--- Log opened Wed Aug 09 00:00:31 2017 | ||
-!- wiking [~wiking@huwico/staff/wiking] has quit [Quit: ZNC 1.6.3 - http://znc.in] | 03:44 | |
-!- wiking [~wiking@huwico/staff/wiking] has joined #shogun | 04:50 | |
-!- mode/#shogun [+o wiking] by ChanServ | 04:50 | |
@sukey | [https://github.com/shogun-toolbox/shogun] Pull Request https://github.com/shogun-toolbox/shogun/pull/3954 synchronized by olinguyen | 05:41 |
---|---|---|
@wiking | olinguyen, oh you are still awake? :) | 05:42 |
olinguyen | yea lo | 05:42 |
olinguyen | lol* | 05:42 |
@wiking | thnx for the change | 05:42 |
olinguyen | np! | 05:43 |
@wiking | as soon as it's green i'll merge it but it should be green :) | 05:43 |
olinguyen | awesome, thanks! | 05:43 |
@wiking | nw | 05:44 |
@sukey | [https://github.com/shogun-toolbox/shogun] Pull Request https://github.com/shogun-toolbox/shogun/pull/3960 synchronized by MikeLing | 06:03 |
@sukey | [https://github.com/shogun-toolbox/shogun] Pull Request https://github.com/shogun-toolbox/shogun/pull/3954 synchronized by olinguyen | 07:29 |
-!- olinguyen [81615ad9@gateway/web/freenode/ip.129.97.90.217] has quit [Quit: Page closed] | 07:36 | |
-!- mikeling [uid89706@gateway/web/irccloud.com/x-vsniquqegldoincf] has joined #shogun | 09:09 | |
mikeling | wiking: ping | 09:09 |
@wiking | ping | 09:09 |
@wiking | pong | 09:09 |
mikeling | wiking: hi, does all these changes in kernel.h https://github.com/shogun-toolbox/shogun/pull/3960/files/520198d7b89d133b6602d9a39ff30666eb730347#diff-512b8f6e61391930b28fe03be5f01bdbR184 make sense to you? | 09:10 |
mikeling | I just want to make sure these are necessary changes so I can change other places related to kernel. :) | 09:10 |
@wiking | lemme check | 09:10 |
@wiking | yes | 09:11 |
@wiking | those are necessary | 09:11 |
mikeling | ok great | 09:12 |
mikeling | thank you | 09:12 |
-!- Netsplit *.net <-> *.split quits: sheogorath27 | 10:49 | |
mikeling | wiking: what's this part of parameters for https://github.com/shogun-toolbox/shogun/blob/develop/src/shogun/distributions/HMM.h#L1331? Do we want them as index_t? | 11:19 |
-!- wiking [~wiking@huwico/staff/wiking] has quit [Quit: ZNC 1.6.3 - http://znc.in] | 11:20 | |
-!- wiking [~wiking@huwico/staff/wiking] has joined #shogun | 11:27 | |
-!- mode/#shogun [+o wiking] by ChanServ | 11:27 | |
@wiking | mikeling, ok so that's a good question | 11:27 |
-!- sukey [~nodebot@ks312251.kimsufi.com] has quit [Remote host closed the connection] | 11:33 | |
-!- sukey [~nodebot@ks312251.kimsufi.com] has joined #shogun | 11:34 | |
-!- mode/#shogun [+o sukey] by ChanServ | 11:34 | |
-!- sheogorath27 [sheogorath@gateway/shell/firrre/x-rdgcmdvvjpibutry] has joined #shogun | 11:34 | |
-!- HeikoS [~heiko@host-92-0-169-11.as43234.net] has joined #shogun | 12:22 | |
-!- mode/#shogun [+o HeikoS] by ChanServ | 12:22 | |
@wiking | mikeling, if you are here | 13:25 |
@wiking | i'd say keep it int32t | 13:25 |
@wiking | for the time being | 13:25 |
@wiking | HeikoS, man that hmm needs cleanup | 13:26 |
@wiking | HeikoS, are you around by any chance? | 13:26 |
@sukey | [https://github.com/shogun-toolbox/shogun] Pull Request https://github.com/shogun-toolbox/shogun/pull/3906 merged by vigsterkr | 13:28 |
