| --- Log opened Tue May 21 00:00:11 2013 | ||
| -!- zxtx [~zv@ool-ae2cb78a.dyn.optonline.net] has joined #shogun | 00:16 | |
| -!- zxtx [~zv@ool-ae2cb78a.dyn.optonline.net] has quit [Ping timeout: 256 seconds] | 00:24 | |
| -!- van51 [~van51@athedsl-320452.home.otenet.gr] has left #shogun ["QUIT :Leaving."] | 00:29 | |
| hushell | sonney2k, are you still here? | 01:38 |
|---|---|---|
| hushell | sonney2k, how do I update the documentation? It's my bad forgetting the change of comment. | 01:39 |
| -!- zxtx [~zv@ool-457e751d.dyn.optonline.net] has joined #shogun | 01:50 | |
| -!- nube [~rho@49.244.76.69] has joined #shogun | 03:04 | |
| -!- nube [~rho@49.244.76.69] has quit [Ping timeout: 248 seconds] | 03:34 | |
| -!- nube [~rho@36.252.251.34] has joined #shogun | 03:36 | |
| -!- pickle27 [~kevin@rcv3-lab-pc.ee.queensu.ca] has quit [Quit: Leaving] | 03:40 | |
| -!- nube [~rho@36.252.251.34] has quit [Ping timeout: 256 seconds] | 03:44 | |
| -!- nube [~rho@49.244.67.77] has joined #shogun | 03:54 | |
| -!- nube [~rho@49.244.67.77] has quit [Ping timeout: 252 seconds] | 04:12 | |
| -!- foulwall [~foulwall@2001:da8:215:503:4172:782d:f853:e6d6] has joined #shogun | 04:56 | |
| -!- nube [~rho@116.90.239.3] has joined #shogun | 05:44 | |
| -!- nube [~rho@116.90.239.3] has quit [Quit: Leaving.] | 06:18 | |
| -!- nube [~rho@116.90.239.3] has joined #shogun | 06:19 | |
| -!- lambday [3ba2174c@gateway/web/cgi-irc/kiwiirc.com/ip.59.162.23.76] has joined #shogun | 06:55 | |
| -!- nube [~rho@116.90.239.3] has quit [Ping timeout: 264 seconds] | 07:43 | |
| -!- nube [~rho@116.90.239.3] has joined #shogun | 07:47 | |
| -!- hushell [~hushell@8-92.ptpg.oregonstate.edu] has quit [Ping timeout: 248 seconds] | 08:12 | |
| -!- hushell [~hushell@c-67-189-100-116.hsd1.or.comcast.net] has joined #shogun | 08:29 | |
| -!- shogun-notifier- [~irker@7nn.de] has joined #shogun | 08:35 | |
| shogun-notifier- | shogun: Soeren Sonnenburg :develop * 6acfcf4 / tests/unit/mathematics/Random_unittest.cc: https://github.com/shogun-toolbox/shogun/commit/6acfcf44b9a3750530111202c11c3c8e9ec4e60f | 08:35 |
| shogun-notifier- | shogun: add unit tests for some CMath::random functions | 08:35 |
| @sonney2k | hushell, the documentation is just the doxygen stuff in every .h file! | 08:35 |
| @sonney2k | gtg | 08:35 |
| @sonney2k | brb | 08:35 |
| -!- lisitsyn [~lisitsin@mxs.kg.ru] has joined #shogun | 08:50 | |
| -!- mode/#shogun [+o lisitsyn] by ChanServ | 08:50 | |
| -!- iglesiasg [c1934d18@gateway/web/freenode/ip.193.147.77.24] has joined #shogun | 09:10 | |
| -!- iglesiasg [c1934d18@gateway/web/freenode/ip.193.147.77.24] has quit [Ping timeout: 250 seconds] | 09:14 | |
| -!- sonne|work [~sonnenbu@sams-office-nat.tomtomgroup.com] has joined #shogun | 09:15 | |
| -!- iglesias [c1934d18@gateway/web/freenode/ip.193.147.77.24] has joined #shogun | 09:16 | |
| -!- iglesias is now known as iglesiasg | 09:17 | |
| -!- mode/#shogun [+o iglesiasg] by ChanServ | 09:17 | |
| -!- nube [~rho@116.90.239.3] has quit [Read error: Connection reset by peer] | 09:18 | |
| -!- nube [~rho@116.90.239.3] has joined #shogun | 09:18 | |
| -!- iglesiasg [c1934d18@gateway/web/freenode/ip.193.147.77.24] has quit [Ping timeout: 250 seconds] | 09:25 | |
| foulwall | morrning sonne|work | 09:37 |
| sonne|work | foulwall: good morning! | 09:37 |
| foulwall | sonne|work: I've no idea of which data I use in the demo. | 09:39 |
| foulwall | '/toy/label_train_regression.dat' | 09:40 |
| sonne|work | foulwall: what I have in mind is that the user gets a listing with available data sets (e.g files in data/toy/*) | 09:40 |
| sonne|work | and then can select the one | 09:41 |
| sonne|work | foulwall: yes we have to standardize the format / filename otherwise it is a mess | 09:41 |
| foulwall | sonne|work: all right. | 09:43 |
| sonne|work | foulwall: it is kind of difficult to make a decision here since there are tons of data formats | 09:50 |
| foulwall | sonne|work: I'll standardize them at first. | 09:51 |
| sonne|work | foulwall: I *think* the best way forward would be only support toy/*.h5 for now | 09:53 |
| sonne|work | and then use h5py to read in the data | 09:53 |
| sonne|work | foulwall: this is how you can read it http://mldata.org/about/hdf5/example/ | 09:54 |
| @lisitsyn | sonne|work: eine wuenderschoenen guten morgen | 09:55 |
| @lisitsyn | :D | 09:55 |
| foulwall | sonne|work: ok | 09:55 |
| sonne|work | lisitsyn: yeah weather cooled down | 09:56 |
| @lisitsyn | sonne|work: was ist das wetter da? | 09:57 |
| foulwall | sonne|work: I've got a bad headache, progress might not be so fast today... | 09:57 |
| @lisitsyn | we have cooled to +15 or +20 here | 09:57 |
| sonne|work | lisitsyn: yeah 14 C here too | 10:10 |
| sonne|work | foulwall: then get some sleep! | 10:10 |
| foulwall | sonne|work: :) thanks | 10:11 |
| foulwall | sonne|work: I slept too much and get the headache. | 10:12 |
| lambday | hi sonne|work:, lisitsyn:... good morning :) | 10:12 |
| sonne|work | foulwall: I wish I could say that | 10:12 |
| sonne|work | lambday: morning! | 10:12 |
| lambday | sonne|work: if we use an external library (arprec in this case), what changes shall I make so that shogun configures for it? | 10:14 |
| -!- foulwall [~foulwall@2001:da8:215:503:4172:782d:f853:e6d6] has left #shogun [] | 10:15 | |
| -!- foulwall [~foulwall@2001:da8:215:503:4172:782d:f853:e6d6] has joined #shogun | 10:15 | |
