 | sonney2k: dim is used in the comparison of vector lengths | 20:11 |
van51 | and I think it's more intuitive to specify the dimension than the number of bits | 20:11 |
@sonney2k | van51, yeah but not when reducing the hash | 20:11 |
@sonney2k | it should be hash % dim then | 20:12 |
@sonney2k | but it is a speed penalty | 20:12 |
@sonney2k | not much but it is | 20:12 |
van51 | sonney2k: also for the number of nnz features isn't dim the smallest upper bound? | 20:13 |
@sonney2k | van51, you should always assume that string lenght << dim | 20:13 |
@wiking | sonney2k: lol ok cool | 20:13 |
@wiking | sonney2k: the server has 100mbit/sec uplink from where we'd stream | 20:13 |
@wiking | sonney2k: so it can handle quite some connections | 20:14 |
@wiking | sonney2k: ok then the only thing is that lets have enough spare batteries for that device.... | 20:14 |
van51 | sonney2k: why is that? a document can have an arbitrary number of tokens, but you hash it in a specific dimension size, namely dim | 20:14 |
-!- HeikoS [~heiko@nat-191-96.internal.eduroam.ucl.ac.uk] has quit [Quit: Leaving.] | 20:15 | |
@sonney2k | van51, yes but it is usually sparse! | 20:16 |
van51 | sonney2k: ok, I get what you 're saying | 20:17 |
@sonney2k | van51, so if you string is say 10k long and you dim is 256**8 (for 8-grams) you would get still get about only 10k as nnz features == upper bound of number of n-grams in the string | 20:17 |
@sonney2k | van51, but IMHO you can just drop dim | 20:18 |
@sonney2k | it is totally sufficient to do this bit based | 20:18 |
van51 | sonney2k: ok, but I would like to leave in in the constructor | 20:18 |
van51 | sonney2k: so that it can be specified in this way as well | 20:18 |
@sonney2k | van51, but then add a check that it is a dim=2**n | 20:25 |
van51 | sonney2k: I just explain it in the documentation after all | 20:27 |
-!- nube [~rho@36.252.243.176] has joined #shogun | 20:27 | |
van51 | sonney2k: I'll test it and repush | 20:28 |
-!- nube1 [~rho@49.126.118.140] has quit [Ping timeout: 276 seconds] | 20:29 | |
@sonney2k | van51, hmmhh maybe nbits is better and less misleading than a parameter that can be used wrongly | 20:30 |
@sonney2k | van51, let me check again | 20:30 |
van51 | sonney2k: yeah I meant I changed it to num_bits and explain in the documentation what it means for the dimensionality of the feature space | 20:30 |
@sonney2k | van51, well you could do return hash % dim; | 20:31 |
@sonney2k | I don't know about nowadays processors maybe it even is as fast as hash & mask; | 20:31 |
@sonney2k | nowadays | 20:31 |
van51 | sonney2k: I have no idea :p | 20:32 |
@sonney2k | van51, well benchmark it :D | 20:32 |
van51 | sonney2k: hehe true! | 20:33 |
@sonney2k | just use the webspam data set for that | 20:33 |
van51 | sonney2k: ok. then when I update it again you can merge and if the benchmark suggests something else, we update it? | 20:34 |
@wiking | sonney2k: http://dvswitch.alioth.debian.org/wiki/ | 20:34 |
@sonney2k | van51, yes | 20:35 |
@sonney2k | wiking, excellent | 20:36 |
@sonney2k | wiking, problem is only that we might have people showing demos | 20:36 |
@sonney2k | so it is not clear if they run on each and every machine | 20:36 |
@sonney2k | wiking, it looks perfect though | 20:37 |
@wiking | sonney2k: yeah that's why is this shit there: TwinPact100 vga capture | 20:37 |
@wiking | ;) | 20:37 |
-!- van511 [~van51@athedsl-408350.home.otenet.gr] has joined #shogun | 20:37 | |
@sonney2k | wiking, wait you have one that does vga capture too? | 20:37 |
-!- van51 [~van51@athedsl-408350.home.otenet.gr] has quit [Quit: Leaving.] | 20:37 | |
@wiking | sonney2k: noup :( | 20:38 |
@sonney2k | wiking, too bad... that would have been excellent | 20:38 |
@wiking | sonney2k: i'll try to find it... maybe one of my friends has something similar | 20:38 |
@wiking | doh it's pretty expensive shit | 20:38 |
-!- nube [~rho@36.252.243.176] has quit [Ping timeout: 256 seconds] | 20:39 | |