@sukey | [https://github.com/shogun-toolbox/shogun] vigsterkr pushed 10 commits: | 13:28 |
@sukey | https://github.com/shogun-toolbox/shogun/commit/2d3324e5dd25e172f83f5b7712992555914b05b8 | 13:28 |
@sukey | https://github.com/shogun-toolbox/shogun/commit/fc97d5d2f4dc09fdaac7b30be91e96022a83a35d | 13:28 |
@sukey | https://github.com/shogun-toolbox/shogun/commit/e5a20fe8e67a54a465a1502cd561237791caff0c | 13:28 |
@sukey | https://github.com/shogun-toolbox/shogun/commit/a3ebf996e016fb85c1f4356213d3b21a24d5523c | 13:28 |
@sukey | https://github.com/shogun-toolbox/shogun/commit/80f7376150eb80981d9c47d1eded32edd3c28fdf | 13:28 |
@sukey | https://github.com/shogun-toolbox/shogun/commit/b2276776cb52962ae62645d7ab36de42026435a4 | 13:28 |
@sukey | https://github.com/shogun-toolbox/shogun/commit/c70f43d4304d0d65e035e042bf867ceac463b8d1 | 13:28 |
@sukey | https://github.com/shogun-toolbox/shogun/commit/817eeeb3a40e3358b6cf7b52c44ca7fedf8311c2 | 13:28 |
@sukey | https://github.com/shogun-toolbox/shogun/commit/5af0493da417d0d8e0fcfcb13f0a70dbcd960b54 | 13:28 |
@sukey | https://github.com/shogun-toolbox/shogun/commit/ef16d24f81fed90057991decce65c82c7e45983c | 13:28 |
@wiking | mooooo | 13:28 |
@wiking | aaaaa | 13:29 |
@HeikoS | wiking: yes | 13:33 |
@wiking | ok so do you maybe have 10 mins? | 13:33 |
@HeikoS | sure | 13:33 |
@HeikoS | just finished chapter 2 :D | 13:34 |
@wiking | hahaha | 13:34 |
@wiking | maaan | 13:34 |
@wiking | dont even remind me | 13:34 |
@wiking | anyhow so i was just thinking about | 13:36 |
@wiking | our solvers | 13:36 |
@wiking | uff btw in the meanwhile i realised | 13:37 |
@wiking | that maybe we should restructure our unti test | 13:37 |
@wiking | because now it's a bit stupid | 13:37 |
@wiking | that gpl related unit tests are in the main repo | 13:37 |
@wiking | but anyhow | 13:37 |
@HeikoS | wiking: why? | 13:38 |
@wiking | because it'd be better | 13:38 |
@HeikoS | I thikn thats fine | 13:38 |
@wiking | that they are with | 13:38 |
@wiking | gpl | 13:38 |
@HeikoS | why? | 13:38 |
@HeikoS | the tests are gpl | 13:38 |
@HeikoS | ehm | 13:38 |
@wiking | because they belong to each other? | 13:38 |
@HeikoS | not gpl | 13:38 |
@wiking | i mean no use to have a source around | 13:38 |
@HeikoS | but the gpl repo is not even a real project | 13:38 |
@wiking | if the corresponding | 13:38 |
@HeikoS | just source files | 13:38 |
@wiking | yeah | 13:38 |
@HeikoS | no cmake or anything | 13:38 |
@wiking | but we could foollow | 13:38 |
@HeikoS | no build | 13:38 |
@wiking | what most of the c++ project do nowadays | 13:38 |
@wiking | especially the bigger ones | 13:38 |
@HeikoS | I really see this as a workaround for the gpl invasive nature | 13:38 |
@HeikoS | so I see it as the same thing | 13:39 |
@HeikoS | but anyhow | 13:39 |
@HeikoS | feel free to change it | 13:39 |
@HeikoS | this is a minimal solution towars BSD so we can move on | 13:39 |
@wiking | yeah i mean i was thinking about changing | 13:39 |
@wiking | it as well | 13:39 |
-!- travis-ci [~travis-ci@ec2-54-211-49-115.compute-1.amazonaws.com] has joined #shogun | 13:39 | |