| sonne|work | lambday: for now I would just say you do a #ifdef HAVE_ARPREC ... #endif in your code and manually change the .config file to contain the appropriate compiler/link flags | 10:16 |
| @lisitsyn | lambday: what is arprec? | 10:16 |
| sonne|work | lambday: if using arprec turns out to be useful we add some small code snippet to shogun | 10:16 |
| sonne|work | lisitsyn: lmgify | 10:16 |
| @lisitsyn | sonne|work: what? | 10:16 |
| lambday | sonne|work: manually change the configure file - okay I'm checking | 10:17 |
| @lisitsyn | I can add | 10:17 |
| @lisitsyn | that thing | 10:17 |
| @lisitsyn | lambday: | 10:17 |
| sonne|work | lisitsyn: http://lmgtfy.com/?q=arprec | 10:17 |
| lambday | lisitsyn: arprec is for arbitrary precision | 10:17 |
| lambday | for higher precision | 10:17 |
| @lisitsyn | sonne|work: I can do that ;) | 10:17 |
| lambday | lisitsyn: thanks.. | 10:18 |
| @lisitsyn | lambday: but tonight | 10:18 |
| @lisitsyn | lambday: just do the thing that evil sonne|work suggests | 10:18 |
| lambday | actually very high precision is needed for this - 1E-19, and std math functions show differences than the arprec version | 10:18 |
| sonne|work | lambday: could you perhaps work in log-space? | 10:19 |
| sonne|work | I did so when I had thousands of products of small probabilities | 10:19 |
| lambday | sonne|work: nice idea! let me check | 10:19 |
| sonne|work | lambday: the other trick is to scale the data (e.g. carray some multiplier with it) | 10:20 |
| sonne|work | carry | 10:20 |
| lambday | sonne|work: hmm... actually we need to use math functions - sin, cos, sinh, tanh etc.. would the carry thing work in that? | 10:22 |
| lambday | lisitsyn: lol@evil sonne|work: :D | 10:22 |
| sonne|work | lambday: lisitsyn needs his morning vodka to get his brain started *eg* | 10:23 |
| lambday | lol... :D | 10:23 |
| lambday | lisitsyn: I'll also have some :D | 10:23 |
| @lisitsyn | I see no package for arprec in debian | 10:24 |
| sonne|work | lambday: you have to think it through... it might make things more complex / might not work for all needed operations | 10:24 |
| lambday | sonne|work: okay... let me check... may be arprec thing would be easy | 10:24 |
| @lisitsyn | lambday: can we stick to GNU MP? | 10:24 |
| sonne|work | lambday: if it is not in debian - then almost no one will have this installed | 10:25 |
| @lisitsyn | it is maintained | 10:25 |
| sonne|work | so a different solution would be much perferred | 10:25 |
| lambday | sonne|work: ummm... yes you're right | 10:25 |
| lambday | lisitsyn: I haven't worked in mp, I'll check today then... sounds better | 10:26 |
| @lisitsyn | sonne|work: what about you, have you already got your beer and sausages? :D | 10:26 |
| sonne|work | lisitsyn: no but I wear my lederhosen as usal | 10:28 |
| sonne|work | usual | 10:28 |
| lambday | its 32 C here and we're burning :( | 10:30 |
| @lisitsyn | lambday: just go to votjakovr's place you are going to like it | 10:31 |
| lambday | he's also from russia right? | 10:31 |
| @lisitsyn | lambday: yes barnaul, north-east from you | 10:32 |
| lambday | lisitsyn: lol I'll drop by his place in the eve then, have some beer and vodka :D | 10:35 |
| sonne|work | lambday: but watch the polar bears! | 10:37 |
| @lisitsyn | sonne|work: they are *polar* | 10:37 |
| @lisitsyn | not like phi rho | 10:37 |
| @lisitsyn | I mean they are much northern :D | 10:37 |
| lambday | whoa! | 10:38 |
| lambday | awesome! | 10:38 |
| lambday | lisitsyn: does gmp support complex numbers? | 10:38 |
| @lisitsyn | lambday: I don't know I just guessed something more maintained :) | 10:38 |
| lambday | lisitsyn: okay I'm checking | 10:38 |
| sonne|work | lambday: and long double is not enough? | 10:41 |
| sonne|work | ... floatmax_t in shogun | 10:41 |
| sonne|work | can be anything from 96 bits to 128bits | 10:41 |
| lambday | sonne|work: oh.. I haven't checked with that... I used float64_t... let me try | 10:42 |
| lambday | nope - this is what I get using arprec (m=0.00186744273288133783884890082305219038971699774265289306640625), and this is using long double (m=0.00186744273288133848937020431435485079418867826461791992187500) | 10:46 |
| sonne|work | lambday: but seriously best would be you could work in logspace or so - that doeasn't seem right... | 10:47 |
| @lisitsyn | hahah precise guy | 10:47 |
| lambday | sonne|work: hmmm... yes.. I'll think about it | 10:48 |
| lambday | lisitsyn: I really want to get the difference below 1E-19 :( | 10:48 |
| @lisitsyn | lambday: what's that? an inverse of $ GSoC student receives in the welcome package? | 10:49 |
| lambday | lisitsyn: LOL I think that's 0.0002 :P | 10:50 |
| @lisitsyn | lambday: want whole money in the welcome package? | 10:51 |
| @lisitsyn | :D | 10:51 |
| lambday | lisitsyn: oh I overlooked "welcome" lol | 10:51 |
| @lisitsyn | 500 - 2250 - 2250 | 10:51 |
| lambday | that makes it 0.002 | 10:51 |
| lambday | lol | 10:51 |
| -!- iglesiasg [~iglesias@193.147.49.60] has joined #shogun | 11:00 | |
| -!- mode/#shogun [+o iglesiasg] by ChanServ | 11:00 | |
| @lisitsyn | bom dia iglesiasg | 11:00 |
| gsomix | good morning | 11:00 |
| @iglesiasg | lisitsyn, hola! | 11:01 |
| @lisitsyn | iglesiasg: como estas? | 11:02 |
| @iglesiasg | lisitsyn, bien, y tú? | 11:04 |
| @iglesiasg | learning some Spanish? | 11:04 |
| @lisitsyn | iglesiasg: no :D | 11:04 |
| @lisitsyn | iglesiasg: I am at dos cervezas por favor level in both spanish/portuguese | 11:05 |
| @lisitsyn | with no progress :D | 11:05 |
| @iglesiasg | haha | 11:06 |
| @iglesiasg | fair enough | 11:06 |
