--- Log opened Wed Jun 25 00:00:47 2014 | ||
-!- shogun-notifier- [~irker@7nn.de] has quit [Quit: transmission timeout] | 00:01 | |
-!- pickle27 [~pickle27@216.191.231.218] has quit [Quit: This computer has gone to sleep] | 00:01 | |
iglesiasg | good night, people | 00:05 |
---|---|---|
-!- iglesiasg [524ae0a7@gateway/web/freenode/ip.82.74.224.167] has quit [Quit: Page closed] | 00:05 | |
-!- HeikoS [~heiko@0543f17e.skybroadband.com] has quit [Quit: Leaving.] | 00:56 | |
-!- vivekrai [uid10305@gateway/web/irccloud.com/x-rierccogfnxhdkpd] has quit [Quit: Connection closed for inactivity] | 01:00 | |
-!- khalednasr [~k.nasr92@41.69.175.62] has quit [Quit: Leaving] | 01:24 | |
-!- soumyaC [uid15286@gateway/web/irccloud.com/x-klcafoygsdzrnomv] has joined #shogun | 01:29 | |
-!- PirosB3 [~pirosb3@host76-224-dynamic.52-79-r.retail.telecomitalia.it] has joined #shogun | 02:47 | |
shogun-buildbot | build #731 of nightly_all is complete: Failure [failed compile test] Build details are at http://buildbot.shogun-toolbox.org/builders/nightly_all/builds/731 blamelist: Viktor Gal <viktor.gal@maeth.com>, Fernando Iglesias <fernando.iglesiasg@gmail.com>, Parijat Mazumdar <mazumdarparijat@gmail.com> | 03:04 |
-!- zxtx_ [~zv@129-79-241-185.dhcp-bl.indiana.edu] has quit [Ping timeout: 240 seconds] | 03:04 | |
-!- PirosB3 [~pirosb3@host76-224-dynamic.52-79-r.retail.telecomitalia.it] has quit [Quit: PirosB3] | 03:08 | |
-!- zxtx_ [~zv@c-98-223-196-32.hsd1.in.comcast.net] has joined #shogun | 04:16 | |
-!- Saurabh7 [~Saurabh7@59.88.36.131] has joined #shogun | 04:44 | |
-!- soumyaC [uid15286@gateway/web/irccloud.com/x-klcafoygsdzrnomv] has quit [Quit: Connection closed for inactivity] | 04:46 | |
-!- Saurabh7 [~Saurabh7@59.88.36.131] has quit [Ping timeout: 272 seconds] | 04:49 | |
-!- Saurabh7 [~Saurabh7@117.200.183.17] has joined #shogun | 05:01 | |
-!- Saurabh7 [~Saurabh7@117.200.183.17] has quit [Read error: Connection reset by peer] | 05:19 | |
-!- Saurabh7 [~Saurabh7@117.203.116.51] has joined #shogun | 05:34 | |
-!- hushell [~hushell@c-71-56-149-179.hsd1.wa.comcast.net] has quit [Ping timeout: 240 seconds] | 06:09 | |
shogun-buildbot | build #835 of nightly_default is complete: Failure [failed test notebooks] Build details are at http://buildbot.shogun-toolbox.org/builders/nightly_default/builds/835 blamelist: Viktor Gal <viktor.gal@maeth.com>, Fernando Iglesias <fernando.iglesiasg@gmail.com>, Parijat Mazumdar <mazumdarparijat@gmail.com> | 06:11 |
-!- Saurabh7 [~Saurabh7@117.203.116.51] has quit [Quit: Leaving] | 06:26 | |
-!- kislay [~Abhijeet@103.240.206.163] has joined #shogun | 07:23 | |
-!- lambday [67157d4c@gateway/web/freenode/ip.103.21.125.76] has joined #shogun | 08:02 | |
-!- vivekrai [uid10305@gateway/web/irccloud.com/x-cgqgzdzrgbemfwzq] has joined #shogun | 08:11 | |
-!- hushell [~hushell@c-71-56-149-179.hsd1.wa.comcast.net] has joined #shogun | 08:48 | |
-!- lambday [67157d4c@gateway/web/freenode/ip.103.21.125.76] has quit [Ping timeout: 246 seconds] | 09:26 | |
-!- hushell [~hushell@c-71-56-149-179.hsd1.wa.comcast.net] has quit [Ping timeout: 240 seconds] | 09:43 | |
-!- hushell [~hushell@c-71-56-149-179.hsd1.wa.comcast.net] has joined #shogun | 09:55 | |
-!- HeikoS1 [~heiko@0543f17e.skybroadband.com] has joined #shogun | 10:01 | |