travis-ci | it's Viktor Gal's turn to pay the next round of drinks for the massacre he caused in shogun-toolbox/shogun: https://travis-ci.org/shogun-toolbox/shogun/builds/262623450 | 13:39 |
-!- travis-ci [~travis-ci@ec2-54-211-49-115.compute-1.amazonaws.com] has left #shogun [] | 13:39 | |
@wiking | because it's a bit shitty now | 13:39 |
@wiking | taht you have to move back and forth | 13:39 |
@wiking | between | 13:39 |
@wiking | dir structures | 13:39 |
@wiking | (trees) | 13:39 |
@HeikoS | it is | 13:39 |
@wiking | for the implementation and it's corresponding | 13:39 |
@wiking | untittest | 13:39 |
@wiking | *unittest | 13:40 |
@wiking | bigger c++ project | 13:40 |
@wiking | interleave these things | 13:40 |
@wiking | _impl.cc _unittest.cc | 13:40 |
@HeikoS | yeah sure | 13:40 |
@wiking | if we'd have this | 13:40 |
@wiking | then we'd simply just put all the gpl stuff | 13:40 |
@wiking | into the gpl place | 13:40 |
@wiking | and that's all | 13:40 |
@HeikoS | no guarding | 13:40 |
@wiking | yeah etc | 13:40 |
@HeikoS | well in practice, this is hard | 13:40 |
@wiking | ? | 13:40 |
@HeikoS | because of how things are convoluted | 13:41 |
@wiking | why hard? | 13:41 |
@HeikoS | with meta examples as well | 13:41 |
@wiking | mmm yeah meta is a different story | 13:41 |
@wiking | dont wanna touch that | 13:41 |
@wiking | just unittest for the time being | 13:41 |
@HeikoS | I think it is ok as things are now actually | 13:41 |
@wiking | k | 13:41 |
@HeikoS | but anyhow | 13:41 |
@wiking | i'll change if you dont have any strong feelings | 13:41 |
@HeikoS | what about the solvers? | 13:41 |
@wiking | as i'd like to have | 13:41 |
@HeikoS | no go ahead | 13:41 |
@wiking | the impl + unittest | 13:41 |
@wiking | in one place | 13:41 |
@HeikoS | sure do it if you want | 13:41 |
@wiking | this way we know as well | 13:41 |
@HeikoS | I prefer that in fact | 13:41 |
@HeikoS | but I am also fine with it is now | 13:42 |
@wiking | wha tthe hell is being tested | 13:42 |
@wiking | kk | 13:42 |
@wiking | so this just came up because i was looking for our friend | 13:42 |
@wiking | DualLibQPBMSOSVM | 13:42 |
@wiking | and the whole solver story in shogun | 13:43 |
@wiking | and as well the solvers in tf for example | 13:43 |
@HeikoS | yeah should be splot | 13:44 |
@HeikoS | with sep api | 13:44 |
@HeikoS | it is so annoying that I cannot use a QP solver for the MMD stuff | 13:44 |
@HeikoS | since the only one that is there is GPL | 13:44 |
@HeikoS | and the others are tangled with some other stuff I dont care about | 13:44 |
@wiking | and then we have this beauty | 13:45 |
@wiking | https://github.com/shogun-toolbox/shogun/blob/270ff9200daeb273bec6c82543d82d5a275b39d3/src/shogun/machine/StructuredOutputMachine.h#L122 | 13:45 |
@HeikoS | yeah | 13:45 |
@HeikoS | I mean | 13:45 |
@HeikoS | SO | 13:45 |
@HeikoS | wtf | 13:45 |
@HeikoS | that is dead effectively | 13:45 |
@HeikoS | the community moved on to deep nets ;) | 13:45 |
@wiking | mmm | 13:46 |
@wiking | i dont see why it has to be dead | 13:46 |
@wiking | :) | 13:46 |
@wiking | i mean yeah it's note being used | 13:46 |
@wiking | but precisely for the fact | 13:46 |
