--- Log opened Wed May 09 00:00:37 2012 | ||
wiking | lol | 00:01 |
---|---|---|
wiking | i've just checked the git blame classifier_qda.cpp | 00:01 |
wiking | how it morphed into this non-working code :))) | 00:01 |
wiking | https://github.com/shogun-toolbox/shogun/pull/513 | 00:04 |
blackburn | arrghh | 00:04 |
CIA-113 | shogun: Viktor Gal master * r7dfdcba / examples/undocumented/libshogun/classifier_qda.cpp : Fix classifier_qda example - http://git.io/Ep7fDw | 00:05 |
wiking | mkl is still a magic for me | 00:07 |
CIA-113 | shogun: Sergey Lisitsyn master * r585ebed / src/shogun/converter/Isomap.cpp : Fix crasher in Isomap - http://git.io/RrIpuA | 00:13 |
-!- sonney2k_ [~shogun@7nn.de] has quit [Ping timeout: 260 seconds] | 00:16 | |
-!- sonney2k [~shogun@7nn.de] has joined #shogun | 00:17 | |
-!- shogun-buildbot [~shogun-bu@7nn.de] has quit [Ping timeout: 255 seconds] | 00:18 | |
-!- shogun-buildbot [~shogun-bu@7nn.de] has joined #shogun | 00:18 | |
blackburn | wiking: still there? | 00:34 |
wiking | y | 00:37 |
-!- PhilTillet [~Philippe@157.159.42.46] has quit [Quit: Leaving] | 00:37 | |
blackburn | wiking: you made me recall one anecdote about putin | 00:37 |
blackburn | hmm let me try to tell it :D | 00:37 |
blackburn | moscow, traffic jam, a lot of cars | 00:38 |
blackburn | one guy is running and shout - putin is captured by aliens | 00:38 |
blackburn | "they threaten to pour him with gasoline and set on fire!" | 00:39 |
blackburn | "and they ask money" | 00:39 |
blackburn | he comes to one car and driver says 'how much do you need?' - that guys says 'how much you can' - and driver answers - 'i can give you 5 liters' | 00:40 |
blackburn | :D | 00:40 |
wiking | :DDDD | 00:40 |
CIA-113 | shogun: Sergey Lisitsyn master * r276fa54 / (4 files): Fixed crashers in PRC/ROC - http://git.io/L9NjOQ | 00:50 |
-!- cronor [~cronor@g231219248.adsl.alicedsl.de] has joined #shogun | 00:59 | |
-!- blackburn [~qdrgsm@83.234.54.222] has quit [Quit: Leaving.] | 01:00 | |
CIA-113 | shogun: Sergey Lisitsyn master * ra2c0e99 / (2 files): Fixed sizes of ROC and PRC graphs - http://git.io/8sLYSw | 01:01 |
-!- wiking [~wiking@huwico/staff/wiking] has quit [Quit: wiking] | 01:23 | |
-!- wiking [~wiking@78-23-189-112.access.telenet.be] has joined #shogun | 01:24 | |
-!- wiking [~wiking@78-23-189-112.access.telenet.be] has quit [Client Quit] | 01:24 | |
-!- av3ngr [av3ngr@nat/redhat/x-esiwbjbcmoextztr] has joined #shogun | 01:57 | |
-!- vikram360 [~vikram360@117.192.173.175] has joined #shogun | 02:38 | |
-!- cronor [~cronor@g231219248.adsl.alicedsl.de] has quit [Read error: Connection reset by peer] | 03:05 | |
-!- vikram360 [~vikram360@117.192.173.175] has quit [Ping timeout: 240 seconds] | 03:14 | |
-!- cronor [~cronor@g231219248.adsl.alicedsl.de] has joined #shogun | 04:00 | |
-!- wiking [~wiking@78-23-189-112.access.telenet.be] has joined #shogun | 04:43 | |
-!- wiking [~wiking@78-23-189-112.access.telenet.be] has quit [Changing host] | 04:43 | |
-!- wiking [~wiking@huwico/staff/wiking] has joined #shogun | 04:43 | |
-!- wiking [~wiking@huwico/staff/wiking] has quit [Quit: wiking] | 05:06 | |
-!- vikram360 [~vikram360@117.192.181.205] has joined #shogun | 05:15 | |
-!- mode/#shogun [+o sonney2k] by ChanServ | 05:28 | |
-!- wiking [~wiking@78-23-189-112.access.telenet.be] has joined #shogun | 05:45 | |
-!- wiking [~wiking@78-23-189-112.access.telenet.be] has quit [Changing host] | 05:45 | |
-!- wiking [~wiking@huwico/staff/wiking] has joined #shogun | 05:45 | |
-!- wiking [~wiking@huwico/staff/wiking] has quit [Quit: wiking] | 06:15 | |
-!- wiking [~wiking@78-23-189-112.access.telenet.be] has joined #shogun | 06:24 | |
-!- wiking [~wiking@78-23-189-112.access.telenet.be] has quit [Changing host] | 06:24 | |
-!- wiking [~wiking@huwico/staff/wiking] has joined #shogun | 06:24 | |
-!- gsomix [~gsomix@178.45.78.190] has joined #shogun | 06:31 | |
-!- vikram360 [~vikram360@117.192.181.205] has quit [Ping timeout: 265 seconds] | 06:35 | |
-!- gsomix [~gsomix@178.45.78.190] has quit [Quit: Ex-Chat] | 06:36 | |
-!- gsomix [~gsomix@178.45.78.190] has joined #shogun | 06:37 | |
gsomix | morning | 06:37 |
wiking | heheey | 06:37 |
-!- gsomix [~gsomix@178.45.78.190] has quit [Remote host closed the connection] | 06:56 | |
-!- gsomix [~gsomix@178.45.78.190] has joined #shogun | 06:59 | |
wiking | gsomix: what'sup? | 07:06 |
wiking | actually how's that little task going that we've talked some weeks before? | 07:06 |
gsomix | wiking, in progress. ^__^'' | 07:14 |
wiking | yeey | 07:15 |
wiking | lol man | 07:20 |
wiking | "Yahoo Board Member Out Over CEO's Bogus R?sum?" | 07:20 |
wiking | i wonder how can u mix up your degree by mistake :D | 07:20 |
-!- n4nd0 [~nando@s83-179-44-135.cust.tele2.se] has joined #shogun | 07:51 | |
n4nd0 | good morning! | 07:51 |
n4nd0 | from Sweden finally :) | 07:51 |
@sonney2k | wiking, yeah indeed I am not happy about std:: crap in pluskids patch | 08:00 |
@sonney2k | since we have alternatives it is not needed | 08:00 |
@sonney2k | gsomix, btw please bring back CSet! | 08:01 |
wiking | sonney2k: heheheh yeah i guessed... | 08:05 |
-!- vikram360 [~vikram360@117.192.181.205] has joined #shogun | 08:07 | |
wiking | sonney2k: so u would go with CMap and CSet i guess, right? | 08:07 |
wiking | as soon as we have CSet as well | 08:07 |
@sonney2k | wiking, we had CSet - no idea why gsomix removed it... | 08:13 |
wiking | sonney2k: yeah i saw it has morphed into CMap | 08:13 |
@sonney2k | wiking, hmmhh I am wondering if 20 byte overhead for a sparse feature vector / string are ok | 08:14 |
wiking | sonney2k: hahahah really? :))) | 08:14 |
@sonney2k | wiking, you don't? | 08:15 |
wiking | well 20 bytes | 08:15 |
wiking | it's really about 5 pointers :) | 08:15 |
@sonney2k | ptr is 8 byte | 08:16 |
@sonney2k | 2 ptrs + one int | 08:16 |
@sonney2k | something is weird with multiclasssvms | 08:16 |
@sonney2k | I cannot call apply() w/o getting an error | 08:16 |
@sonney2k | wiking, for sgvectors we internally do't have a list of vectors | 08:17 |
@sonney2k | but just a double* aka SGMatrix | 08:17 |
@sonney2k | so that doesn't matter | 08:17 |
@sonney2k | but for SGSparseMatrix we wouldh ave a list of SGSparseVector.... | 08:17 |
wiking | sonney2k: on 64bits... | 08:17 |
wiking | otherwise it's 4 bytes | 08:17 |
@sonney2k | it is nice that one can share it | 08:17 |
@sonney2k | but bad memory wise... | 08:18 |
@sonney2k | thinking of that I trained on 50 million strings | 08:18 |
@sonney2k | that would mean 1G overhead | 08:18 |
@sonney2k | ok got to leave train | 08:19 |
@sonney2k | brb | 08:19 |
gsomix | oh, added | 08:26 |
gsomix | but I will commit in car | 08:27 |
gsomix | going to Samara | 08:27 |
wiking | :))))) | 08:27 |
wiking | commit in car :> | 08:27 |
sonne|work | gsomix: push please :) | 08:30 |
wiking | sonne|work: arrived to work? :) | 08:30 |
sonne|work | yeah | 08:31 |
-!- gsomix [~gsomix@178.45.78.190] has quit [Ping timeout: 265 seconds] | 08:32 | |
wiking | seems gsomix is in the car commiting | 08:32 |
wiking | mmmm | 08:36 |
wiking | kernel_matrix = 0x0, | 08:36 |
wiking | lhs = 0x0, | 08:36 |
wiking | rhs = 0x0, | 08:36 |
wiking | on a supposedly trained kernel | 08:36 |
* wiking wonders | 08:36 | |
wiking | oooh yeah | 08:38 |