-!- soumyaC [uid15286@gateway/web/irccloud.com/x-frxgkxwkqifybuww] has joined #shogun | 10:09 | |
-!- hushell [~hushell@c-71-56-149-179.hsd1.wa.comcast.net] has quit [Ping timeout: 244 seconds] | 10:11 | |
-!- lisitsyn [~lisitsin@mxs.kg.ru] has joined #shogun | 10:12 | |
-!- hushell [~hushell@c-71-56-149-179.hsd1.wa.comcast.net] has joined #shogun | 10:17 | |
-!- HeikoS1 [~heiko@0543f17e.skybroadband.com] has quit [Ping timeout: 240 seconds] | 10:24 | |
-!- PirosB3 [~pirosb3@host76-224-dynamic.52-79-r.retail.telecomitalia.it] has joined #shogun | 10:43 | |
-!- hushell [~hushell@c-71-56-149-179.hsd1.wa.comcast.net] has quit [Ping timeout: 255 seconds] | 10:54 | |
-!- HeikoS1 [~heiko@0543f17e.skybroadband.com] has joined #shogun | 10:56 | |
-!- vivekrai [uid10305@gateway/web/irccloud.com/x-cgqgzdzrgbemfwzq] has quit [Quit: Connection closed for inactivity] | 11:00 | |
-!- hushell [~hushell@c-71-56-149-179.hsd1.wa.comcast.net] has joined #shogun | 11:17 | |
-!- HeikoS1 [~heiko@0543f17e.skybroadband.com] has quit [Ping timeout: 264 seconds] | 11:40 | |
-!- PirosB3 [~pirosb3@host76-224-dynamic.52-79-r.retail.telecomitalia.it] has quit [Quit: PirosB3] | 12:14 | |
-!- vivekrai [uid10305@gateway/web/irccloud.com/x-udvjjbgmsbzjzznj] has joined #shogun | 12:37 | |
-!- hushell [~hushell@c-71-56-149-179.hsd1.wa.comcast.net] has quit [Ping timeout: 245 seconds] | 12:57 | |
-!- hushell [~hushell@c-71-56-149-179.hsd1.wa.comcast.net] has joined #shogun | 13:21 | |
-!- thoralf [~thoralf@91-65-216-252-dynip.superkabel.de] has quit [Quit: Konversation terminated!] | 14:12 | |
-!- HeikoS1 [~heiko@nat-187-60.internal.eduroam.ucl.ac.uk] has joined #shogun | 14:53 | |
-!- hushell [~hushell@c-71-56-149-179.hsd1.wa.comcast.net] has quit [Ping timeout: 245 seconds] | 15:23 | |
-!- hushell [~hushell@c-71-56-149-179.hsd1.wa.comcast.net] has joined #shogun | 15:30 | |
-!- pickle27 [~pickle27@192-0-136-118.cpe.teksavvy.com] has joined #shogun | 15:52 | |
-!- hushell [~hushell@c-71-56-149-179.hsd1.wa.comcast.net] has quit [Ping timeout: 244 seconds] | 15:58 | |
-!- pickle27 [~pickle27@192-0-136-118.cpe.teksavvy.com] has quit [Remote host closed the connection] | 16:04 | |
-!- pickle27 [~pickle27@216.191.231.218] has joined #shogun | 16:13 | |
-!- lambday [67157d4c@gateway/web/freenode/ip.103.21.125.76] has joined #shogun | 16:16 | |
-!- yorkerlin [18d4da3a@gateway/web/freenode/ip.24.212.218.58] has joined #shogun | 16:41 | |
-!- hushell [~hushell@c-71-56-149-179.hsd1.wa.comcast.net] has joined #shogun | 16:43 | |
-!- vivekrai [uid10305@gateway/web/irccloud.com/x-udvjjbgmsbzjzznj] has quit [Quit: Connection closed for inactivity] | 17:00 | |
-!- lambday [67157d4c@gateway/web/freenode/ip.103.21.125.76] has quit [Ping timeout: 246 seconds] | 17:32 | |
-!- PirosB3 [~pirosb3@host76-224-dynamic.52-79-r.retail.telecomitalia.it] has joined #shogun | 17:42 | |
-!- HeikoS1 [~heiko@nat-187-60.internal.eduroam.ucl.ac.uk] has quit [Quit: Leaving.] | 18:03 | |
-!- pickle27 [~pickle27@216.191.231.218] has quit [Quit: This computer has gone to sleep] | 18:18 | |
-!- pickle27 [~pickle27@216.191.231.218] has joined #shogun | 18:23 | |
-!- hushell [~hushell@c-71-56-149-179.hsd1.wa.comcast.net] has quit [Ping timeout: 240 seconds] | 18:25 | |