@wiking | that for example there | 13:46 |
@HeikoS | I remember how thoralf told me numerous times that this whole framework is fitting things into SVM framework with hammers | 13:46 |
@HeikoS | haha | 13:46 |
@HeikoS | anyhow | 13:46 |
@HeikoS | we have it | 13:46 |
@wiking | you had to generate your subgrad | 13:46 |
@wiking | which is why everybody is | 13:46 |
@wiking | fuck that | 13:46 |
@wiking | :) | 13:46 |
@wiking | they are like man i have to code for my specific case the subgrad | 13:47 |
@wiking | lalala | 13:47 |
@wiking | but i mean | 13:47 |
@wiking | couldn't we actually use now some nice solvers :) | 13:47 |
@HeikoS | would be cool to director class that | 13:47 |
@HeikoS | sure | 13:47 |
@HeikoS | that would be cool | 13:47 |
@wiking | that would actually calculate | 13:47 |
@wiking | the subgrads | 13:47 |
@wiking | because apparently we. could | 13:47 |
@wiking | no? | 13:47 |
@wiking | or am i being delusional? | 13:48 |
@HeikoS | how? | 13:51 |
@HeikoS | I don't know that framework too well | 13:51 |
@HeikoS | but aren't these things problem specific? | 13:51 |
@wiking | well yeah you need to define your | 13:52 |
@wiking | delta over the SO | 13:52 |
@wiking | which is actually what you wanna most tweak | 13:52 |
@wiking | and that is problem specific | 13:52 |
@HeikoS | so what do you want to do there? | 13:53 |
@wiking | and then there's the fact of how to represent the PSI(x,y) | 13:54 |
@wiking | but then it's a PO without any conditions | 13:54 |
@wiking | so surely tf could solve that for us no? | 13:54 |
@wiking | sorry | 13:55 |
@HeikoS | you mena the tf solvers? | 13:55 |
@wiking | i'm in the meanwhile cooking and rebasing feature/random-refactor | 13:55 |
@HeikoS | so you want to re-build the SO framework in tf effectively, which will give all the solvers from there, as well as the gradient stuff? | 13:55 |
@wiking | yeah i mean now comes the question | 13:55 |
@wiking | how this would/should look like | 13:55 |
@wiking | because surely we would like to use | 13:56 |
@wiking | the solvers one way or another | 13:56 |
@wiking | from tf | 13:56 |
@wiking | not only in SO | 13:56 |
@wiking | the other somebody had a problem that was very SO-like | 13:56 |
@wiking | remembered shit | 13:56 |
@wiking | then i realised that fuck with all those manual subgradient calculation/coding | 13:56 |
@wiking | this is horrible | 13:56 |
@HeikoS | yep | 13:57 |
@HeikoS | I get it | 13:57 |
@HeikoS | would be very nice | 13:57 |
@wiking | but then i was wondering if we could actually use TF here | 13:57 |
@wiking | or stan for that matter | 13:57 |
@wiking | whichever would be the better fit | 13:57 |
@wiking | i mean i have no idea what is exposed | 13:57 |
@wiking | by stan's c++ api | 13:57 |
@wiking | but i guess | 13:57 |
@wiking | you can do anything you want no? | 13:58 |
@HeikoS | yeah sure | 13:58 |
@HeikoS | stan's api is not the nicest | 13:58 |
@HeikoS | also they dont have this graph stuff | 13:58 |
@HeikoS | I dont think that is the way | 13:58 |
@HeikoS | tf is the way | 13:58 |
@wiking | yeah tf was my immediate idea | 13:58 |
@HeikoS | ok then | 13:59 |
@HeikoS | but in order to do this | 13:59 |