| @lisitsyn | iglesiasg: for some reason I remember | 11:09 |
| @lisitsyn | acento agudo acento grave | 11:10 |
| @lisitsyn | acento circumflexo | 11:10 |
| @lisitsyn | :D | 11:10 |
| @iglesiasg | lisitsyn, wow that's something funny to remember probably | 11:12 |
| @iglesiasg | lisitsyn, but that's French I think | 11:12 |
| -!- van51 [~van51@195.251.235.71] has joined #shogun | 11:13 | |
| @lisitsyn | iglesiasg: portuguese! | 11:13 |
| @lisitsyn | a` a' a~ a^ | 11:13 |
| @iglesiasg | aham | 11:13 |
| @lisitsyn | iglesiasg: are there such things in spanish? | 11:14 |
| @iglesiasg | iglesiasg, I didn't know they have ^ in Portuguese | 11:14 |
| @lisitsyn | ha! | 11:14 |
| @iglesiasg | talking to me again | 11:14 |
| @iglesiasg | anyway | 11:14 |
| @iglesiasg | lisitsyn, we have this one á | 11:15 |
| @iglesiasg | for every vowel | 11:15 |
| @lisitsyn | iglesiasg: I see '?' | 11:15 |
| @lisitsyn | :) | 11:15 |
| @lisitsyn | what about ~? | 11:15 |
| @iglesiasg | no | 11:15 |
| @iglesiasg | just over the n hehe | 11:15 |
| @iglesiasg | ñ | 11:15 |
| @lisitsyn | like in sa~o paulo | 11:15 |
| @lisitsyn | espan~a | 11:15 |
| @lisitsyn | I have t-shirt | 11:16 |
| @lisitsyn | espan~a | 11:16 |
| @lisitsyn | iglesiasg: so it sounds espanha? | 11:17 |
| @iglesiasg | lisitsyn, sort of | 11:17 |
| @lisitsyn | like sanh paulu | 11:17 |
| lambday | espanya? | 11:17 |
| -!- vgorbati_ [5b8ea25a@gateway/web/freenode/ip.91.142.162.90] has joined #shogun | 11:18 | |
| @lisitsyn | iglesiasg: what about the first part? | 11:18 |
| @lisitsyn | like t[ea] or [e]nd? | 11:19 |
| -!- vgorbati [c3ee5cb1@gateway/web/freenode/ip.195.238.92.177] has quit [Quit: Page closed] | 11:19 | |
| @iglesiasg | lisitsyn, end | 11:19 |
| @lisitsyn | iglesiasg: got it | 11:20 |
| @iglesiasg | pretty much like Spain :P | 11:20 |
| @lisitsyn | iglesiasg: but that [e] is here still? | 11:20 |
| hushell | sonney2k, you mean I send a PR for the documentation? | 11:20 |
| @iglesiasg | yes | 11:20 |
| @lisitsyn | iglesiasg: well russia is not russia but ro:ssiya | 11:21 |
| @lisitsyn | :D | 11:21 |
| @lisitsyn | so sometimes it differs | 11:22 |
| vgorbati_ | lisitsyn: hi, I looked through more kernels, and it looks like only string kernels and dot kernels share this pattern. I don't like the idea of subclassing both of them from some parent, instead I want to implement it this way: http://pastebin.com/PBZWmrU8. What do you think? | 11:22 |
| @lisitsyn | iglesiasg: like Austria and Oesterreich | 11:22 |
| @lisitsyn | vgorbati_: but what generality it offers? | 11:23 |
| -!- iglesiasg [~iglesias@193.147.49.60] has quit [Quit: Saliendo] | 11:24 | |
| vgorbati_ | lisitsyn: it will work for both string kernels and dot kernels | 11:24 |
| vgorbati_ | lisitsyn: for string kernels the call will look like | 11:24 |
| @lisitsyn | vgorbati_: but string kernels are using SGString right? | 11:24 |
| @lisitsyn | not SGVector I mean | 11:24 |
| vgorbati_ | lisitsyn: https://github.com/shogun-toolbox/shogun/blob/develop/src/shogun/kernel/string/LocalityImprovedStringKernel.cpp#L66 | 11:25 |
| -!- iglesiasg [~iglesias@193.147.49.92] has joined #shogun | 11:26 | |
| -!- mode/#shogun [+o iglesiasg] by ChanServ | 11:26 | |
| vgorbati_ | lisitsyn: they use char*, or uint16_t*, or uint32_t* | 11:26 |
| @lisitsyn | ah alright | 11:26 |
| vgorbati_ | lisitsyn: so we can use SGVector of char, uint_16 or uint_32 | 11:26 |
| @lisitsyn | yeah sounds like a way | 11:26 |
| @lisitsyn | vgorbati_: just to avoid one template parameter here | 11:27 |
| vgorbati_ | lisitsyn: we can use template template arguments? | 11:28 |
| @lisitsyn | vgorbati_: you may try to add a typedef | 11:28 |
| @lisitsyn | to templated features | 11:28 |
| @lisitsyn | like typedef ST ElementType; | 11:28 |
| @lisitsyn | then you can get it back in that function | 11:28 |
| @lisitsyn | with FeatureType::ElementType | 11:28 |
| vgorbati_ | lisitsyn: thanks, great idea | 11:29 |
| -!- lambday [3ba2174c@gateway/web/cgi-irc/kiwiirc.com/ip.59.162.23.76] has quit [Quit: http://www.kiwiirc.com/ - A hand crafted IRC client] | 11:32 | |
| -!- shogun-notifier- [~irker@7nn.de] has quit [Quit: transmission timeout] | 11:35 | |
| vgorbati_ | lisitsyn: then I will do it this way. I am also curious, why do StringFeatures (https://github.com/shogun-toolbox/shogun/blob/develop/src/shogun/features/StringFeatures.cpp#L222) return a copy of feature vector, while DenseFeatures (https://github.com/shogun-toolbox/shogun/blob/develop/src/shogun/features/DenseFeatures.cpp#L170) return the vector without copying? | 11:36 |
| @lisitsyn | vgorbati_: I don't remember if there is a particular reason to do that | 11:38 |
| @lisitsyn | may be due to legacy (TM) | 11:38 |
| vgorbati_ | lisitsyn: who may remember it?:) | 11:39 |
| -!- nube1 [~rho@116.90.239.3] has joined #shogun | 11:39 | |
| -!- nube [~rho@116.90.239.3] has quit [Read error: No route to host] | 11:39 | |
| @lisitsyn | vgorbati_: sonne|work may be | 11:39 |
| sonne|work | vgorbati_: lisitsyn went berserk when trying to fix that | 11:47 |
| -!- vgorbati_ [5b8ea25a@gateway/web/freenode/ip.91.142.162.90] has quit [Ping timeout: 250 seconds] | 11:48 | |
| sonne|work | vgorbati_: yes we intend to never copy ... | 11:48 |
| hushell | Hey guys, should I send a PR for a minor change of documentation? | 11:50 |
| @lisitsyn | hushell: of course | 11:54 |
| @lisitsyn | very welcome | 11:54 |
| hushell | That seems overkill, but also the only way ... | 12:01 |