-!- lisitsyn [~lisitsin@mxs.kg.ru] has quit [Quit: Leaving.] | 18:52 | |
-!- thoralf [~thoralf@91-65-141-60-dynip.superkabel.de] has joined #shogun | 19:43 | |
thoralf | Hello. | 19:43 |
-!- thoralf [~thoralf@91-65-141-60-dynip.superkabel.de] has quit [Quit: Konversation terminated!] | 19:55 | |
-!- HeikoS1 [~heiko@nat-161-71.internal.eduroam.ucl.ac.uk] has joined #shogun | 20:13 | |
-!- yorkerlin [18d4da3a@gateway/web/freenode/ip.24.212.218.58] has quit [Ping timeout: 246 seconds] | 20:21 | |
kislay | hey HeikoS1 | 20:27 |
HeikoS1 | kislay: hey there! | 20:27 |
kislay | HeikoS1, can you look over #2325. It's done most probably | 20:27 |
HeikoS1 | kislay: yep | 20:28 |
HeikoS1 | let me check | 20:28 |
HeikoS1 | kislay: commented quite a few things | 20:38 |
kislay | HeikoS1, yup | 20:38 |
HeikoS1 | kislay: most important is the matrix inversion, that should be cholesky, and never compute and store inverse | 20:39 |
HeikoS1 | the rest is cosmetics | 20:39 |
HeikoS1 | (sorry for making you clean up the code of others, but thats the we we improve :) ) | 20:39 |
kislay | HeikoS1, it will help me for my future work. so its good | 20:40 |
HeikoS1 | kislay: thats good | 20:40 |
HeikoS1 | kislay: see david barbers book | 20:43 |
HeikoS1 | 16.2 Fisher's Linear Discriminant | 20:43 |
HeikoS1 | kislay: the method you implement is described there, and also when it fails. you should put the solution (regularisation) also in | 20:44 |
HeikoS1 | kislay: and document properly | 20:44 |
HeikoS1 | kislay: maybe you can even put in example 16.1 into the notebook? :) | 20:45 |
HeikoS1 | kislay: sorry for asking for more and more :) | 20:45 |
HeikoS1 | CanonVar.m: Canonical Variates | 20:45 |
HeikoS1 | demoCanonVarDigits.m: Demo for Canonical Variates | 20:45 |
HeikoS1 | check his implementation also | 20:45 |
HeikoS1 | kislay: very worth understanding and very worth to have in Shogun | 20:48 |
kislay | HeikoS1, coooool | 20:52 |
kislay | HeikoS1, I needed some docs! | 20:52 |
HeikoS1 | kislay: this book is a good source for all those basic probabilistic methods | 20:52 |
kislay | HeikoS1, all the time following OpenCV was getting tougher! | 20:52 |
HeikoS1 | hehe | 20:53 |
HeikoS1 | kislay: let me know if you need any help or so | 20:53 |
kislay | HeikoS1, sure! | 20:53 |
kislay | hey HeikoS1 ... Is it good to have a preprocessor of LDA? | 20:53 |
HeikoS1 | kislay: yes totally | 20:53 |
kislay | or should I revert back and try to add this in MCLDA | 20:54 |
HeikoS1 | kislay: whats the story with that one? | 20:54 |
HeikoS1 | why revert back? | 20:54 |
kislay | no. maybe its repetative. | 20:54 |
HeikoS1 | kislay: please give me some context, I have no idea what you are talking about :) | 20:54 |
kislay | actually we have a Multiclass LDA implemented already. But it is way too different than Fisher LDA | 20:55 |
kislay | so I came up with this. It's okay? | 20:55 |
HeikoS1 | kislay: came up with? | 20:56 |
HeikoS1 | kislay: we should have a std LDA and a multiclass LDA | 20:56 |
HeikoS1 | but if they share lots of things, they should be dealt with in one class | 20:57 |
-!- lisitsyn [~qdrgsm@80.252.20.67] has joined #shogun | 20:57 | |
kislay | then this is standard LDA, I would think! | 20:58 |
HeikoS1 | kislay: what do you mean? | 20:59 |
kislay | I came across this implementation a lot . | 20:59 |