sonne|work | ? | 08:42 |
-!- uricamic [~uricamic@2001:718:2:1634:f2de:f1ff:fecf:a6a5] has joined #shogun | 08:44 | |
wiking | i guess we are missing a kernel.init () in classifier_mklmulticlass.cpp | 08:44 |
wiking | that's why both lhs and rhs is NULL | 08:44 |
wiking | and that's why it dies with an assertion failure | 08:45 |
sonne|work | wiking: I was looking at that example but there is more that is missing it seems | 08:46 |
sonne|work | needs to be debugged... | 08:46 |
sonne|work | but had to leave train before finishing | 08:46 |
wiking | index_t is unsigned | 08:47 |
wiking | right? | 08:47 |
wiking | mmm not :) | 08:47 |
wiking | why not: shogun/lib/DataType.h:typedef int32_t index_t; ? | 08:47 |
sonne|work | ? | 08:48 |
sonne|work | yes int32_t | 08:48 |
sonne|work | wiking: just because you need quite a bit of extra care when dealing with it | 08:48 |
wiking | but why is an index type is signed? | 08:49 |
sonne|work | when you do things like idx - 100 | 08:49 |
wiking | i mean for sure there's a reason | 08:49 |
wiking | i just wonder why | 08:49 |
sonne|work | you cannot check < 0 | 08:49 |
sonne|work | err >=0 | 08:49 |
wiking | yes but you can check idx < 100 | 08:50 |
wiking | ;) | 08:50 |
sonne|work | sure | 08:50 |
wiking | it's essentially the same amount of branching | 08:50 |
sonne|work | but it needs extra care | 08:50 |
wiking | either this or that way it'll point u to a bad memory if u don't do a check | 08:51 |
wiking | i.e. something[idx-100] where idx < 100 | 08:51 |
wiking | it's the same in case of a signed and an unsigned integer... | 08:51 |
wiking | just the bad memory is pointed into a different place :) | 08:52 |
sonne|work | no | 08:52 |
sonne|work | if any operation happens to idx | 08:52 |
sonne|work | checking idx will not rescue | 08:53 |
sonne|work | you | 08:53 |
wiking | i'm just saying since a) essential negative index has no meaning b) u get an extra bit :) | 08:53 |
sonne|work | give me an example where we need the extra bit | 08:53 |
wiking | i mean that's 2* amount of indexable | 08:53 |
sonne|work | I know | 08:54 |
wiking | and since you were talking here about a 20byte overhead | 08:54 |
wiking | and a huge dataset | 08:54 |
wiking | if we really want to support those size | 08:54 |
sonne|work | 50 million << 2**31 | 08:54 |
wiking | that 2* will count | 08:54 |
sonne|work | no problem | 08:54 |
wiking | yes i know | 08:54 |
sonne|work | I would need 2 billion | 08:54 |
sonne|work | and that is still orders of magnitude bigger | 08:54 |
wiking | mmm | 08:55 |
sonne|work | and I would rather switch to using int64_t if we have to switch | 08:55 |
wiking | ok i think i cannot have right | 08:55 |
wiking | so i give up | 08:55 |
wiking | fine then just let's try to use the same type, i.e. index_t for the variable that is actually doing the indexing | 08:56 |
wiking | so if one day index_t is actually going to become int64_t then there's not problem | 08:56 |
sonne|work | wiking: yeah - problem is that this would need a review of *all* code | 08:57 |
sonne|work | I suspect 1% uses index_t | 08:57 |
wiking | sonne|work: yeah sure i know | 08:57 |
wiking | but still we can take care of the ones that are pull requested from now on | 08:57 |
sonne|work | who wants to do that just for fun? | 08:57 |
sonne|work | yeah | 08:57 |
sonne|work | that makes sense | 08:57 |
sonne|work | index_t is a bit confusing though | 08:58 |
sonne|work | as it is not what we use in SG MALLOC etc | 08:58 |
sonne|work | there we of course use 64bit... | 08:58 |
wiking | well yeah size_t | 08:59 |
wiking | git grep vlen|grep for|wc -l | 09:01 |
wiking | 95 | 09:01 |
wiking | :D | 09:01 |
wiking | ahhaha this is heavy shit: shogun/lib/SGVector.h:for (int32_t i=0; i<vlen; i++) | 09:02 |
wiking | :> | 09:02 |
wiking | mmm and even the api is defined | 09:02 |
wiking | as len is int32_T | 09:03 |
wiking | yeah u are right it'll take some time | 09:03 |
-!- cronor [~cronor@g231219248.adsl.alicedsl.de] has quit [Read error: Connection reset by peer] | 09:03 | |
sonne|work | wiking: you have to know that we didn't (and still don't) even use SGVector that much | 09:03 |
sonne|work | lots of stuff operates on double* | 09:04 |
wiking | sonne|work: yeah i can see | 09:04 |
sonne|work | and some len | 09:04 |
sonne|work | I hope that will change now | 09:04 |
wiking | i mean 95 lines is not so much compared to the whole LoC of shogun | 09:04 |
sonne|work | but still it is a looong way | 09:04 |
sonne|work | alright work!@ | 09:04 |
-!- cronor [~cronor@g231219248.adsl.alicedsl.de] has joined #shogun | 09:05 | |
-!- gsomix [~gsomix@85.26.235.21] has joined #shogun | 09:48 | |
gsomix | hehe | 09:48 |
gsomix | Programming in car is bad idea | 09:49 |
sonne|work | gsomix: don't drive and code | 09:49 |
sonne|work | there are not enough traffic jams I guess :) | 09:49 |
gsomix | aha | 09:51 |
gsomix | I will come to Samara in an hour | 09:56 |
gsomix | 113km from home | 09:58 |
-!- av3ngr [av3ngr@nat/redhat/x-esiwbjbcmoextztr] has quit [Quit: That's all folks!] | 09:59 | |
-!- gsomix [~gsomix@85.26.235.21] has quit [Read error: Connection reset by peer] | 09:59 | |
-!- gsomix [~gsomix@85.26.235.21] has joined #shogun | 09:59 | |
gsomix | 113km from home | 09:59 |
gsomix | bad internet :) | 10:00 |
-!- gsomix [~gsomix@85.26.235.21] has quit [Ping timeout: 256 seconds] | 10:11 | |
-!- cronor [~cronor@g231219248.adsl.alicedsl.de] has quit [Quit: cronor] | 10:13 | |
-!- wiking [~wiking@huwico/staff/wiking] has quit [Quit: wiking] | 10:47 | |
CIA-113 | shogun: Soeren Sonnenburg master * r2c888f8 / (3 files in 2 dirs): | 10:51 |
CIA-113 | shogun: Merge pull request #509 from pluskid/kernelmc-store-model | 10:51 |
CIA-113 | shogun: (+5 more commits...) - http://git.io/Gm3g3g | 10:51 |
-!- cronor [~cronor@fb.ml.tu-berlin.de] has joined #shogun | 10:52 | |
-!- wiking [~wiking@we02c096.ugent.be] has joined #shogun | 11:00 | |
-!- wiking [~wiking@we02c096.ugent.be] has quit [Changing host] | 11:00 | |
-!- wiking [~wiking@huwico/staff/wiking] has joined #shogun | 11:00 | |
-!- blackburn [~qdrgsm@83.234.54.222] has joined #shogun | 11:30 | |
-!- gsomix [~gsomix@188.168.4.56] has joined #shogun | 11:58 | |
gsomix | ok, I'm in Samara | 11:58 |
-!- cronor [~cronor@fb.ml.tu-berlin.de] has quit [Ping timeout: 252 seconds] | 12:01 | |
-!- cronor [~cronor@fb.ml.tu-berlin.de] has joined #shogun | 12:05 | |
blackburn | gsomix: is it you knock-knock? :) | 12:16 |
gsomix | blackburn, aha | 12:16 |
blackburn | I had been kind of sleeping yet this time | 12:16 |
blackburn | @egorov I'm at Reichstag w/ @kantaria and 81,034 others | 12:18 |
blackburn | :D | 12:18 |
blackburn | sonne|work: hmm I have no clang complete | 12:39 |
blackburn | do you use it? | 12:39 |
-!- vikram360 [~vikram360@117.192.181.205] has quit [Ping timeout: 260 seconds] | 12:41 | |
-!- n4nd0 [~nando@s83-179-44-135.cust.tele2.se] has quit [Ping timeout: 245 seconds] | 12:43 | |
wiking | nyipp this clang_complete seems sweet | 12:48 |
blackburn | wiking: how to use it? | 12:51 |
blackburn | :D | 12:51 |
wiking | blackburn: do u use vim or any derivative of vim like gvim? | 12:51 |
blackburn | vim | 12:51 |
wiking | https://github.com/Rip-Rip/clang_complete | 12:51 |