| hushell | Thanks for making sure this | 12:01 |
| @lisitsyn | hushell: what seems? | 12:04 |
| @lisitsyn | ahh PR? | 12:04 |
| @lisitsyn | yes but that's the only way so far | 12:04 |
| hushell | I got an error when make unit-tests ../src/shogun/mathematics/Math.h:458: undefined reference to `shogun::CMath::rand_state' | 12:47 |
| hushell | any idea how to fix it? | 12:47 |
| hushell | It's strange after fetching the upstream the unit-tests don't compile, but make tests are okay | 12:57 |
| hushell | ./classifier/svm/SVMOcas_unittest.cc.o: In function `shogun::CMath::init_random(unsigned int)': | 12:57 |
| hushell | /home/hushell/working/git-dir/shogun-test/tests/unit/../../src/shogun/mathematics/Math.h:458: undefined reference to `shogun::CMath::rand_state' | 12:58 |
| @iglesiasg | hushell, not sure if a make clean might help there | 13:00 |
| hushell | haha you made it | 13:02 |
| hushell | the Make can detect new changes ... :( | 13:02 |
| hushell | cannot | 13:02 |
| hushell | Thanks iglesiasg | 13:03 |
| @iglesiasg | hushell, ;) | 13:03 |
| hushell | okay, I finish the PR, now I can go to sleep :) Hope you have a nice day! | 13:07 |
| -!- hushell [~hushell@c-67-189-100-116.hsd1.or.comcast.net] has quit [Quit: WeeChat 0.3.7] | 13:07 | |
| -!- iglesiasg [~iglesias@193.147.49.92] has quit [Quit: Saliendo] | 13:15 | |
| -!- sijin [~smuxi@144.214.222.109] has joined #shogun | 14:01 | |
| -!- lambday [3ba2174c@gateway/web/cgi-irc/kiwiirc.com/ip.59.162.23.76] has joined #shogun | 14:04 | |
| -!- nube1 [~rho@116.90.239.3] has quit [Quit: Leaving.] | 14:11 | |
| -!- iglesiasg [~iglesias@193.147.49.92] has joined #shogun | 14:21 | |
| -!- mode/#shogun [+o iglesiasg] by ChanServ | 14:21 | |
| -!- zxtx [~zv@ool-457e751d.dyn.optonline.net] has quit [Ping timeout: 256 seconds] | 14:26 | |
| -!- vgorbati [4d2f8427@gateway/web/freenode/ip.77.47.132.39] has joined #shogun | 14:28 | |
| vgorbati | lisitsyn: here? | 14:50 |
| @lisitsyn | vgorbati: yes | 14:50 |
| vgorbati | lisitsyn: what if we go even further and also add a typedef for a used features type inside the kernel, then we will have only one template parameter? | 14:51 |
| @lisitsyn | vgorbati: why not | 14:51 |
| vgorbati | lisitsyn: I am also wondering where to put this function. Is 'Kernel.h' a suitable place? | 14:52 |
| @lisitsyn | vgorbati: as it is template it has to go to the header | 14:54 |
| -!- thoralf [~thoralf@enki.zib.de] has joined #shogun | 14:55 | |
| thoralf | Hey. | 14:55 |
| vgorbati | lisitsyn: you mean I should create a new header file, or could I just use 'Kernel.h'? | 14:56 |
| sonne|work | thoralf: hey | 15:03 |
| sonne|work | vgorbati, lisitsyn what are you talking about? | 15:05 |
| -!- vgorbati_ [4d2f8427@gateway/web/freenode/ip.77.47.132.39] has joined #shogun | 15:06 | |
| -!- vgorbati_ [4d2f8427@gateway/web/freenode/ip.77.47.132.39] has quit [Client Quit] | 15:06 | |
| -!- vgorbati_ [4d2f8427@gateway/web/freenode/ip.77.47.132.39] has joined #shogun | 15:06 | |
| -!- vgorbati [4d2f8427@gateway/web/freenode/ip.77.47.132.39] has quit [Ping timeout: 250 seconds] | 15:06 | |
| -!- vgorbati_ is now known as vgorbati | 15:08 | |
| -!- nube [~rho@36.252.225.240] has joined #shogun | 15:16 | |
| -!- lambday [3ba2174c@gateway/web/cgi-irc/kiwiirc.com/ip.59.162.23.76] has quit [Quit: http://www.kiwiirc.com/ - A hand crafted IRC client] | 15:18 | |
| -!- nube1 [~rho@49.244.29.35] has joined #shogun | 15:38 | |
| -!- vgorbati [4d2f8427@gateway/web/freenode/ip.77.47.132.39] has quit [Ping timeout: 250 seconds] | 15:40 | |
| -!- nube [~rho@36.252.225.240] has quit [Ping timeout: 256 seconds] | 15:41 | |
| gsomix | sonney2k, sonne|work: hi. are you there? how are you? | 15:42 |
| -!- van51 [~van51@195.251.235.71] has quit [Quit: Leaving.] | 15:43 | |
| -!- HeikoS [~heiko@nat-163-207.internal.eduroam.ucl.ac.uk] has joined #shogun | 16:09 | |
| -!- mode/#shogun [+o HeikoS] by ChanServ | 16:09 | |
| -!- pickle27 [~kevin@75.98.19.132] has joined #shogun | 16:41 | |
| pickle27 | hey everyone | 16:41 |
| pickle27 | really like the new wrapped output from the makefile btw! | 16:41 |
| pickle27 | iglesiasg, did you see the merged MCLDA with eigen3? | 16:42 |
| -!- votjakovr [~votjakovr@host-46-241-3-209.bbcustomer.zsttk.net] has joined #shogun | 16:50 | |
| -!- lambday [3ba2174e@gateway/web/cgi-irc/kiwiirc.com/ip.59.162.23.78] has joined #shogun | 17:00 | |
| -!- sumit [75e247a5@gateway/web/freenode/ip.117.226.71.165] has joined #shogun | 17:00 | |
| lambday | HeikoS: hi | 17:01 |
| @HeikoS | lambday: hi! | 17:01 |
| @HeikoS | how is it going? | 17:01 |
| lambday | without using high precision thing, we cannot achieve that much high precision it seems | 17:02 |
| lambday | like arprec or so | 17:02 |
| @HeikoS | lambday: I saw | 17:02 |
| @iglesiasg | pickle27, I have just taken a look | 17:02 |
| @HeikoS | so whats this other library? | 17:02 |
| lambday | HeikoS: locally I configured shogun with arprec | 17:03 |
| lambday | log-scale thing is tough it seems | 17:03 |
| lambday | using gmc is another option | 17:03 |
| @HeikoS | lambday anything speaks against arpec? | 17:04 |
| @HeikoS | acprec | 17:04 |
| @HeikoS | lisitsyn: what do you think? you know all these libs | 17:04 |
| lambday | well, its not in debian... | 17:04 |
| lambday | HeikoS: I was thinking - if we could have methods which uses arprec when its there otherwise just uses shogun's float, it should be fine, right? | 17:04 |
| lambday | that's what I'm trying to do | 17:05 |