HeikoS1 | kislay: could you be a bit more precise, I dont get what you want to say | 21:01 |
kislay | ah! nothing actually. I just got confused above. | 21:03 |
kislay | will complete it. | 21:03 |
HeikoS1 | kislay: so whats the plan on those classes then? | 21:11 |
kislay | HeikoS1, I will complete the PR with the docs you suggested | 21:12 |
kislay | specially the canonival variates that you suggested. | 21:14 |
HeikoS1 | kislay: cool | 21:15 |
HeikoS1 | kislay: but fix the other things first and make unit tests pass /write some, ebfore doing the new thing, thats just a minor modiufication, we want to approach the step-wise | 21:15 |
kislay | HeikoS1, thanks man. This will be more solid now! | 21:15 |
kislay | HeikoS1, yup. sure | 21:15 |
HeikoS1 | kislay: commenting similar things on the other PR :) | 21:18 |
-!- PirosB3 [~pirosb3@host76-224-dynamic.52-79-r.retail.telecomitalia.it] has quit [Quit: PirosB3] | 21:28 | |
-!- PirosB3 [~pirosb3@host76-224-dynamic.52-79-r.retail.telecomitalia.it] has joined #shogun | 21:31 | |
-!- shogun-notifier- [~irker@7nn.de] has joined #shogun | 21:37 | |
shogun-notifier- | shogun: lambday :develop * 69340bc / data: https://github.com/shogun-toolbox/shogun/commit/69340bc951d08375b0d413a3f7644064d4c309c8 | 21:37 |
shogun-notifier- | shogun: Reverting changes in data to fix broken build on develop | 21:37 |
shogun-notifier- | shogun: Heiko Strathmann :develop * 4c9c918 / data: https://github.com/shogun-toolbox/shogun/commit/4c9c918b9a991eda716b9f42c0e629906a58adc2 | 21:37 |
shogun-notifier- | shogun: Merge pull request #2330 from lambday/develop | 21:37 |
shogun-notifier- | shogun: | 21:37 |
shogun-notifier- | shogun: Reverting changes in data to fix broken build on develop | 21:37 |
shogun-notifier- | shogun: Heiko Strathmann :develop * 625b0ca / src/shogun/features/SubsetStack.cpp: https://github.com/shogun-toolbox/shogun/commit/625b0ca93f0001c9379f94af81ba5a5f6fc76f7e | 21:41 |
shogun-notifier- | shogun: Update SubsetStack.cpp | 21:41 |
shogun-notifier- | shogun: | 21:41 |
shogun-notifier- | shogun: remove warning, update license | 21:41 |
-!- kislay [~Abhijeet@103.240.206.163] has quit [Quit: Leaving] | 21:43 | |
shogun-buildbot | build #2399 of bsd1 - libshogun is complete: Failure [failed test] Build details are at http://buildbot.shogun-toolbox.org/builders/bsd1%20-%20libshogun/builds/2399 blamelist: Heiko Strathmann <heiko.strathmann@gmail.com>, lambday <heavensdevil6909@gmail.com> | 21:45 |
shogun-buildbot | build #372 of deb4 - python3 is complete: Success [build successful] Build details are at http://buildbot.shogun-toolbox.org/builders/deb4%20-%20python3/builds/372 | 21:50 |
shogun-notifier- | shogun: Saurabh :develop * 0bf54f3 / doc/ipython-notebooks/regression/Regression.ipynb: https://github.com/shogun-toolbox/shogun/commit/0bf54f302f55050a7889cec52345891f72e728e8 | 21:50 |
shogun-notifier- | shogun: Regression nb | 21:50 |
shogun-notifier- | shogun: Saurabh :develop * a7372bc / doc/ipython-notebooks/regression/Regression.ipynb: https://github.com/shogun-toolbox/shogun/commit/a7372bc850b074a65ac7a6f3141d87bdcd9817fe | 21:50 |
shogun-notifier- | shogun: lars section | 21:50 |