wiking | use it | 12:51 |
blackburn | yeah already | 12:51 |
wiking | or of course here's the http://www.vim.org/scripts/script.php?script_id=3302 | 12:52 |
blackburn | how to generate that ctags file? :D | 12:52 |
wiking | quote: "You don't need any ctags for it to work! Only clang is needed." | 12:52 |
blackburn | wiking: okay so how to ask him for completion ? :D | 12:54 |
wiking | have u actually read this: http://www.vim.org/scripts/script.php?script_id=3302 | 12:54 |
wiking | After a . , -> and :: it is automatically trying to complete the code. If you want to "force" the completion, use <C-X><C-U> which is the mapping for user defined completion. | 12:55 |
blackburn | no :D | 12:55 |
blackburn | okay thanks | 12:55 |
blackburn | cool1 | 12:55 |
wiking | and here's some stuff about what to put into .clang_complete | 12:56 |
wiking | http://vim.wikia.com/wiki/Script:3302 | 12:56 |
wiking | if u are having problemoooos | 12:56 |
blackburn | wiking: hmm nicht problemos | 12:57 |
blackburn | works like a charm | 12:57 |
wiking | ;) | 12:57 |
blackburn | cool | 12:57 |
wiking | yeah it is | 12:57 |
blackburn | did you know about it? | 12:58 |
wiking | no | 12:58 |
wiking | first time i'veheard now there on | 12:58 |
wiking | pull request | 12:58 |
wiking | pluskid should share more funky shit :) | 12:58 |
blackburn | yeah I guess he knows something we do not | 12:59 |
wiking | :>>> | 12:59 |
blackburn | may be he even has script for shogun development | 12:59 |
blackburn | I admit it takes some time :D | 12:59 |
wiking | heheheheh | 12:59 |
wiking | yeah this way doing coding is much faster | 13:00 |
wiking | i mean for me a lot of times i had to do git grep functioncall | 13:00 |
wiking | so that i know what the fuck is the return value and the argu | 13:00 |
wiking | e | 13:00 |
blackburn | no I mean may be even some make-painful-sgvector-transition commands | 13:00 |
blackburn | :D | 13:00 |
blackburn | wiking: how does it work?? | 13:05 |
wiking | what? | 13:05 |
blackburn | wiking: how can it be so fast :D | 13:05 |
wiking | libclang | 13:06 |
blackburn | some C plugin for vim I had had for a while is much more slower | 13:06 |
blackburn | but it was a bunch of templates only | 13:06 |
wiking | nonono | 13:08 |
wiking | this is because of libclang | 13:08 |
wiking | i'm not 100% sure | 13:08 |
wiking | but afaik they do a binary based indexing as well | 13:08 |
wiking | the story is of course here that apple had a very shitty autocomplete mechanism for c/c++ | 13:09 |
wiking | in xcode | 13:09 |
wiking | and now they did it within clang | 13:09 |
wiking | since they dropped gcc because of gplv3 | 13:09 |
blackburn | wiking: does frame on top close after completion? | 13:10 |
blackburn | it stays here :( | 13:12 |
wiking | mmm | 13:12 |
wiking | weird | 13:12 |
-!- pluskid [~pluskid@li400-235.members.linode.com] has joined #shogun | 13:16 | |
-!- n4nd0 [~nando@n173-p203.kthopen.kth.se] has joined #shogun | 13:17 | |
-!- n4nd0 [~nando@n173-p203.kthopen.kth.se] has quit [Client Quit] | 13:17 | |
CIA-113 | shogun: Chiyuan Zhang master * r2e99d05 / src/shogun/machine/KernelMulticlassMachine.cpp : Use index_t instead of int32_t for index type. - http://git.io/3j6kVg | 13:24 |
CIA-113 | shogun: Soeren Sonnenburg master * rf91c456 / src/shogun/machine/KernelMulticlassMachine.cpp : | 13:24 |
CIA-113 | shogun: Merge pull request #514 from pluskid/kernelmc-store-model | 13:24 |
CIA-113 | shogun: - http://git.io/ygXRvg | 13:24 |
sonne|work | blackburn: I had a look at SGString/SparseVector | 13:28 |
sonne|work | IMHO the overhead of 20 bytes for each is too much... | 13:28 |
sonne|work | because we store stringlist/sparsematrix as list of sgstring/sparse vector | 13:29 |
sonne|work | for SGVector itself things are not soo bad | 13:29 |
sonne|work | because we really rarely needed this | 13:30 |
sonne|work | for very small vecs | 13:30 |
sonne|work | pluskid: ahh you around! | 13:30 |
sonne|work | pluskid: I wanted to ask one thing - in current sgvector we have this refcount ptr | 13:30 |
sonne|work | now all incref'ing is done automagically | 13:30 |
blackburn | sonne|work: then sparse matrix should be sparse in a different way! | 13:31 |
sonne|work | blackburn: ? | 13:31 |
blackburn | not as collection of sparse vectors | 13:31 |
pluskid | automagically :D, good word | 13:31 |
sonne|work | blackburn: why that? | 13:31 |
blackburn | sonne|work: but that will enable us to do refcounted matrix.. | 13:31 |
sonne|work | pluskid: and unrefing too except that maybe someone wants to *manually* unref | 13:32 |
sonne|work | blackburn: SGSparseMatrix will always work | 13:32 |
-!- vikram360 [~vikram360@117.192.181.205] has joined #shogun | 13:32 | |
blackburn | yes but not with that overhead | 13:32 |
sonne|work | pluskid: so in this case it may happen that someone calls x.unref() | 13:32 |
sonne|work | pluskid: more than once | 13:32 |
sonne|work | pluskid: so what I do is set all the ptrs to the data to NULL | 13:32 |
sonne|work | pluskid: so that works ... but now the problem is when someone wants to do x.ref(); x.ref(); multiple times | 13:33 |
sonne|work | then he cannot unref() it twice | 13:33 |
sonne|work | so I made ref() protected | 13:34 |
pluskid | sonne|work: you mean, on a single unref() call, you reset all ref-counters? | 13:34 |
pluskid | for example, ref-count is 5 currently, when user *manually* call unref, you set it to 0? | 13:34 |
sonne|work | pluskid: it is not needed anyway but somehow not my intuition | 13:34 |
sonne|work | pluskid: no | 13:34 |
sonne|work | pluskid: I set the ptr pointing to ref count to NULL | 13:35 |
pluskid | ah, then when will the data be free-ed? | 13:35 |
wiking | pluskid: awessoooome! again! | 13:35 |
pluskid | it looks strange | 13:35 |
pluskid | wiking: ?? | 13:35 |
sonne|work | pluskid: when the last object's refcount gets 0 | 13:35 |
wiking | pluskid: your pull req! i really needed that | 13:36 |
sonne|work | blackburn: well if we keep SGSparseVector as is there is no problem | 13:36 |
pluskid | sonne|work: oh, I see, you set the ref-count pointer of the current object to NULL | 13:36 |
sonne|work | wiking: waiting for your patch to introduce index_t in SG* :) | 13:36 |
blackburn | sonne|work: why not to change sgsparsematrix? | 13:36 |
sonne|work | pluskid: exactly | 13:36 |
wiking | sonne|work: you mean the replacement of the bad indexing variables? | 13:37 |
sonne|work | blackburn: I said not change sparse vector - not sparse matrix | 13:37 |
pluskid | sonne|work: do you decrease ref_count before reseting the pointer to NULL? | 13:37 |
sonne|work | wiking: int32_t -> index_t | 13:37 |
sonne|work | pluskid: of course | 13:37 |
blackburn | sonne|work: ?? | 13:37 |
wiking | sonne|work: yep yep that one | 13:37 |
sonne|work | pluskid: I just wanted your opinion on that | 13:37 |
pluskid | sonne|work: a bit strange, let me think for 2 min :D | 13:37 |
sonne|work | pluskid: it cannot leak memory - that is the good thing | 13:37 |
sonne|work | pluskid: only problem is that ref() can never be called explicitly | 13:38 |
sonne|work | blackburn: we have sparse *vectors* and sparse *matrices | 13:38 |
pluskid | sonne|work: if ref() is not call explicitly, why unref() will be called explicitly? | 13:38 |