| @HeikoS | lambday: yes, that would be a good way, but maybe a lot of work to code up | 17:05 |
| @HeikoS | we could also just say, install lib to use the log-det samplers | 17:05 |
| @HeikoS | do you think you can manage to do this quickly? | 17:05 |
| lambday | HeikoS: we need it just for these two methods - won't take long | 17:05 |
| @lisitsyn | HeikoS: what speaks against arprec | 17:05 |
| @lisitsyn | is its absence in debian | 17:05 |
| lambday | we don't need ellPJC for real datatype anyway - just the complex version is fine | 17:06 |
| @lisitsyn | no package no honey | 17:06 |
| @lisitsyn | :D | 17:06 |
| pickle27 | ah got to go, later! | 17:06 |
| -!- pickle27 [~kevin@75.98.19.132] has quit [Quit: Leaving] | 17:06 | |
| lambday | lisitsyn: may be, those who want to use these two particular methods better use arprec :D | 17:06 |
| @HeikoS | lambday: okay then | 17:07 |
| @lisitsyn | lambday: lets chat about it a little bit later if you are online | 17:07 |
| @HeikoS | and what about http://gmplib.org/? | 17:07 |
| @lisitsyn | I am moving out from job | 17:07 |
| lambday | lisitsyn: sure.. | 17:07 |
| @lisitsyn | to my great 12m2 residence | 17:07 |
| @lisitsyn | :D | 17:07 |
| @lisitsyn | see you | 17:07 |
| -!- lisitsyn [~lisitsin@mxs.kg.ru] has quit [Quit: Leaving.] | 17:07 | |
| lambday | :) | 17:08 |
| lambday | HeikoS: gmp has gmc for complex datatypes | 17:08 |
| lambday | but its in C, which makes the work terrible | 17:08 |
| @HeikoS | lambday: woulnt that be better than? | 17:08 |
| @HeikoS | ah | 17:08 |
| @HeikoS | I see | 17:08 |
| lambday | in arprec, most of the operators are overloaded | 17:08 |
| lambday | so, spares a lot of hard work | 17:08 |
| lambday | we could use same code for most of the places as of shogun's datatype and arprec's datatypes | 17:09 |
| @HeikoS | lambday: okay then, arprec and fallback seems to be the best option | 17:09 |
| lambday | HeikoS: alright then :) | 17:09 |
| lambday | HeikoS: by the way, I added a check_arprec() in configure script for arprec, locally though | 17:10 |
| lambday | and it works it seems | 17:10 |
| lambday | shall I include that too in the PR? | 17:10 |
| @HeikoS | lambday: cool, that very good | 17:10 |
| @HeikoS | yes please | 17:10 |
| @HeikoS | so I need to install it right? ;) | 17:10 |
| lambday | HeikoS: okay... | 17:10 |
| lambday | HeikoS: lol.. yes, probably | 17:10 |
| lambday | HeikoS: I'm rewriting the methods as needed and then sending the PR | 17:11 |
| @HeikoS | okay nice | 17:11 |
| @HeikoS | you were quite fast with this | 17:11 |
| @HeikoS | really looking forward to the framework draft | 17:11 |
| lambday | HeikoS: yes, I'll come up with a class diagram for that as soon as I'm done with this | 17:12 |
| @HeikoS | lambday: good, try to keep it quite general (and clean) | 17:12 |
| @HeikoS | not talk about pre/post CG calls for now | 17:13 |
| @HeikoS | first version should be very big picture | 17:13 |
| @HeikoS | and then well start with dense matrices etc for a simple check | 17:13 |
| lambday | HeikoS: okay | 17:13 |
| lambday | I'll finish this stuff by tonight/tomorrow at the latest I'm hoping... will have get into the real work as soon as I can | 17:15 |
| lambday | HeikoS: by the way, how was the climbing? :) | 17:16 |
| @HeikoS | cool! | 17:16 |
| @HeikoS | climbing was nice | 17:16 |
| @HeikoS | unfortunately bad whether | 17:16 |
| lambday | ah! | 17:16 |
| @HeikoS | so we ended up staying in Germany, not going to Italy | 17:16 |
| @HeikoS | but we were on a pretty wall one day | 17:17 |
| @HeikoS | and two days climbing inside | 17:17 |
| @HeikoS | which is also awesome in munic | 17:17 |
| @HeikoS | 30m building | 17:17 |
| @HeikoS | hundreds of routers | 17:17 |
| lambday | wow! | 17:17 |
| @HeikoS | yeah very nice, if you are ever in munic, try it :) | 17:18 |
| @HeikoS | munich | 17:18 |
| lambday | HeikoS: ah, dying to go to europe once | 17:18 |
| lambday | I haven't been anywhere outside of india ever | 17:18 |
| @HeikoS | lambday: maybe after this gsoc, you can go to Norway to the stats guys :) | 17:19 |
| @HeikoS | they are desperately searching for phds :) | 17:19 |
| lambday | HeikoS: lol.. will think about it :) | 17:19 |
| @sonney2k | hey HeikoS ! | 17:21 |
| @sonney2k | back in town? | 17:21 |
| @HeikoS | sonney2k: yes, hi! | 17:21 |
| @HeikoS | how are things? | 17:22 |
| @sonney2k | HeikoS, we have a green build - including travis :D | 17:22 |
| @HeikoS | sonney2k: yeah! :) | 17:22 |
| @HeikoS | equals is not yet being used right? | 17:22 |
| @sonney2k | HeikoS, btw today evening we wanted to discuss with sergey right? | 17:23 |
| @sonney2k | HeikoS, nope | 17:23 |
| @HeikoS | sonney2k: is there a list of things that I need to fix apart from the sum? | 17:23 |
| @sonney2k | but it would be the real fix | 17:23 |
| @sonney2k | HeikoS, sum and product | 17:23 |
| @sonney2k | HeikoS, but it should be the same thing | 17:23 |
| gsomix | sonney2k, guck-guck :) | 17:25 |
| @sonney2k | gsomix, hey! any news on the line reader? | 17:27 |
| gsomix | sonney2k, yep, workin' on csv reader. I have some questions about IO in shogun. | 17:28 |
| @sonney2k | gsomix, well a line reader would already be a big step and worth a PR | 17:29 |
| gsomix | sonney2k, hm, ok. I don't really understood where my getline should be situated. | 17:30 |
| @HeikoS | sonney2k: I just discussed with Sergey a little | 17:30 |
| gsomix | In CFile? or in AsciiFile? | 17:31 |