shogun-notifier- | shogun: Saurabh :develop * e3ee8e2 / doc/ipython-notebooks/regression/Regression.ipynb: https://github.com/shogun-toolbox/shogun/commit/e3ee8e238291a0b5f44d9508a0f670dc0aa54590 | 21:50 |
shogun-notifier- | shogun: refine lars and intro | 21:50 |
shogun-notifier- | shogun: Saurabh :develop * b7cf615 / doc/ipython-notebooks/regression/Regression.ipynb: https://github.com/shogun-toolbox/shogun/commit/b7cf615c5de6e33ffb017863c33fcb941824bac0 | 21:50 |
shogun-notifier- | shogun: add krr, examples and math | 21:50 |
shogun-notifier- | shogun: Heiko Strathmann :develop * 7c16bf8 / doc/ipython-notebooks/regression/Regression.ipynb: https://github.com/shogun-toolbox/shogun/commit/7c16bf8a4d901e7d5a118ff01823ade7655a6542 | 21:50 |
shogun-notifier- | shogun: Merge pull request #2268 from Saurabh7/regressnb | 21:50 |
shogun-notifier- | shogun: | 21:50 |
shogun-notifier- | shogun: Regression nb | 21:50 |
-!- yorkerlin [18d4da3a@gateway/web/freenode/ip.24.212.218.58] has joined #shogun | 21:54 | |
HeikoS1 | yorkerlin: hi! | 21:55 |
yorkerlin | hi | 21:55 |
HeikoS1 | yorkerlin: how are things? | 21:56 |
yorkerlin | the fourth KL method is passed unit tests | 21:56 |
HeikoS1 | yorkerlin: super nice! | 21:56 |
HeikoS1 | yorkerlin: did you see my comments on the PR? | 21:57 |
yorkerlin | just saw | 21:57 |
yorkerlin | the method has a default value NULL | 21:58 |
yorkerlin | if the function value is NULL, it will be activated | 21:58 |
HeikoS1 | yorkerlin: could you just write some tests that ensure this? | 21:58 |
yorkerlin | ok | 21:58 |
yorkerlin | in fact, the first three KL method is used the orignial l-bfgs. I will add some simple example as unit test. | 22:02 |
HeikoS1 | yorkerlin: you dont need to test complicated examples, you can use pathalogical ones, just to make sure that your new option works as desired | 22:14 |
yorkerlin | I will use this example min {x_1^2+x_2^2} without constraint the opt value is 0. if the constraint is all x_i >=1, the opt value is 2 | 22:15 |
HeikoS1 | yep thats good! | 22:15 |
yorkerlin | do you think this example is enough? | 22:15 |
HeikoS1 | yeah I was just really more after checking the few new lines you added. obvious bugs | 22:16 |
yorkerlin | where? | 22:16 |
yorkerlin | please name them. I will correct them. | 22:17 |
HeikoS1 | yorkerlin: nono, I did not mean I found bugs, I just want to make sure there are no obvious ones | 22:17 |
yorkerlin | ok. | 22:17 |
HeikoS1 | yorkerlin: you know, like they happen when one adds an option to code :) | 22:17 |
yorkerlin | yes. I know | 22:18 |
shogun-buildbot | build #2387 of deb3 - modular_interfaces is complete: Success [build successful] Build details are at http://buildbot.shogun-toolbox.org/builders/deb3%20-%20modular_interfaces/builds/2387 | 22:22 |
HeikoS1 | shogun-buildbot: build nightly_all | 22:23 |
HeikoS1 | shogun-buildbot: force build nightly_all | 22:24 |
HeikoS1 | shogun-buildbot: commands | 22:24 |
shogun-buildbot | buildbot commands: commands, dance, destroy, force, hello, help, last, list, mute, notify, source, status, stop, unmute, version, watch | 22:24 |
HeikoS1 | shogun-buildbot: force build nightly_all | 22:25 |
shogun-buildbot | The build has been queued, I'll give a shout when it starts | 22:25 |
HeikoS1 | force build nightly_default | 22:25 |
HeikoS1 | force build osx2_python | 22:25 |