pluskid | people should call ref() and unref() in pairs, or else it is a bug | 13:39 |
sonne|work | blackburn: I am saying exploding sparse vectors is not good because we use them in sparse matrix | 13:39 |
blackburn | sonne|work: obvious - why not to make sparse matrix independant from vector? | 13:39 |
sonne|work | pluskid: argh | 13:39 |
blackburn | there are other ways to store sparse matrices | 13:39 |
pluskid | sonne|work: ? | 13:40 |
sonne|work | pluskid: no one ever needs to call unref() if he can live with the object being automagically destroyed at some point | 13:40 |
sonne|work | pluskid: however sometimes you want to get rid of the old object | 13:40 |
sonne|work | *now* | 13:40 |
pluskid | sonne|work: then just re-assign it to an empty one | 13:40 |
pluskid | the operator= will call unref() | 13:41 |
sonne|work | pluskid: or call unref | 13:41 |
blackburn | why to unref anything? | 13:41 |
pluskid | sonne|work: calling unref() is strange here, the object still exist but becomes invalid | 13:41 |
pluskid | sg_vec = SGVector(length=0) will do, that's the standard way | 13:42 |
blackburn | second | 13:42 |
pluskid | and this will solve your headache of ref/unref | 13:42 |
sonne|work | pluskid: so you would make ref/unref private | 13:42 |
pluskid | sonne|work: yes, as private as possible | 13:43 |
sonne|work | and instead of writing sg_vec.unref() do sg_vec=SGVector<float64>() | 13:43 |
pluskid | unless you have to interactive with some legacy code or some other lang (SWIG? maybe I don't know) | 13:43 |
pluskid | sonney2k: yes | 13:43 |
pluskid | sonne|work: btw, STL vector has a clear() method :p | 13:43 |
sonne|work | pluskid: no swig will only ref/unref SGOBjects | 13:44 |
sonne|work | not data types | 13:44 |
sonne|work | pluskid: what does it do? | 13:44 |
pluskid | OK | 13:44 |
pluskid | sonne|work: it virtually do vec = std::vector() | 13:44 |
sonne|work | STL vector is not even guaranteed to be contiguous in memory | 13:44 |
pluskid | sonne|work: no, it is | 13:44 |
pluskid | vector should be continguous | 13:44 |
sonne|work | it is but not in the standard | 13:44 |
pluskid | really? | 13:45 |
pluskid | I don't think so | 13:45 |
sonne|work | at least when I considered it 10 years ago | 13:45 |
pluskid | vector iterator is random-access iterators | 13:45 |
sonne|work | >10 years ago | 13:45 |
* sonne|work is old | 13:45 | |
pluskid | for a random-access iterators to work, it should guarantee contiguous memory | 13:45 |
wiking | "From the C++03 standard (23.2.4.1): | 13:45 |
wiking | The elements of a vector are stored contiguously, meaning that if v is a vector where T is some type other than bool, then it obeys the identity &v[n] == &v[0] + n for all 0 <= n < v.size(). | 13:45 |
wiking | " | 13:45 |
pluskid | sonne|work: maybe time to re-consider STL and become young again :p | 13:46 |
sonne|work | never | 13:46 |
sonne|work | :D | 13:46 |
blackburn | sonne|work: just imagine you are 20 again | 13:46 |
blackburn | :D | 13:46 |
blackburn | you should accept our suggestion! | 13:46 |
sonne|work | hah I just learned: C++98 standard didn't completely guarantee contiguity, the C++03 standard was amended to explicitly add the guarantee. | 13:48 |
sonne|work | of contiguous memory | 13:48 |
pluskid | considering refactoring of CLabels, I start to prefer not doing it right now | 13:50 |
sonne|work | pluskid: what is the issue? | 13:51 |
pluskid | Because the current class works fine with regression, binary and mc | 13:51 |
pluskid | refactoring won't change much | 13:51 |
pluskid | I'm the current CLabels is just good enough, if we don't consider SO | 13:52 |
pluskid | I'm expecting huge changes to our Label system when more SO learning is involved | 13:52 |
blackburn | my suggestion is | 13:53 |
blackburn | to develop SOLabels standalone for now | 13:53 |
blackburn | and machine too | 13:53 |
blackburn | we can always merge that kind of baselines | 13:53 |
blackburn | and this would enable n4ndo to work on that | 13:53 |
blackburn | not requiring waiting for the changes | 13:54 |
pluskid | I mean, if we ignore SO, CLabels work fine, so it is not urge to refactor CLabels; if we consider SO, then they will also need to refactor CLabels, our refactoring needs to be compatible with theirs, and since ours are rather small, we'd better follow the SO's side: what kind of class hierarchy they need, etc. | 13:55 |
sonne|work | pluskid: I don't see what SO and regression/binary/mc labels could change except get_num_labels() and is_valid() | 13:56 |
blackburn | I still think template labels could be cool there | 13:57 |
pluskid | sonne|work: float64_t get_label(int32_t) will be meaningless in SOLabels, they should return a more complex object I think | 13:57 |
sonne|work | pluskid: that is why this function is not in the shared baseclass | 13:58 |
sonne|work | blackburn: I don't | 13:58 |
blackburn | sonne|work: why? | 13:58 |
sonne|work | blackburn: what do you think you gain by templates? | 13:58 |
pluskid | sonne|work: so the shared baseclass becomes almost empty, only have functions like get_num_labels() | 13:59 |
blackburn | sonne|work: generality | 13:59 |
sonne|work | pluskid: yes | 13:59 |
blackburn | sonne|work: even SO can be done with template | 13:59 |
sonne|work | but then the template needs to be of a certain class type | 14:00 |
sonne|work | not int / double etc | 14:00 |
pluskid | sonne|work: that's OK, only a little painful because have to do cast almost everywhere | 14:00 |
blackburn | sonne|work: why? | 14:00 |
sonne|work | horrible | 14:00 |
sonne|work | blackburn: how do you implement 'is_valid' ? | 14:00 |
blackburn | sonne|work: specialize | 14:00 |
sonne|work | how | 14:01 |
blackburn | I am sorry i can't see problem there - what is ? | 14:01 |
sonne|work | blackburn: give a more concrete example how you think it works | 14:01 |
blackburn | Labels<int32_t> | 14:02 |
blackburn | is multiclass labels | 14:02 |
pluskid | sonne|work: if we use template, then we'll not use a shared base class for various kind of labels | 14:02 |
blackburn | the only problem is that machine contains pointer to labels | 14:02 |
blackburn | we can mimic boost::any here somehow | 14:02 |
blackburn | pluskid: do you agree? | 14:02 |
pluskid | blackburn: I'd rather prefer we use only MCLabels in MCClass, SOLabels in SOClass, no base class | 14:03 |
blackburn | no base class at all? | 14:03 |
pluskid | blackburn: yes | 14:03 |
sonne|work | pluskid: we need a base class | 14:03 |
blackburn | sonne|work: why do we? | 14:03 |
sonne|work | for machines returning labels | 14:03 |
pluskid | sonne|work: only because we have a generic CMachine, that is almost empty | 14:03 |
sonne|work | apply() | 14:03 |
blackburn | sonne|work: right | 14:03 |
sonne|work | yes | 14:04 |
pluskid | sonne|work: we can split | 14:04 |
pluskid | sonne|work: MCMachine.apply() returns MCLabels | 14:04 |
pluskid | SOMachine.apply() returns SOLabels | 14:04 |
blackburn | pluskid: we will lost CMachine interface then | 14:04 |
pluskid | etc | 14:04 |
sonne|work | yeah - I prefer having the interface | 14:04 |
sonne|work | it is like apply(CFeatures*) | 14:04 |
pluskid | blackburn: why do we need a generic apply()? | 14:05 |
sonne|work | and train(CFeatures*) | 14:05 |