@sonney2k | yeah | 20:40 |
@sonney2k | no idea what the c-base beamer has | 20:40 |
@wiking | sonney2k: i'll try to look around during this week | 20:42 |
@wiking | sonney2k: since this way of course it doesn't matter who and what beams to the beamer :P | 20:42 |
@sonney2k | van511, ok so please drop the num_bits stuff for now and do instead of return hash & ((1 << num_bits) - 1) return hash % dim; then we will see | 20:44 |
van511 | sonney2k: baah I just finished doing it the other way :p | 20:44 |
@sonney2k | van511, well also fine! | 20:44 |
@sonney2k | van511, when you do benchmarking jsut make sure you don't use some const dim=128 when you do hash % dim later | 20:45 |
@sonney2k | van511, the compiler would do a hash & (mask) thing then | 20:45 |
van511 | sonney2k: you mean that a const in a mod operation is transformed to a mask? | 20:46 |
@sonney2k | van511, alright looking forward to the benchmark - please do it systematically as in use some script to run it on a few strings up to 'lots' and then store timings | 20:46 |
@sonney2k | van511, exactly - compilers are clever these days | 20:47 |
@sonney2k | van511, for a few bit sizes like 8,16,24,32 bit | 20:48 |
@sonney2k | van511, and remember to compile with optimizations enabled! | 20:48 |
van511 | sonney2k: ok, got it! | 20:48 |
@sonney2k | van511, we could then even use these tables in a blog post / create a figure out of it comparing % vs mask | 20:49 |
@wiking | sonney2k: can u ask the exact type of the beamer they have @ c-base plz? | 20:49 |
@sonney2k | van511, and later you can just train liblinear with a) converted features b) with dotfeatures on that data set | 20:49 |
@sonney2k | and see how it goes | 20:49 |
@wiking | sonney2k: maybe i can just bring a simple vga2usb adapter | 20:49 |
@sonney2k | wiking, I have no idea | 20:49 |
van511 | sonney2k: the benchmark will be with the dotfeatures though | 20:50 |
@wiking | sonney2k: well let's just ask the type of beamer they have and i'll try to figure out something | 20:50 |
van511 | sonney2k: that last you said is just for comparison between the two | 20:50 |
-!- iglesiasg [~iglesias@s83-179-44-135.cust.tele2.se] has joined #shogun | 20:54 | |
-!- mode/#shogun [+o iglesiasg] by ChanServ | 20:54 | |
@wiking | sonney2k: or we get another camera that is zoomed and directed on the projectors screen ;P | 20:56 |
-!- nube [~rho@49.244.79.68] has joined #shogun | 21:00 | |
lisitsyn | votjakovr: weekly report please ;) | 21:01 |
lisitsyn | foulwall: weekly report please | 21:01 |
lisitsyn | hushell: weekly report please | 21:01 |
@wiking | lisitsyn: get us a cheap russian solution for vga/dvi to usb :) | 21:01 |
lisitsyn | wiking: vga to usb? | 21:01 |
@wiking | lisitsyn: yep | 21:02 |
@sonney2k | wiking, yeah I asked. I asked a bit too much maybe hope we get a response :) | 21:02 |
@wiking | sonney2k: cool | 21:02 |
lisitsyn | wiking: is that possible? | 21:02 |
@wiking | lisitsyn: yeas for 300 USD ;) | 21:02 |
@wiking | lisitsyn: http://www.amazon.com/Epiphan-Systems-VGA2USB-External-Capture/dp/B000UO94DU/ref=pd_cp_pc_2 | 21:03 |
@wiking | it's basically a video4linux device | 21:03 |
@wiking | it's just a dv stream... | 21:04 |
@sonney2k | van511, well sure makes sense. I suspect you would win the competition if you do this with shogun then ;) | 21:05 |
-!- van511 is now known as van51 | 21:05 | |
@sonney2k | lisitsyn, anything from gsomix? | 21:06 |
van51 | sonney2k: that would be cool :D | 21:06 |
lisitsyn | sonney2k: he is out of touch for me :) | 21:07 |
van51 | sonney2k: btw I have pushed and I am waiting for travis | 21:07 |
@sonney2k | van51, I am serious here yahoo & chap are the ones that won :) | 21:08 |
van51 | sonney2k: yea I saw the results | 21:08 |
van51 | sonney2k: wasn't chap at yahoo at the time? :P | 21:08 |
@sonney2k | van51, and we have liblinear sgdqn etc in shogun but the feature representation helps a *lot* wrt performance | 21:08 |