HeikoS1 | shogun-buildbot: force build nightly_default | 22:25 |
shogun-buildbot | The build has been queued, I'll give a shout when it starts | 22:25 |
HeikoS1 | shogun-buildbot: force build osx2_python | 22:26 |
shogun-buildbot | no such builder 'osx2_python' | 22:26 |
HeikoS1 | shogun-buildbot: force build osx2 - python | 22:26 |
shogun-buildbot | no such builder 'osx2' | 22:26 |
HeikoS1 | shogun-buildbot: force build osx2-python | 22:26 |
shogun-buildbot | no such builder 'osx2-python' | 22:26 |
HeikoS1 | shogun-buildbot: force build "osx2 - python" | 22:26 |
shogun-buildbot | no such builder 'osx2 - python' | 22:26 |
shogun-buildbot | build #2400 of bsd1 - libshogun is complete: Failure [failed test] Build details are at http://buildbot.shogun-toolbox.org/builders/bsd1%20-%20libshogun/builds/2400 blamelist: Heiko Strathmann <heiko.strathmann@gmail.com>, Saurabh <saurabh.mahindre@gmail.com> | 22:26 |
HeikoS1 | shogun-buildbot: force build rpm1 | 22:26 |
shogun-buildbot | no such builder 'rpm1' | 22:26 |
HeikoS1 | yorkerlin: whats the state of the notebook btw? any drafts already? | 22:29 |
-!- khalednasr [~k.nasr92@41.69.128.186] has joined #shogun | 22:30 | |
HeikoS1 | yorkerlin: you mentioned that you are interested in doing multiclass Laplace Approx? | 22:33 |
yorkerlin | yes | 22:33 |
HeikoS1 | yorkerlin: I think that woul dbe nice | 22:34 |
HeikoS1 | yorkerlin: in fact, we could do a bunch of multiclass things? | 22:34 |
yorkerlin | My plan is to implement Laplace and 2 Kl methods | 22:34 |
HeikoS1 | yorkerlin: like some EP or variational stuff | 22:34 |
HeikoS1 | yorkerlin: then we finally have multiclass GP :) | 22:34 |
yorkerlin | for multiclass | 22:34 |
yorkerlin | for GSoC period | 22:34 |
HeikoS1 | which KL? | 22:35 |
yorkerlin | the fast dual | 22:35 |
yorkerlin | the cholesky | 22:35 |
yorkerlin | possible the approx diagonal | 22:35 |
HeikoS1 | do these work seamlessly for multiclass likelihoods? | 22:35 |
yorkerlin | fast dual is yes | 22:36 |
yorkerlin | I think laplace is the basis for multiclass | 22:36 |
yorkerlin | that is why I implement this first | 22:36 |
HeikoS1 | yorkerlin: yeah its a good baseline (and doesnt work well ;) ) | 22:37 |
HeikoS1 | yorkerlin: btw we could also use some sparse approximations to GP classification, if you are interested in doing something after gsoc :) | 22:38 |
yorkerlin | yes. I know and I am interested. I will finish the basis first | 22:38 |
HeikoS1 | yorkerlin: for regression we already have the FITC, which is quite nice | 22:39 |
yorkerlin | laplace has naive implementation but I go beyond that | 22:39 |
HeikoS1 | but there are also variational extensions to learn the inducing points | 22:39 |
HeikoS1 | yorkerlin: what do you mean naive implementation? | 22:39 |
yorkerlin | multiclass depends on class decoding | 22:39 |
HeikoS1 | yorkerlin: what do you mean by that? | 22:40 |
yorkerlin | if we use 0/1 vector for multiclass | 22:41 |
HeikoS1 | yorkerlin: ah I see | 22:41 |
HeikoS1 | yeah, its basically one GP per class | 22:41 |
yorkerlin | yeah | 22:41 |
HeikoS1 | and then model their covariance jointly | 22:41 |
HeikoS1 | what do you want to do? | 22:41 |
yorkerlin | no | 22:41 |
yorkerlin | I think the GPML has a pusedo code for efficient implementation | 22:42 |