sonne|work | pluskid: it is the interface of all machines | 14:05 |
blackburn | pluskid: to make all machines similar | 14:05 |
blackburn | in some kind | 14:05 |
pluskid | sonne|work: but they are actually not similar, MC machine apply to MClabels, SO machine applys to MCClass | 14:05 |
blackburn | having general interface seems to be cool anyway | 14:06 |
sonne|work | pluskid: they are - they all work on features and labels | 14:06 |
blackburn | all features are features | 14:06 |
blackburn | and all labels are labels | 14:06 |
blackburn | concept stays | 14:06 |
pluskid | CMachine can become empty, just like the base class CLabels can become empty | 14:06 |
blackburn | pluskid: yes interface fate of that classes :D | 14:07 |
pluskid | blackburn: one option is to make CMachine empty; another option is to make CLabels empty | 14:07 |
blackburn | why empty? | 14:07 |
pluskid | almost empty | 14:08 |
sonne|work | pluskid: you have the same problem with kernels | 14:08 |
blackburn | apply is here | 14:08 |
sonne|work | and kernel machines | 14:08 |
sonne|work | they have an apply() method | 14:08 |
sonne|work | and can work with *any* kind of features | 14:08 |
sonne|work | that fit the kernel | 14:08 |
pluskid | hmm, seems we have discussed about this before | 14:09 |
pluskid | anyway, I'm OK with either solution | 14:09 |
pluskid | brb | 14:10 |
blackburn | sonne|work: clang seems to be *very* cool | 14:11 |
blackburn | complete* | 14:11 |
blackburn | however there is a bug I'll try to resolve | 14:11 |
blackburn | ah not a bug option | 14:12 |
-!- blackburn [~qdrgsm@83.234.54.222] has left #shogun [] | 14:12 | |
-!- blackburn [~qdrgsm@83.234.54.222] has joined #shogun | 14:12 | |
blackburn | oops | 14:12 |
blackburn | wiking: g:clang_complete_auto = 1 | 14:13 |
blackburn | nooooo | 14:13 |
blackburn | g:clang_close_preview = 1 | 14:13 |
blackburn | I want to make preview display doc | 14:14 |
blackburn | pluskid: we want you to share more funky stuff :) | 14:21 |
pluskid | blackburn: pathogen neocomplcache? | 14:29 |
blackburn | pluskid: sounds like an illness name | 14:30 |
blackburn | :D | 14:30 |
pluskid | pathogen is an elegent way of managing vim plugins | 14:30 |
pluskid | neocomplcache is a plugin that's cool I think | 14:30 |
pluskid | no perfect plugins, but many are cool, clang_complete is one with promising future | 14:31 |
blackburn | pluskid: do you use clang complete + neocomplcache? | 14:31 |
blackburn | I am really excited with clang complete | 14:31 |
pluskid | yes I use the two | 14:31 |
blackburn | okay I'll try it now | 14:32 |
-!- eric_ [2e1fd566@gateway/web/freenode/ip.46.31.213.102] has joined #shogun | 14:32 | |
eric_ | hi all | 14:32 |
wiking | eric_: try multi class now | 14:32 |
wiking | i mean multi class xval | 14:33 |
eric_ | wiking: I saw it in the logs, thx :p | 14:33 |
eric_ | wiking: one question about it: I get one error : "MulticlassLibSVM does not support locking" | 14:34 |
wiking | eric_: mmm have u checked the example? | 14:34 |
wiking | evaluation_cross_validation_multiclass.cpp | 14:34 |
eric_ | wiking: what's locking ? Is it possible to run multiclass Xval with MulticlassLibSVM or MKLmulticlass ? | 14:35 |
blackburn | pluskid: did you configure neocomplcache? | 14:35 |
eric_ | wiking: sure | 14:35 |
pluskid | blackburn: yes, a bit | 14:35 |
wiking | eric_: u should be able to run multiclasslibsvm | 14:35 |
blackburn | pluskid: can you share a config? | 14:35 |
wiking | in xval | 14:35 |
pluskid | blackburn: a min | 14:35 |
gsomix | sonne|work, CSet is on github now. | 14:35 |
wiking | just replace basically CMulticlassLibLinear -> CMulticlassLibSVM | 14:36 |
wiking | a | 14:36 |
wiking | n | 14:36 |
wiking | d | 14:36 |
wiking | use | 14:36 |
wiking | 14:36 | |
wiking | kernel of course | 14:36 |
wiking | fucking hell what's with this irc client | 14:36 |
wiking | i've tested yesterday and workedforme | 14:36 |
pluskid | blackburn: https://gist.github.com/2644215 | 14:36 |
CIA-113 | shogun: Soeren Sonnenburg master * r98c94f9 / (6 files in 2 dirs): | 14:36 |
CIA-113 | shogun: Merge pull request #515 from gsomix/CSet | 14:36 |
CIA-113 | shogun: (+5 more commits...) - http://git.io/t4nN6A | 14:36 |
eric_ | wiking: that's exaclty what I have done, it returns "MulticlassLibSVM does not support locking. Autolocking is skipped. Set autolock flag to false to get rid of warning." | 14:37 |
blackburn | pluskid: cool! thanks | 14:37 |
gsomix | sonne|work, wow, so fast. tnx. | 14:37 |
pluskid | blackburn: ur wel~ :) hope you find interesting thing | 14:38 |
gsomix | unfortunately I have to study the methods for solving the eikonal equation | 14:38 |
blackburn | pluskid: yes - is git status a some plugin? | 14:38 |
gsomix | cu later | 14:38 |
wiking | eric_: ok, then add this: cross->set_autolock (false); to line 123 in evaluation_cross_validation_multiclass.cpp | 14:38 |
wiking | i'm trying here as well | 14:38 |
pluskid | blackburn: let me find | 14:38 |
wiking | but i still need to recompile shogun | 14:39 |
pluskid | blackburn: vim-powerline | 14:39 |
pluskid | blackburn: actually, you might find many useful plugin here: https://github.com/spf13/spf13-vim | 14:39 |
pluskid | blackburn: vim-fugitive is also needed for the git status line to work I think | 14:39 |
eric_ | wiking: warning disappear but got seg fault | 14:40 |
wiking | pluskid: i've just used janus till now | 14:40 |
wiking | eric_: i'll try it myself as well | 14:40 |
blackburn | wow cool thanks, pluskid | 14:40 |
wiking | still waiting for the compilation | 14:40 |
blackburn | wiking: do you use ccache? | 14:41 |
wiking | blackburn: y | 14:41 |
blackburn | ok :) | 14:42 |
pluskid | wiking: I use spf13 until this morning. Crazy by various mysterious behaviors and not being able to find where they are configured, I re-write my own vimrc today | 14:42 |
pluskid | that's what blackburn saw just now :D | 14:42 |
pluskid | btw: clang+ccache is cool | 14:42 |
wiking | wooohooo | 14:43 |
wiking | Program received signal EXC_ARITHMETIC, Arithmetic exception. | 14:44 |
wiking | 0x000000010023294d in shogun::CMap<int, int>::hash (this=0x7fff5fbfdfa0, key=<value temporarily unavailable, due to optimizations>) at Map.h:249 | 14:44 |
wiking | eric_: sorry i have tested it with the version where still std::map was used ;) | 14:44 |
blackburn | pluskid: did you patch your fonts for powerline? | 14:45 |
pluskid | blackburn: no, it's fine | 14:45 |
blackburn | hmm droid? | 14:45 |
pluskid | yeah | 14:45 |
blackburn | I have troubles with liberation fonts, will try droid | 14:46 |
pluskid | Droid sans mono | 14:46 |
blackburn | hehe looks like a crap still | 14:46 |
pluskid | overlapping? | 14:47 |
blackburn | pluskid: no colors, some crazy chars | 14:47 |
pluskid | hmm... no idea, just works here | 14:47 |
blackburn | I am using guake here | 14:48 |
pluskid | KDE fan here | 14:48 |
pluskid | brb | 14:49 |
wiking | ooooh loool stupid IDE handled % hash_size as comment | 14:52 |
wiking | wtf?! | 14:52 |
eric_ | wiking: I got "Floating point exception" error | 14:52 |
wiking | eric_: no worries i think i know what's the problem | 14:52 |
wiking | there's a modulo 0 | 14:52 |
wiking | gsomix: here? | 14:53 |
gsomix | wiking, yep. | 14:54 |