van51 | what's wrt? | 21:09 |
@sonney2k | and that was just sparse doubles IIRC | 21:09 |
@sonney2k | van51, with respect oo | 21:10 |
* sonney2k thunderstorm! | 21:10 | |
@sonney2k | van51, and dont' forget your weekly report! | 21:11 |
van51 | sonney2k: it's sent! | 21:11 |
@sonney2k | van51, one more thing with our PR - you always have to use the *SAME* seed | 21:13 |
van51 | sonney2k: I got that from Olivier's paper | 21:13 |
@sonney2k | van51, otherwise the order of n-grams will matter | 21:13 |
van51 | sonney2k: they were saying that for each feature you use a different hash function | 21:14 |
@sonney2k | hash = calculate_token_hash(&sv.vector[start], end-start, num_bits, hash); | 21:14 |
van51 | sonney2k: and that you can achieve that by having the seed set to the previous hash value | 21:14 |
@sonney2k | yeah but it depends on the previous feature here | 21:14 |
@sonney2k | hmmhh | 21:15 |
-!- nube [~rho@49.244.79.68] has quit [Ping timeout: 248 seconds] | 21:15 | |
@sonney2k | when you have 'aaa' 'bbb' it might make a big difference if you compare it to 'bbb' 'aaa' then | 21:16 |
van51 | sonney2k: no | 21:16 |
van51 | sonney2k: sorry wait | 21:16 |
van51 | sonney2k: I;m reading it again | 21:16 |
@sonney2k | it would make sense in my eyes when you have say certain feature classes - say several texts describing one image or so | 21:17 |
@sonney2k | then each class could have its own seed | 21:17 |
van51 | sonney2k: you are right | 21:17 |
van51 | sonney2k: it just says to use a different seed for each feature | 21:18 |
van51 | like the index of the feature | 21:18 |
@sonney2k | van51, it can make sense btw | 21:18 |
@sonney2k | van51, on the DNA data set it would probably | 21:19 |
van51 | sonney2k: yeah but this is for general purposes | 21:19 |
@sonney2k | van51, there is a tradeoff sometimes order of strings matters a lot sometimes not. for now just use some constant like 0xdeadbeaf | 21:20 |
@sonney2k | as seed | 21:21 |
van51 | sonney2k: ok! | 21:21 |
van51 | sonney2k: good catch btw | 21:21 |
van51 | sonney2k: I will also change it in the converter now then | 21:25 |
van51 | so that the tests won't fail | 21:26 |
-!- nube [~rho@49.244.27.183] has joined #shogun | 21:27 | |
* sonney2k Zzzzz | 21:31 | |
lisitsyn | okay sonney2k falls asleep I wake up | 21:32 |
-!- nube [~rho@49.244.27.183] has quit [Ping timeout: 276 seconds] | 21:34 | |
-!- lisitsyn [~lisitsyn@83.234.54.206] has quit [Quit: Leaving.] | 21:36 | |
-!- nube [~rho@49.244.102.148] has joined #shogun | 21:47 | |
-!- lisitsyn [~lisitsyn@83.234.54.206] has joined #shogun | 21:54 | |
-!- lisitsyn [~lisitsyn@83.234.54.206] has quit [Quit: Leaving.] | 21:55 | |
-!- lisitsyn [~lisitsyn@83.234.54.206] has joined #shogun | 21:57 | |
-!- nube [~rho@49.244.102.148] has quit [Ping timeout: 264 seconds] | 21:58 | |
pickle27 | lisitsyn: hey! | 22:06 |
lisitsyn | pickle27: hey | 22:07 |
pickle27 | I was thinking a bit about the shogun interface for my GSoC project | 22:07 |
pickle27 | and I was looking over some older emails with Andreas | 22:07 |
pickle27 | what I'm thinking now is that the AJD algorithms should be preproccessors | 22:08 |
pickle27 | and then the BSS algorithms (which use the AJD algs) should be converters | 22:08 |
lisitsyn | pickle27: ahh so like init / transform? | 22:08 |
lisitsyn | nice | 22:09 |
pickle27 | sort of I am still looking through the pre processors | 22:09 |
pickle27 | ADJ is kind of like the best PCA for several matrices | 22:09 |
pickle27 | so the result would be applicable to new matrices that are expected to be part of the same group | 22:10 |
pickle27 | however unlike PCA AJD takes a SGNDArray instead of a single Matrix | 22:10 |
pickle27 | so we could make an object very similar to PCA | 22:11 |
pickle27 | anyways I messaged Andreas back to see if he could elaborate on the use case for that | 22:11 |
lisitsyn | what is n-dim input here? | 22:12 |
pickle27 | a series of matrics | 22:12 |