| gsomix | btw there is implementation of getline into CAsciiFile, but it's not used in shogun code anywhere. | 17:32 |
| -!- sumit [75e247a5@gateway/web/freenode/ip.117.226.71.165] has quit [Ping timeout: 250 seconds] | 17:33 | |
| -!- iglesiasg [~iglesias@193.147.49.92] has quit [Quit: Saliendo] | 18:01 | |
| lambday | I am getting weird errors in make unit-tests after I did git pull --rebase upstream... in existing unit-test for MCLDA_unittest.cc :-/ | 18:07 |
| lambday | "error: ‘CMCLDA’ was not declared in this scope", ...and so on! | 18:08 |
| @HeikoS | lambday: try make clean before | 18:12 |
| lambday | HeikoS: I did... also did git clean -fdx, didn't help :( | 18:13 |
| -!- gsomix_ [~gsomix@37.61.180.77] has joined #shogun | 18:13 | |
| -!- gsomix [~gsomix@37.61.179.131] has quit [Disconnected by services] | 18:13 | |
| -!- gsomix_ is now known as gsomix | 18:13 | |
| @HeikoS | in the unit test folder? | 18:13 |
| lambday | no in the src | 18:13 |
| -!- van51 [~van51@athedsl-320452.home.otenet.gr] has joined #shogun | 18:18 | |
| -!- lambday [3ba2174e@gateway/web/cgi-irc/kiwiirc.com/ip.59.162.23.78] has quit [Quit: http://www.kiwiirc.com/ - A hand crafted IRC client] | 18:23 | |
| -!- sijin_ [~smuxi@144.214.222.109] has joined #shogun | 18:23 | |
| -!- sijin [~smuxi@144.214.222.109] has quit [Ping timeout: 264 seconds] | 18:26 | |
| -!- lambday [e8b6149@gateway/web/cgi-irc/kiwiirc.com/x-qycumxkxgmepyzsx] has joined #shogun | 18:40 | |
| thoralf | mkl.set_interleaved_optimization_enabled(true); produces an endless loop? | 18:41 |
| thoralf | Infinitly printing [DEBUG] entering CCombinedKernel::get_subkernel_weights(), [DEBUG] not appending kernel weights | 18:41 |
| thoralf | Is it a known bug or should it be working? | 18:42 |
| -!- lisitsyn [~blackburn@37.61.180.77] has joined #shogun | 18:42 | |
| -!- mode/#shogun [+o lisitsyn] by ChanServ | 18:42 | |
| thoralf | setting interleaved optimization to "false" works as expected. | 18:43 |
| lambday | HeikoS: may be I'm doing something wrong.. but unit-tests not working :( | 18:43 |
| lambday | HeikoS: just for testing, I cloned from shogun-toolbox/shogun into a separate directory and it gives same error | 18:45 |
| lambday | HeikoS: oh I see... that unit-test doesn't check if Eigen3 exists, I configured without Eigen3 for making current things fast and it was giving the error | 18:50 |
| lambday | HeikoS: I think unit tests should check if Eigen3 exists too, no? | 18:51 |
| @HeikoS | yes! | 18:52 |
| @HeikoS | could you do that? | 18:52 |
| @HeikoS | :) | 18:52 |
| lambday | HeikoS: yes :) | 18:52 |
| lambday | I'll check if any other unit-test fails in absense of Eigen3 | 18:52 |
| @lisitsyn | HeikoS: lambday: so what did you decide? | 19:03 |
| lambday | lisitsyn: to use arprec when its there, else fallback to shogun's datatypes | 19:04 |
| @lisitsyn | I want to warn you ~ nobody is going to use that | 19:04 |
| @lisitsyn | people don't like that | 19:04 |
| @HeikoS | lisitsyn: maybe you are right, lambday, how much effort to use gmp? would it be possible to use that? dont really know whats best here | 19:05 |
| @HeikoS | gotta go now, will be back tomorrow (full day then) | 19:06 |
| @HeikoS | bye | 19:06 |
| @lisitsyn | bye | 19:06 |
| lambday | HeikoS: okay | 19:06 |
| lambday | HeikoS: bye | 19:06 |
| lambday | lisitsyn: you are right - but I was thinking, since its just for two methods, and its in C++ and the operators are overloaded.. for complex number gmp provides gmc... which is in C | 19:07 |
| lambday | HeikoS: I'll check it then | 19:07 |
| -!- hushell [~hushell@c-67-189-100-116.hsd1.or.comcast.net] has joined #shogun | 19:14 | |
| lambday | lisitsyn: is it okay to work with arprec for now (cause high precision is really needed for this)... and later may be, replace it with gmp/gmc? (just for two methods, won't take too much work later) | 19:15 |
| @lisitsyn | lambday: yes sure | 19:15 |
| lambday | lisitsyn: okay then | 19:15 |
| lambday | :) | 19:15 |
| @lisitsyn | lambday: it is just a matter of your will to deliver the software | 19:16 |
| @lisitsyn | I mean if it is easier to get it to work it is usually better | 19:16 |
| lambday | lisitsyn: arprec is easier to work with in c++, gmc - I need to check out | 19:18 |
| -!- lambday [e8b6149@gateway/web/cgi-irc/kiwiirc.com/x-qycumxkxgmepyzsx] has quit [Quit: http://www.kiwiirc.com/ - A hand crafted IRC client] | 19:19 | |
| @sonney2k | HeikoS, sorry had to leave train | 19:26 |
| -!- shogun-notifier- [~irker@7nn.de] has joined #shogun | 19:27 | |
| shogun-notifier- | shogun: hushell :develop * b0b6b6b / src/shogun/multiclass/MulticlassStrategy.h: https://github.com/shogun-toolbox/shogun/commit/b0b6b6b7c2000b1fdb145c4a4f14de5e4d9ea10f | 19:27 |
| shogun-notifier- | shogun: update doc of rescale_outputs | 19:27 |
| shogun-notifier- | shogun: hushell :develop * ae06fc1 / src/shogun/multiclass/MulticlassStrategy.h: https://github.com/shogun-toolbox/shogun/commit/ae06fc14f643d80269d107249244b0a34edf46ff | 19:27 |
| shogun-notifier- | shogun: fix indent issues | 19:27 |
| shogun-notifier- | shogun: Soeren Sonnenburg :develop * efbbbc4 / src/shogun/multiclass/MulticlassStrategy.h: https://github.com/shogun-toolbox/shogun/commit/efbbbc48e2ad9c8b136855a501e7ce6cb1491492 | 19:27 |
| shogun-notifier- | shogun: Merge pull request #1118 from hushell/develop | 19:27 |
| shogun-notifier- | shogun: | 19:27 |
| shogun-notifier- | shogun: fix a wrong documentation of CMulticlassStrategy::rescale_outputs | 19:27 |