wiking | gsomix: u wrote Map.h right? | 14:54 |
gsomix | yep =___= what's wrong? | 14:54 |
wiking | gsomix: return CHash::MurmurHash2((uint8_t*)(&key), sizeof(key), 0xDEADBEEF) % hash_size; | 14:54 |
wiking | Map.h:249 | 14:54 |
wiking | it can actually happen that hash_size = 0 | 14:55 |
wiking | :S | 14:55 |
gsomix | wiking, show me code. | 14:55 |
wiking | check line 249 in Map.h | 14:55 |
wiking | and then if the map is constructed like this: CMap<int32_t, int32_t> all_sv; | 14:56 |
wiking | then basically hash_size = 0 | 14:56 |
wiking | => arithmetic error | 14:56 |
gsomix | ah, I got it. | 14:56 |
wiking | BOOOM | 14:56 |
gsomix | thanks | 14:56 |
wiking | everything booooms | 14:56 |
wiking | so i guess | 14:56 |
wiking | in void add(const K& key, const T& data) | 14:56 |
wiking | you should do something with hash_size or something :) | 14:57 |
gsomix | wiking, I'll fix it. please, use custom constructor | 14:59 |
gsomix | at now | 14:59 |
wiking | ^^ pluskid | 14:59 |
pluskid | oops, forgot to recommend ctrlp and a.vim -- the two best vim plugin in my knowledge :D | 15:04 |
blackburn | pluskid: I use A | 15:06 |
blackburn | pluskid: I have problem with clang complete and necompl | 15:06 |
blackburn | neocompl | 15:06 |
pluskid | blackburn: what's that? | 15:06 |
blackburn | pluskid: it fails because they both define completeFunc function | 15:07 |
pluskid | hmm? I didn't get any error | 15:07 |
gsomix | wiking, fixed. | 15:08 |
gsomix | I think I need to introduce rehash method for increase hash_size. | 15:09 |
blackburn | pluskid: hmm interesting - it fails only after :A | 15:09 |
pluskid | blackburn: ah, I see, also failing here | 15:10 |
-!- n4nd0 [~nando@n173-p203.kthopen.kth.se] has joined #shogun | 15:14 | |
CIA-113 | shogun: Evgeniy Andreev master * rbb6c2a7 / (src/shogun/lib/Map.h src/shogun/lib/Set.h): fixes in default constructors - http://git.io/gVXHyA | 15:20 |
CIA-113 | shogun: Evgeniy Andreev master * rf416fe9 / (src/shogun/lib/Map.h src/shogun/lib/Set.h): another fixes in default constructors - http://git.io/sSDK0g | 15:20 |
CIA-113 | shogun: Soeren Sonnenburg master * r191a0b1 / (src/shogun/lib/Map.h src/shogun/lib/Set.h): | 15:20 |
CIA-113 | shogun: Merge pull request #516 from gsomix/CSet | 15:20 |
CIA-113 | shogun: - http://git.io/HLuaeQ | 15:20 |
wiking | gsomix: testing | 15:23 |
-!- nicococo [~nico@lacedcoffee.ml.tu-berlin.de] has joined #shogun | 15:24 | |
blackburn | http://img151.imageshack.us/img151/3855/vlcsnap2012050916h57m10.jpg | 15:25 |
wiking | eric_: should work now | 15:28 |
wiking | even without calling set_autolock(false) | 15:29 |
eric_ | wiking: I pull, test, and come back. | 15:33 |
blackburn | pluskid: how to get these colors like in powerline screenshots? | 15:34 |
blackburn | I have no colors here at all | 15:34 |
blackburn | do you have? | 15:34 |
eric_ | ahah improving your vimrc config ? | 15:36 |
eric_ | about that: does any1 get autocompletion with boost lib ? | 15:37 |
eric_ | I never succeed in having completion working with "template" library | 15:38 |
blackburn | eric_: yeah we solved this issue with using no boost :D | 15:38 |
pluskid | blackburn: http://imageshack.us/photo/my-images/194/gvim.png | 15:39 |
pluskid | I changed nothing | 15:39 |
blackburn | pluskid: I see - some bug here :( | 15:40 |
blackburn | in my config may be | 15:40 |
eric_ | do you have a common config ? | 15:40 |
eric_ | wiking: mmhh I got a seg fault. I will check my code again. | 15:44 |
wiking | eric_: i've just changed that example code to use gaussian with multiclasslibsvm | 15:44 |
wiking | and it ran w/o any errors | 15:45 |
gsomix | wiking, works? | 15:45 |
wiking | gsomix: yep like a charm thnx | 15:45 |
* gsomix breathed out | 15:46 | |
-!- wiking [~wiking@huwico/staff/wiking] has quit [Quit: wiking] | 15:47 | |
-!- wiking [~wiking@we02c096.ugent.be] has joined #shogun | 15:48 | |
-!- wiking [~wiking@we02c096.ugent.be] has quit [Changing host] | 15:48 | |
-!- wiking [~wiking@huwico/staff/wiking] has joined #shogun | 15:48 | |
-!- cronor [~cronor@fb.ml.tu-berlin.de] has quit [Ping timeout: 250 seconds] | 15:49 | |
eric_ | wiking: still got seg fault | 15:52 |
eric_ | wiking: [0.096233,0.099767] with alpha=0.400000, mean=0.098000 Segmentation fault | 15:52 |
eric_ | wiking: ok solved. my fault. | 15:56 |
blackburn | pluskid: export TERM="xterm-256color" solved it! | 15:57 |
blackburn | yay | 15:57 |
pluskid | cool | 16:00 |
eric_ | wiking: SG_UNREF(cross) is the issue | 16:00 |
eric_ | wiking: because it control unref of labels etc .. | 16:02 |
wiking | eric_: great, now you can xval :D | 16:12 |
wiking | have fun | 16:12 |
eric_ | wiking: thx again! | 16:13 |
wiking | eric_: thanx to pluskid | 16:13 |
eric_ | pluskid: +1 | 16:13 |
blackburn | time to shogun! | 16:13 |
blackburn | vim is moar awesome now | 16:14 |
blackburn | :D | 16:14 |
wiking | gunsho | 16:14 |
pluskid | u r wel! though I'm not sure what's happening here :p | 16:14 |
pluskid | blackburn: solved clang_complete vs. neocomplcache conflict? | 16:14 |
blackburn | pluskid: no just dropped neocomplcache | 16:14 |
blackburn | for now | 16:14 |
pluskid | blackburn: considering that, too | 16:14 |
wiking | gunshow: say it fast repeatedly | 16:15 |
pluskid | though it's very nice for completing header file path etc. | 16:15 |
wiking | blackburn: SAY IT | 16:15 |
blackburn | gunshow? | 16:15 |
wiking | yeah fast | 16:15 |
wiking | and repeatedly | 16:15 |
blackburn | soo? | 16:16 |
wiking | shogun | 16:16 |
wiking | :> | 16:16 |
wiking | if u know what i mean? :) | 16:16 |
blackburn | ah :D | 16:16 |
pluskid | blackburn: btw, maybe this could help: https://github.com/osyo-manga/neocomplcache-clang_complete | 16:16 |
blackburn | I start to think most of github projects are vim plugins :D | 16:17 |
blackburn | pluskid: pathogen is simply awesome | 16:18 |
blackburn | I am curious whether sonne|work would accept it :D | 16:18 |
sonne|work | what? | 16:18 |
sonne|work | vim plugin sure | 16:19 |
pluskid | blackburn: it seems someone has mirrored *every* script in the official vim-script site to github | 16:19 |
blackburn | sonne|work: try pathogen - installs plugins automagically | 16:20 |
blackburn | and clang complete + neocomplcache is awesome too | 16:20 |
blackburn | pluskid: aren't these guys authors? | 16:20 |
pluskid | blackburn: https://github.com/Rip-Rip/clang_complete/issues/60 | 16:20 |
pluskid | blackburn: no, see this user: https://github.com/vim-scripts | 16:21 |
pluskid | he has 3,972 github repos | 16:21 |
pluskid | all are vim-scripts | 16:21 |
blackburn | ah | 16:22 |
blackburn | pluskid: that's nice nevertheless | 16:22 |
gsomix | vimers, vimers everywhere | 16:26 |
pluskid | I was a Emacser | 16:28 |
sonne|work | pluskid: poor guy | 16:28 |
sonne|work | but it is good that you converted from the dark side :-) | 16:29 |
pluskid | I love emacs-lisp, I wrote a Emacs plugin that is quite popular (http://code.google.com/p/yasnippet/). Though I use vim nowadays, I never managed to understand the crappy vimscript syntax... | 16:29 |
pluskid | blackburn: it seems installing this plugin can solve the conflict: https://github.com/osyo-manga/neocomplcache-clang_complete | 16:32 |
pluskid | I haven't tested much, but at least the errors seems to be gone | 16:32 |