pickle27 | could also be just an array/list of matrices | 22:12 |
-!- nube [~rho@49.244.124.177] has joined #shogun | 22:13 | |
pickle27 | I need to move out of the matrix realm and into the feature realm soon. Hopefully Andreas will have an example use case | 22:15 |
lisitsyn | pickle27: could you please give me some understanding of these series? | 22:15 |
pickle27 | although internally it will look similar to the pca object | 22:15 |
lisitsyn | what is one matrix here? | 22:15 |
pickle27 | lisitsyn: I'll try but I only really understand what they are at the numerical level | 22:15 |
pickle27 | Its basically a set of matrices and the goal is to find the single matrix V that best diagonalizes all the matrices in the set | 22:16 |
pickle27 | where PCA diagonlizes only a single matrix AJD diagonalizes a set | 22:16 |
lisitsyn | pickle27: where these matrices come from? | 22:17 |
pickle27 | lisitsyn: right now they are just synthetic data | 22:17 |
pickle27 | I was hoping Andreas could provide a machine learning example | 22:17 |
lisitsyn | pickle27: I mean we all get used to feature vectors | 22:18 |
lisitsyn | but it is like feature matrix? | 22:18 |
lisitsyn | see what I mean? | 22:19 |
pickle27 | yeah | 22:19 |
pickle27 | when you're using AJD for BSS its a series of covariance matrices | 22:19 |
lisitsyn | oh! | 22:20 |
pickle27 | Im not sure what it would be in the pre processing application | 22:20 |
lisitsyn | interesting | 22:20 |
-!- shogun-notifier- [~irker@7nn.de] has quit [Quit: transmission timeout] | 22:20 | |
lisitsyn | pickle27: what is the data these cov. matrices are computed from? | 22:20 |
pickle27 | the mixed input signal you are trying to separate | 22:20 |
pickle27 | mixed input signals* | 22:21 |
pickle27 | lisitsyn: I'm looking at the PCA preprocessor now and it works on a cov matrix too | 22:24 |
lisitsyn | pickle27: yes indeed | 22:24 |
pickle27 | lisitsyn: I think that is probably what you would apply AJD too as well | 22:24 |
lisitsyn | but *one* covariance matrix | 22:24 |
pickle27 | yeah | 22:24 |
lisitsyn | where multiple come from? | 22:25 |
pickle27 | yeah I've got to look into that | 22:25 |
pickle27 | Andreas did mention using AJD for pre processing though | 22:25 |
pickle27 | so he has some idea | 22:25 |
-!- iglesiasg [~iglesias@s83-179-44-135.cust.tele2.se] has quit [Quit: Leaving] | 22:28 | |
lisitsyn | pickle27: alright lets see what he answers then | 22:29 |
lisitsyn | I don't get the concept yet | 22:29 |
pickle27 | yeah Im not quite seeing the connection to applications other than BSS | 22:30 |
-!- nube [~rho@49.244.124.177] has quit [Ping timeout: 256 seconds] | 22:30 | |
-!- lisitsyn [~lisitsyn@83.234.54.206] has quit [Quit: Leaving.] | 22:42 | |
-!- lisitsyn1 [~lisitsyn@83.234.54.206] has joined #shogun | 22:42 | |
-!- lambday [67157e4f@gateway/web/freenode/ip.103.21.126.79] has quit [] | 22:42 | |
-!- nube [~rho@49.244.75.183] has joined #shogun | 22:45 | |
-!- zxtx [~zv@rrcs-74-62-200-195.west.biz.rr.com] has quit [Ping timeout: 256 seconds] | 22:52 | |
-!- pickle27 [~Kevin@d67-193-243-174.home3.cgocable.net] has quit [Quit: Leaving] | 22:54 | |
-!- zxtx [~zv@rrcs-74-62-200-195.west.biz.rr.com] has joined #shogun | 22:55 | |
-!- zxtx [~zv@rrcs-74-62-200-195.west.biz.rr.com] has quit [Ping timeout: 248 seconds] | 23:00 | |
-!- lisitsyn1 [~lisitsyn@83.234.54.206] has left #shogun [] | 23:00 | |
-!- votjakovr [~votjakovr@host-46-241-3-209.bbcustomer.zsttk.net] has left #shogun ["Fallen asleep!"] | 23:02 | |
-!- gsomix [~gsomix@109.188.126.165] has joined #shogun | 23:11 | |
gsomix | hello | 23:11 |
-!- zxtx [~zv@rrcs-74-62-200-195.west.biz.rr.com] has joined #shogun | 23:31 | |
-!- gsomix [~gsomix@109.188.126.165] has quit [Ping timeout: 256 seconds] | 23:51 | |
-!- gsomix [~gsomix@109.188.125.167] has joined #shogun | 23:55 | |
--- Log closed Tue Jul 02 00:00:06 2013 |
Generated by irclog2html.py 2.10.0 by Marius Gedminas - find it at mg.pov.lt!