| -!- lambday [3ba2174d@gateway/web/cgi-irc/kiwiirc.com/ip.59.162.23.77] has joined #shogun | 19:34 | |
| -!- lambday [3ba2174d@gateway/web/cgi-irc/kiwiirc.com/ip.59.162.23.77] has quit [Quit: http://www.kiwiirc.com/ - A hand crafted IRC client] | 19:39 | |
| @sonney2k | gsomix, in CAsciiFile a static function | 19:54 |
| @sonney2k | wiking, hey there prng hero | 19:55 |
| @sonney2k | wiking, I am going to rewrite all the prng functions | 19:56 |
| gsomix | sonney2k, yep, got it. I think I should change get_string_list functions and make them work with my getline. right? | 19:58 |
| @lisitsyn | sonney2k: just for fun oder? | 20:01 |
| @sonney2k | lisitsyn, what? | 20:08 |
| @lisitsyn | sonney2k: why to rewrite all prng? | 20:08 |
| @sonney2k | gsomix, *any* line reading stuff will benefit from that | 20:09 |
| @sonney2k | gsomix, but first we need a fast line reader | 20:09 |
| @sonney2k | lisitsyn, now that we have mersenne the PRNG doesn't have to use higher order bits | 20:10 |
| @sonney2k | so we can use lower order bits and stuff like rand_64() % 10 will work *very* nicely | 20:10 |
| -!- votjakovr [~votjakovr@host-46-241-3-209.bbcustomer.zsttk.net] has quit [Quit: Leaving] | 20:10 | |
| @sonney2k | lisitsyn, with the rand_64() % n; way of generating prn's we are like twice as fast | 20:26 |
| @lisitsyn | sonney2k: what is the way with higher order bits you are talkign about? | 20:26 |
| @sonney2k | the old rand() was less random in its lower bits | 20:27 |
| @sonney2k | so doing rand() % n is not a good idea | 20:27 |
| @lisitsyn | got it | 20:29 |
| @sonney2k | lisitsyn, ooh look | 20:31 |
| @sonney2k | the situation is already better at least with newer libc's | 20:31 |
| @sonney2k | NOTES | 20:32 |
| @sonney2k | The versions of rand() and srand() in the Linux C Library use the same random number generator as random(3) and srandom(3), so the lower-order bits should be as random as the higher-order bits. However, on older rand() implementations, and on current imple‐ | 20:32 |
| @sonney2k | mentations on different systems, the lower-order bits are much less random than the higher-order bits. Do not use this function in applications intended to be portable when good randomness is needed. (Use random(3) instead.) | 20:32 |
| @lisitsyn | haha | 20:32 |
| @lisitsyn | not enough entropy | 20:32 |
| @sonney2k | we did use random btw | 20:32 |
| @lisitsyn | sonney2k: the importance of reading TFM? | 20:33 |
| @lisitsyn | TFM from RTFM I mean | 20:34 |
| @sonney2k | lisitsyn, what do you mean? | 20:34 |
| @lisitsyn | sonney2k: I didn't know that | 20:34 |
| @sonney2k | well it is obsolete knowledge | 20:35 |
| @sonney2k | it is like std::vector not being guaranteed to be contiguous memory | 20:35 |
| @sonney2k | no longer true | 20:35 |
| @lisitsyn | haha | 20:35 |
| @lisitsyn | sonney2k: good you believe that now | 20:35 |
| @sonney2k | lisitsyn, I didn't doubt that since I read the standard | 20:36 |
| @lisitsyn | sonney2k: furthermore in C++11 vector doesn't copy things twice | 20:36 |
| @sonney2k | but it was not like this when I started | 20:36 |
| @lisitsyn | sonney2k: bad design decision | 20:36 |
| @sonney2k | lisitsyn, in 10 years you will face the same things(tm) | 20:36 |
| @lisitsyn | sonney2k: vector<bool> | 20:37 |
| @lisitsyn | sonney2k: I am trying to be modern :P | 20:37 |
| @sonney2k | lisitsyn, yeah I have no idea why they didn't require things to contiguous in the first place | 20:37 |
| @lisitsyn | sonney2k: I understand the motivation beneath | 20:37 |
| @sonney2k | lisitsyn, like us 14 years ago :P | 20:37 |
| @sonney2k | look when we started people were screaming templates - hahaha! | 20:38 |
| @sonney2k | you won't get that compiled | 20:38 |
| @lisitsyn | sonney2k: haha | 20:38 |
| @lisitsyn | and now template MPL is usual thing | 20:38 |
| @sonney2k | and indeed compilers had *lots* of issues back then with templates | 20:38 |
| @lisitsyn | sonney2k: http://html9responsiveboilerstrapjs.com/ | 20:38 |
| @lisitsyn | sonney2k: Just attackclone the grit repo pushmerge, then rubygem the lymphnode js shawarma module – and presto! | 20:39 |
| @lisitsyn | that's the future | 20:39 |
| @lisitsyn | :D | 20:39 |
| @sonney2k | alright back to rand & friends | 20:41 |
| @lisitsyn | sonney2k: have you ever seen notation 'R^{++}' | 20:46 |
| @lisitsyn | ? | 20:46 |
| @sonney2k | maybe x \in R but with x >0 ? | 20:57 |
| @lisitsyn | sonney2k: yes that's R+ | 20:57 |
| @lisitsyn | but R++ | 20:57 |
| @lisitsyn | I guess a typo | 20:57 |
| @sonney2k | well for some R^+ might include 0 | 20:57 |
| @lisitsyn | ahh | 20:57 |
| @sonney2k | some also write R^+_0 | 20:57 |
| -!- iglesiasg [d58f329d@gateway/web/freenode/ip.213.143.50.157] has joined #shogun | 21:21 | |
| -!- mode/#shogun [+o iglesiasg] by ChanServ | 21:21 | |
| gsomix | sonney2k, done. PR was sent. | 21:52 |
| @sonney2k | gsomix, that is super slow | 22:08 |
| @sonney2k | you do getc there | 22:08 |
| @sonney2k | so reading character by character | 22:09 |
| @sonney2k | we should rather read big blocks (say 1MB in size) and then return a line | 22:10 |
| gsomix | sonney2k, ok! so, should I use something like fread for reading blocks? | 22:12 |
| @sonney2k | gsomix, yes and then some ptr/cache for the buffer | 22:14 |
| gsomix | sonney2k, ok, I'll do it. tnx for fast review. | 22:15 |
| @sonney2k | lisitsyn, any chance that you use in shogun's tapkee the random functions? | 22:19 |
| @wiking | murmur | 22:24 |
| @sonney2k | wiking, hey there | 22:25 |