blackburn | pluskid: yes errors are gon | 16:34 |
blackburn | e | 16:34 |
gsomix | http://farm2.static.flickr.com/1070/4723428183_f9f1aa7f90.jpg | 16:35 |
blackburn | pluskid: what system do you use btw? | 16:41 |
pluskid | blackburn: Arch Linux | 16:41 |
blackburn | heh cool we cover a lot of distribs in our dev team | 16:42 |
pluskid | what do we have? | 16:42 |
blackburn | I use ubuntu (yet?), sonne|work uses debian, my mentor chris uses opensuse | 16:42 |
blackburn | wiking uses macos | 16:43 |
pluskid | that's a good coverage | 16:43 |
blackburn | yes, no windows but | 16:44 |
blackburn | who needs windows :D | 16:44 |
sonne|work | but shogun doesn't run under windows anywya | 16:44 |
pluskid | cygwin? | 16:44 |
wiking | :D | 16:45 |
blackburn | yes only cygwin | 16:45 |
-!- n4nd0 [~nando@n173-p203.kthopen.kth.se] has quit [Read error: Connection reset by peer] | 16:47 | |
wiking | i recall somebody struggling with shogun under cygwin | 16:47 |
wiking | but then he never came back | 16:47 |
wiking | :) | 16:47 |
blackburn | yes it can happen if you try to compile shogun under cygwin | 16:48 |
* wiking wonders whether that is a sign for success or failure | 16:48 | |
-!- n4nd0 [~nando@n173-p203.kthopen.kth.se] has joined #shogun | 16:48 | |
-!- wiking [~wiking@huwico/staff/wiking] has quit [Quit: wiking] | 16:56 | |
sonne|work | blackburn: he submitted many patches | 16:59 |
sonne|work | and fixed many bugs tough | 16:59 |
sonne|work | though | 16:59 |
blackburn | sonne|work: about guy with cygwin? | 16:59 |
eric_ | sonne|work: any advice ? MulticlassLibSVM::load_serializable(): File contains no parameter version. Seems like your file is from the days before this was introduced. | 17:05 |
eric_ | sonne|work: what type of file should I choose to save/load svm machine trained | 17:06 |
eric_ | I use currently CSerializableAsciiFile | 17:06 |
blackburn | eric_: did you save it with exactly that version? | 17:07 |
blackburn | pluskid: cool! | 17:07 |
eric_ | yes I rerun the trainind step | 17:07 |
blackburn | it is a bug then | 17:08 |
pluskid | nite boys | 17:13 |
blackburn | sonne|work: I am ok with pluskid's patch - sounds very reasonable for me | 17:13 |
blackburn | pluskid: good night | 17:13 |
pluskid | blackburn: maybe he is sonney2k now (on the train), :D | 17:14 |
-!- pluskid [~pluskid@li400-235.members.linode.com] has quit [Quit: Leaving] | 17:14 | |
blackburn | yes should be sonne|train | 17:14 |
n4nd0 | haha | 17:15 |
-!- wiking [~wiking@78-23-189-112.access.telenet.be] has joined #shogun | 17:15 | |
-!- wiking [~wiking@78-23-189-112.access.telenet.be] has quit [Changing host] | 17:15 | |
-!- wiking [~wiking@huwico/staff/wiking] has joined #shogun | 17:15 | |
blackburn | n4nd0: check last pull request | 17:23 |
blackburn | I believe it can be the acceptable solution | 17:23 |
n4nd0 | oh labels, interesting | 17:24 |
eric_ | wiking: remarq = in your Xvalid c++ example, when you apply the trained svm on the same features, the training error your print is actually the accuracy. | 17:29 |
-!- eric_ [2e1fd566@gateway/web/freenode/ip.46.31.213.102] has quit [Quit: Page closed] | 17:32 | |
n4nd0 | blackburn: I am not sure if I understand the functionality of LabelsImpl | 17:35 |
blackburn | n4nd0: just storing/returning/etc | 17:35 |
blackburn | n4nd0: generalize all the stuff | 17:35 |
n4nd0 | ok | 17:37 |
n4nd0 | what can be used as a CLabel but not as a CLabelImpl? | 17:38 |
blackburn | n4nd0: hmm machines have labels | 17:40 |
blackburn | not labelimpl because it is *any* templated | 17:40 |
n4nd0 | ahm I see | 17:42 |
n4nd0 | it looks nice to me | 17:42 |
n4nd0 | I have to go now | 17:43 |
n4nd0 | see you later! | 17:43 |
blackburn | ok | 17:43 |
blackburn | see you | 17:43 |
-!- n4nd0 [~nando@n173-p203.kthopen.kth.se] has quit [Quit: leaving] | 17:43 | |
-!- gsomix [~gsomix@188.168.4.56] has quit [Quit: Ex-Chat] | 17:45 | |
CIA-113 | shogun: Sergey Lisitsyn master * r608f708 / src/shogun/regression/KernelRidgeRegression.cpp : Fixed KRR crasher - http://git.io/AtDzcw | 17:46 |
-!- uricamic [~uricamic@2001:718:2:1634:f2de:f1ff:fecf:a6a5] has quit [Quit: Leaving.] | 18:49 | |
-!- vikram360 [~vikram360@117.192.181.205] has quit [Ping timeout: 240 seconds] | 19:46 | |
-!- Marty28 [~marty@cable-158-181-77-81.cust.telecolumbus.net] has joined #shogun | 19:52 | |
-!- puffin444 [62e3926e@gateway/web/freenode/ip.98.227.146.110] has joined #shogun | 19:57 | |
-!- Marty28 [~marty@cable-158-181-77-81.cust.telecolumbus.net] has quit [Quit: Colloquy for iPad - http://colloquy.mobi] | 20:01 | |
blackburn | sonney2k: you should try these cool plugins pluskid referenced! | 20:24 |
-!- n4nd0 [~nando@s83-179-44-135.cust.tele2.se] has joined #shogun | 20:28 | |
puffin444 | What are these plugins? | 20:30 |
blackburn | puffin444: ctags_complete | 20:31 |
blackburn | fugitive | 20:32 |
blackburn | powerline | 20:32 |
CIA-113 | shogun: Soeren Sonnenburg master * r49462e7 / (15 files in 9 dirs): make unref() protected - http://git.io/2NUNCQ | 20:46 |
blackburn | sonney2k: fixing plifs (fyi) | 20:48 |
CIA-113 | shogun: Sergey Lisitsyn master * rad0868b / (3 files): Fixed Plif and PlifMatrix SGVectors handling routines - http://git.io/z5igmQ | 20:52 |
-!- blackburn [~qdrgsm@83.234.54.222] has quit [Quit: Leaving.] | 20:53 | |
-!- PhilTillet [~philippe@157.159.42.73] has joined #shogun | 20:55 | |
-!- blackburn [~qdrgsm@83.234.54.222] has joined #shogun | 20:56 | |
-!- ckwidmer [~chris@HSI-KBW-046-005-237-106.hsi8.kabel-badenwuerttemberg.de] has joined #shogun | 21:01 | |
-!- PhilTillet [~philippe@157.159.42.73] has quit [Read error: Operation timed out] | 21:05 | |
-!- gsomix [~gsomix@188.168.4.80] has joined #shogun | 21:10 | |
gsomix | crazy day | 21:10 |
gsomix | too many people in Samara | 21:10 |
blackburn | yeah fire and gas can help there I think | 21:11 |
gsomix | blackburn, hehe | 21:20 |
* gsomix have returned from dance floor | 21:20 | |
@sonney2k | blackburn, I have some general ML question | 21:23 |
blackburn | for me? | 21:24 |
@sonney2k | do you have any idea how one can compare two distributions | 21:24 |
@sonney2k | I have some very simple ratios | 21:24 |
@sonney2k | x,y x+y=1 | 21:24 |
@sonney2k | and I know what values x & y should have | 21:25 |
@sonney2k | but at some point they start to deviate | 21:25 |
@sonney2k | how can I measure deviation | 21:25 |
@sonney2k | blackburn, yes | 21:25 |
-!- vikram360 [~vikram360@117.192.181.205] has joined #shogun | 21:25 | |
blackburn | hmm? | 21:25 |
@sonney2k | I thought about sth like relative entropy | 21:25 |
@sonney2k | aka kullback leiber divergence | 21:26 |
@sonney2k | KL div | 21:26 |
blackburn | yeah know that | 21:26 |
@sonney2k | not so sure if that is the best chocie thoug | 21:26 |
ckwidmer | sonney2k, what do you think the downside is? | 21:26 |
@sonney2k | ckwidmer, no idea | 21:27 |
@sonney2k | I haven't tried it yet | 21:27 |
blackburn | well KL is a kind of standard metric between distributions afaik? | 21:28 |
ckwidmer | I used it once in a very simple setting, worked ok for what I was doing (checking how different distributions for different tasks are) | 21:28 |
blackburn | sonney2k: do you want ->R function? or to find out whether they deviate? | 21:29 |