yorkerlin | the GPML text book | 22:42 |
HeikoS1 | yorkerlin: of multiclass laplace? | 22:42 |
HeikoS1 | yorkerlin: yeah | 22:42 |
yorkerlin | yes | 22:42 |
HeikoS1 | yorkerlin: but dont base your code on that | 22:42 |
HeikoS1 | yorkerlin: many caveats | 22:42 |
HeikoS1 | yorkerlin: rather follow an existing implementation, like GPStuff | 22:42 |
HeikoS1 | of pyGP | 22:42 |
HeikoS1 | or | 22:43 |
yorkerlin | yes. I will follow some existing implementation. | 22:43 |
yorkerlin | but the basic idea should be the same. | 22:43 |
HeikoS1 | yorkerlin: yeah sure, its all about these nasty little numerical linear algebra tricks ;) | 22:44 |
yorkerlin | in fact, I can derive all KL methods by hand. | 22:46 |
HeikoS1 | yorkerlin: I would talk to emt before doing that. It is not clear whether things work nicely for multiclass. There is quite a bit of literature on this | 22:46 |
yorkerlin | This is good for me to extend some unimpelemted idea in shogun | 22:46 |
HeikoS1 | yorkerlin: so we should go for something that is known to work well | 22:47 |
yorkerlin | ok | 22:47 |
HeikoS1 | yorkerlin: see what I mean? | 22:47 |
yorkerlin | yes | 22:47 |
yorkerlin | so, for now I juse continue to do notebook first. right? | 22:47 |
HeikoS1 | yorkerlin: yeah, notebook should have main focus now | 22:48 |
HeikoS1 | yorkerlin: we want a really really cool one | 22:48 |
HeikoS1 | yorkerlin: with nice examples, etc | 22:48 |
yorkerlin | i see | 22:48 |
HeikoS1 | reproduce paper results | 22:48 |
HeikoS1 | etc | 22:48 |
yorkerlin | agree | 22:49 |
yorkerlin | and some bugs may appear and fix for some big real dataset(s) | 22:49 |
HeikoS1 | yorkerlin: exactly | 22:50 |
HeikoS1 | yorkerlin: do you also want to write some benchmarks? | 22:50 |
HeikoS1 | yorkerlin: we have a folder for that, might be good | 22:51 |
HeikoS1 | yorkerlin: ah and then also the linalg interface. We can solve all factorizations and inversions on a GPU if we manage to get this going until then | 22:51 |
yorkerlin | benchmarks for speed or ? | 22:51 |
-!- thoralf [~thoralf@91-65-141-60-dynip.superkabel.de] has joined #shogun | 22:52 | |
thoralf | Hey. | 22:52 |
HeikoS1 | thoralf: hi! | 22:53 |
yorkerlin | ok. | 22:53 |
thoralf | Hey HeikoS1 | 22:53 |
yorkerlin | hi | 22:53 |
-!- travis-ci [~travis-ci@ec2-54-87-245-34.compute-1.amazonaws.com] has joined #shogun | 23:06 | |
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/28439656 | 23:06 |
-!- travis-ci [~travis-ci@ec2-54-87-245-34.compute-1.amazonaws.com] has left #shogun [] | 23:06 | |
-!- PirosB3 [~pirosb3@host76-224-dynamic.52-79-r.retail.telecomitalia.it] has quit [Quit: PirosB3] | 23:29 | |
-!- travis-ci [~travis-ci@ec2-54-237-167-213.compute-1.amazonaws.com] has joined #shogun | 23:31 | |
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/28440057 | 23:31 |
-!- travis-ci [~travis-ci@ec2-54-237-167-213.compute-1.amazonaws.com] has left #shogun [] | 23:31 | |
shogun-buildbot | build #151 of osx2 - python is complete: Failure [failed test python modular] Build details are at http://buildbot.shogun-toolbox.org/builders/osx2%20-%20python/builds/151 blamelist: lambday <heavensdevil6909@gmail.com> | 23:37 |
--- Log closed Thu Jun 26 00:00:48 2014 |
Generated by irclog2html.py 2.10.0 by Marius Gedminas - find it at mg.pov.lt!