| @sonney2k | sec | 22:25 |
| @wiking | yo | 22:25 |
| -!- shogun-notifier- [~irker@7nn.de] has quit [Quit: transmission timeout] | 22:27 | |
| @sonney2k | wiking, I am switching to using your rand_64() % n to generate 0...n-1'prn's | 22:28 |
| @sonney2k | wiking, that really is like 2 times faster than what we had before | 22:29 |
| @sonney2k | and I've written some tests that check if stuff is still uniform | 22:29 |
| @sonney2k | and yes all good | 22:29 |
| @wiking | \o/ | 22:30 |
| @sonney2k | this will kill all our tests though | 22:30 |
| @wiking | you mean integration tests? | 22:30 |
| @sonney2k | yes | 22:30 |
| @sonney2k | since random numbers are no longer comparable | 22:30 |
| @wiking | btw: the CMath::random(...) were using rand_64 already | 22:30 |
| @sonney2k | wiking, but not the way I am using them now | 22:31 |
| @wiking | mmmm how are u using them now? | 22:31 |
| @wiking | i mean what's the diff ? :) | 22:31 |
| @sonney2k | wiking, how do I upload the feature branch? | 22:31 |
| @sonney2k | just git push? | 22:32 |
| @wiking | welll | 22:32 |
| @wiking | yeah git push origin feature/FEATURENAME | 22:32 |
| @wiking | origin = upstream | 22:32 |
| @wiking | or whatever you named the shogun.git | 22:33 |
| @wiking | i mean shogun-toolbox/shogun.git | 22:33 |
| @wiking | of course you can do the fancy | 22:33 |
| @sonney2k | https://github.com/shogun-toolbox/shogun/tree/feature/prng_lower_order_bits | 22:33 |
| @wiking | git flow feature publish <feature> | 22:33 |
| @wiking | but that'll be published on your own repo | 22:33 |
| @sonney2k | wiking, I've moved the content of the CMath:random* functions into CRandom | 22:34 |
| @sonney2k | only calling the random stuff | 22:34 |
| @sonney2k | from CMath | 22:34 |
| @wiking | mmm | 22:35 |
| @wiking | i see the unittests :) | 22:35 |
| @wiking | which commit should i check? | 22:36 |
| @sonney2k | oops forgot to commit | 22:40 |
| @sonney2k | sec | 22:40 |
| -!- shogun-notifier- [~irker@7nn.de] has joined #shogun | 22:41 | |
| shogun-notifier- | shogun: Soeren Sonnenburg :feature/prng_lower_order_bits * 32fbfc8 / / (5 files): https://github.com/shogun-toolbox/shogun/commit/32fbfc8e7410c86968ab57e2e235d33bb90782c3 | 22:41 |
| shogun-notifier- | shogun: use lower order bits in prng | 22:41 |
| shogun-notifier- | shogun: | 22:41 |
| shogun-notifier- | shogun: - This is about 2 times faster compared to the previous rng. | 22:41 |
| shogun-notifier- | shogun: - Add unit tests to ensure things are correct | 22:41 |
| @sonney2k | wiking, this commit | 22:41 |
| @sonney2k | wiking, would be good if you could give some feedback | 22:42 |
| @sonney2k | wiking, btw the closed interval rng [0,1] never really returned 0 nor 1 for me | 22:42 |
| @sonney2k | it is like very close but not exactly 0 nor 1 | 22:43 |
| @sonney2k | (like 1e-6 or 1-1e-6) | 22:43 |
| @wiking | sonney2k: #define RAND_MAX_32 | 22:43 |
| @wiking | where did this go? | 22:43 |
| @wiking | btw | 22:44 |
| @wiking | random_close | 22:44 |
| @wiking | is slow | 22:44 |
| @wiking | i highly suggest to use [0,1) or (0,1] functions for float random | 22:44 |
| @wiking | they are even faster | 22:45 |
| -!- dvalcarce [5b757556@gateway/web/freenode/ip.91.117.117.86] has joined #shogun | 22:46 | |
| @wiking | so either random_open or random_half_open | 22:46 |
| @wiking | as it's using dsfmt | 22:47 |
| @wiking | and not sfmt | 22:47 |
| dvalcarce | hi all | 22:47 |
| @wiking | which is exactly designed for float numbers... while sfmt is designed for integer number generation | 22:47 |
| @sonney2k | wiking, I think for us the integer based floats need to be fastest | 22:48 |
| @sonney2k | wiking, e.g. liblinear uses them to select the next variable | 22:48 |
| @sonney2k | dvalcarce, hey | 22:48 |
| dvalcarce | sonney2k, hi | 22:49 |
| @wiking | sonney2k: yeah right but still | 22:49 |
| @wiking | sonney2k: currently this way dsfmt is not used ever | 22:49 |
| @wiking | sonney2k: or barely ever... so it's useless then to have it in the repo | 22:50 |
| @sonney2k | wiking, well feel free to add random_open functions | 22:51 |
| @sonney2k | wiking, I most of the time needed to generate say initial probabilites so 0...1 (closed intervals) | 22:52 |
| @sonney2k | dvalcarce, how is it going? | 22:52 |
| @sonney2k | dvalcarce, exams on the way bakufo progressing? | 22:52 |
| dvalcarce | sonney2k, I had my first final today | 22:53 |
| dvalcarce | so this night is bakufu's night :) | 22:54 |
| @sonney2k | heh | 22:56 |
| @sonney2k | wiking, alright please give feedback by tomorrow | 22:56 |
| @sonney2k | I need to sleep now... | 22:56 |
| @sonney2k | n8 everyone | 22:56 |
| @iglesiasg | good night sonney2k | 22:57 |
| gsomix | sonney2k, nite | 22:58 |
| dvalcarce | sonney2k, nite | 22:58 |
| -!- hushell [~hushell@c-67-189-100-116.hsd1.or.comcast.net] has quit [Ping timeout: 252 seconds] | 22:59 | |
| gsomix | so, good night, guys | 23:01 |
| @iglesiasg | night | 23:02 |
| -!- travis-ci [~travis-ci@ec2-23-23-66-69.compute-1.amazonaws.com] has joined #shogun | 23:07 | |
| travis-ci | [travis-ci] it's Soeren Sonnenburg'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/7371170 | 23:07 |
| -!- travis-ci [~travis-ci@ec2-23-23-66-69.compute-1.amazonaws.com] has left #shogun [] | 23:07 | |
| -!- hushell [~hushell@8-92.ptpg.oregonstate.edu] has joined #shogun | 23:12 | |
| --- Log closed Wed May 22 00:00:05 2013 | ||
Generated by irclog2html.py 2.10.0 by Marius Gedminas - find it at mg.pov.lt!