@sonney2k | ckwidmer, ok - that sounds pretty similar to my problem | 21:29 |
blackburn | in the second case kolmogorov test would work | 21:29 |
@sonney2k | blackburn, yeah some score telling how much they deviate | 21:29 |
@sonney2k | so not just 'deviates' | 21:30 |
ckwidmer | some people here in tuebingen use MMD (maximum mean discrepancy) for two sample problems | 21:30 |
ckwidmer | http://people.kyb.tuebingen.mpg.de/arthur/mmd.htm | 21:30 |
blackburn | haha | 21:30 |
blackburn | sonney2k: yeah heiko should be expert there | 21:30 |
@sonney2k | heiko is not here :( | 21:30 |
blackburn | ckwidmer: we have gsoc project about that ;) | 21:30 |
ckwidmer | :) | 21:31 |
@sonney2k | heiko is implementing arthurs code in shogun... | 21:31 |
ckwidmer | I see | 21:31 |
ckwidmer | anyway | 21:31 |
ckwidmer | one thing to consider with KL | 21:31 |
blackburn | ckwidmer: I am interested in that too btw | 21:31 |
ckwidmer | is that it's not symmetric | 21:31 |
@sonney2k | ckwidmer, I know | 21:31 |
@sonney2k | but that's ok | 21:31 |
ckwidmer | ok | 21:31 |
blackburn | sonney2k: JS btw | 21:32 |
blackburn | it is symmetric | 21:32 |
blackburn | I do think you can even use JS kernel from shogun to measure it | 21:32 |
ckwidmer | you could also do something like KL(X,Y) + KL(Y,X) | 21:33 |
blackburn | 1/2 | 21:33 |
ckwidmer | optional :) | 21:33 |
blackburn | ckwidmer: I said that already :) | 21:33 |
blackburn | ckwidmer: JS it is | 21:33 |
ckwidmer | guess sometimes, you have to spell it out for clarity | 21:34 |
@sonney2k | funny thing is that this is very close to JS | 21:34 |
blackburn | ckwidmer: :) jensen shannon | 21:34 |
blackburn | if I understand you right | 21:35 |
@sonney2k | ckwidmer, what you wrote | 21:35 |
@sonney2k | JS is KL(X, X+Y) + KL(Y, X+Y) | 21:35 |
blackburn | ??? | 21:35 |
@sonney2k | and a couple of 0.5's | 21:35 |
ckwidmer | yep | 21:35 |
@sonney2k | blackburn, https://en.wikipedia.org/wiki/Jensen%E2%80%93Shannon_divergence | 21:35 |
blackburn | oh okay sorry :D | 21:35 |
gsomix | good night guys | 21:36 |
@sonney2k | gsomix, let me guess your android phone? | 21:36 |
@sonney2k | gsomix, any progress on the array stuff btw? | 21:36 |
gsomix | sonney2k, aha, phone gone crazy :) | 21:37 |
gsomix | sonney2k, a little, just added methods in DynamicArray. now I need to replace all Array# in the code by DynamicArray. | 21:38 |
@sonney2k | gsomix, very good - so thanks for the update and have a nice sleep! | 21:39 |
-!- PhilTillet [~philippe@157.159.47.10] has joined #shogun | 21:39 | |
gsomix | sonney2k, May celebrations have ended -> more time to shogun | 21:39 |
gsomix | nite | 21:39 |
@sonney2k | blackburn where are we SGMatrix transition wise | 21:40 |
@sonney2k | are examples back to normal? | 21:40 |
blackburn | sonney2k: in big black ass :D | 21:40 |
@sonney2k | or do we still have a few crashers? | 21:40 |
blackburn | only a few | 21:40 |
@sonney2k | valgrind errors... | 21:40 |
@sonney2k | blackburn, ahh btw we didn't end the discussion about sparse matrix / sparse vector | 21:41 |
@sonney2k | what I was saying is that the current SGSparseVector should either be kept or renamed to sth else | 21:41 |
@sonney2k | if we want to save 20 bytes / vector | 21:41 |
@sonney2k | same with SGString btw | 21:41 |
@sonney2k | if we decide not to save these 20 bytes | 21:42 |
@sonney2k | we can do crazy things like replacing / fetching vectors | 21:42 |
@sonney2k | and sharing them | 21:42 |
blackburn | sonney2k: do you know boost matrix | 21:52 |
blackburn | has | 21:53 |
blackburn | m(i,j) = x | 21:53 |
blackburn | ? | 21:53 |
blackburn | ehm.. we have too?? | 21:55 |
-!- cronor [~cronor@e178181022.adsl.alicedsl.de] has joined #shogun | 22:06 | |
-!- ckwidmer [~chris@HSI-KBW-046-005-237-106.hsi8.kabel-badenwuerttemberg.de] has quit [Remote host closed the connection] | 22:14 | |
blackburn | sonney2k: about sgsparsematrix? | 22:15 |
-!- puffin444 [62e3926e@gateway/web/freenode/ip.98.227.146.110] has left #shogun [] | 22:27 | |
blackburn | fugitive should be illegal | 22:28 |
blackburn | :D | 22:28 |
blackburn | cool | 22:29 |
CIA-113 | shogun: Sergey Lisitsyn master * r34da81e / src/shogun/converter/MultidimensionalScaling.cpp : Used () matrix element access in the MDS - http://git.io/-CYQ9A | 22:29 |
-!- gsomix [~gsomix@188.168.4.80] has quit [Read error: Operation timed out] | 22:31 | |
blackburn | time to merge some code | 22:32 |
shogun-buildbot | build #904 of libshogun is complete: Failure [failed compile] Build details are at http://www.shogun-toolbox.org/buildbot/builders/libshogun/builds/904 blamelist: blackburn91@gmail.com | 22:33 |
blackburn | really? | 22:33 |
-!- Netsplit *.net <-> *.split quits: PhilTillet | 22:34 | |
CIA-113 | shogun: Sergey Lisitsyn master * rc1bcbbb / src/shogun/converter/MultidimensionalScaling.cpp : Fixed bug in last commit - http://git.io/JY1KHA | 22:34 |
blackburn | shogun-buildbot: forgive me I was wrong it will never happen again | 22:38 |
blackburn | please | 22:38 |
wiking | hehehe it's cool actually | 22:38 |
blackburn | wiking: what is? | 22:39 |
wiking | we can make buildbot say anything | 22:39 |
wiking | :))))) | 22:39 |
blackburn | :D | 22:39 |
blackburn | wiking: last year we had one bot here | 22:39 |
* wiking already knows what's his next commit log will look like | 22:39 | |
blackburn | he was answering with phrases he catched there | 22:39 |
blackburn | it was funny actually | 22:39 |
shogun-buildbot | build #905 of libshogun is complete: Success [build successful] Build details are at http://www.shogun-toolbox.org/buildbot/builders/libshogun/builds/905 | 22:40 |
blackburn | @sonney2kbettyboo, are you always smiling?10:58 | 22:41 |
blackburn | @bettyboosonney2k: but you are interested in shogun?10:58 | 22:41 |
blackburn | @sonney2kbettyboo, yes of course - now I understand the reason why you are smiling.10:59 | 22:41 |
blackburn | @bettyboosonney2k: input set of strings, output n-dimensional representation | 22:41 |
blackburn | wiking: ^ | 22:41 |
wiking | :>>> | 22:41 |
blackburn | wiking: I'll make a 10k LoC commit in ten mins :D | 22:47 |
blackburn | (I renamed everything to blackburn instead of shogun) | 22:47 |
wiking | :DDDDDD | 22:50 |
-!- wiking [~wiking@huwico/staff/wiking] has quit [Quit: wiking] | 22:53 | |
blackburn | hmm got another idea for cool commit | 22:56 |
blackburn | n4nd0: hey did you make any progress? | 22:56 |
-!- Netsplit over, joins: PhilTillet | 23:01 | |
-!- PhilTillet [~philippe@157.159.47.10] has quit [Ping timeout: 245 seconds] | 23:17 | |
blackburn | have somebody ever grep for FUCK word in shogun git index? | 23:21 |
blackburn | :D | 23:21 |
CIA-113 | shogun: Sergey Lisitsyn master * r3df8a13 / (25 files in 7 dirs): | 23:24 |
CIA-113 | shogun: Integrated SLEP (SparseLEarningPackage) core code | 23:24 |
CIA-113 | shogun: - http://git.io/iB5IFg | 23:24 |
blackburn | bam | 23:24 |
-!- cronor_ [~cronor@g231216163.adsl.alicedsl.de] has joined #shogun | 23:40 | |
-!- cronor [~cronor@e178181022.adsl.alicedsl.de] has quit [Ping timeout: 256 seconds] | 23:40 | |
-!- cronor_ is now known as cronor | 23:40 | |
--- Log closed Thu May 10 00:00:37 2012 |
Generated by irclog2html.py 2.10.0 by Marius Gedminas - find it at mg.pov.lt!