@wiking | yeah i mean we need optimizer framework | 13:59 |
@HeikoS | we need to have tf graphs in | 13:59 |
@HeikoS | which is quite a task | 13:59 |
@HeikoS | interfacing with tf | 13:59 |
@HeikoS | would be SO useful for other things as well | 13:59 |
@HeikoS | gps | 13:59 |
@wiking | yeah | 13:59 |
@HeikoS | I mean if you could have a linalg like thing with tf | 13:59 |
@HeikoS | that would do it | 13:59 |
@HeikoS | but we would need to copy the API behind a wrapper | 14:00 |
@HeikoS | then all this is possible | 14:00 |
@HeikoS | or just directly use tf | 14:00 |
@HeikoS | without wrapper | 14:00 |
@HeikoS | but I am hesitant with this as we had the problem of changing backends so many times now | 14:00 |
@wiking | nono | 14:00 |
@wiking | i would wrap it | 14:01 |
@HeikoS | yeah ok | 14:01 |
@wiking | because yeah | 14:01 |
@wiking | what if there's tf++ | 14:01 |
@HeikoS | I mean that would be huge | 14:01 |
@wiking | whatevermachine | 14:01 |
@HeikoS | gamechanger | 14:01 |
@wiking | and we need to redo everything | 14:01 |
@wiking | imo better just do a wrapping | 14:01 |
@wiking | HeikoS, yeah i mean im just planning | 14:01 |
@HeikoS | I would love to do that | 14:01 |
@wiking | if things go well | 14:01 |
@HeikoS | since that really gives a perspective for shogun long term | 14:01 |
@wiking | what could be a long running proj | 14:01 |
@HeikoS | so yeah | 14:02 |
@HeikoS | I discussed this a lot with others before | 14:02 |
@HeikoS | but I don't really know how to go ahead | 14:02 |
@wiking | mmm | 14:02 |
@wiking | gist .h | 14:02 |
@wiking | :D | 14:02 |
@HeikoS | we could spend a hackathon prototyping stuff for that | 14:02 |
@wiking | as usual | 14:02 |
@HeikoS | I just wonder, is it the time for that already? | 14:02 |
@HeikoS | or shouldnt we solve some other problems first | 14:02 |
@HeikoS | like the tags and cereal | 14:02 |
@wiking | what would be more immediate? | 14:03 |
@wiking | mmm tago | 14:03 |
@wiking | yeah | 14:03 |
@wiking | and being able to cereal | 14:03 |
@wiking | yeah | 14:03 |
@wiking | both are valid | 14:03 |
@HeikoS | and this stuff we discussed in zurich | 14:03 |
@wiking | which? | 14:03 |
@HeikoS | no templates in algorithms | 14:03 |
@wiking | aah | 14:03 |
@HeikoS | Matrix m = feats.get_cov | 14:03 |
@wiking | but yeah i wrote something about somewhere a bit | 14:03 |
@wiking | regarding | 14:03 |
@HeikoS | linalg::invert(m) | 14:03 |
@wiking | the typedetection | 14:03 |
@HeikoS | yeah | 14:03 |
@wiking | and how not to do it again | 14:03 |
@wiking | because the | 14:03 |
@wiking | IF type if type if type | 14:03 |
@HeikoS | yes that needs some clever stuff | 14:04 |
@wiking | that horrible | 14:04 |
@HeikoS | I agree | 14:04 |
@HeikoS | I mean but | 14:04 |
@HeikoS | the API doesnt care about that too much I think | 14:04 |
@HeikoS | this is backend stuff | 14:04 |
@wiking | yeah | 14:04 |
@wiking | sure i mean i just dont wanna do this behind the api | 14:04 |
@HeikoS | so doing the API (e.g. this GSoC ) would be good | 14:04 |
@wiking | that in every function call | 14:04 |
@HeikoS | yeah agreed | 14:04 |
@wiking | there's an if type switch | 14:04 |
@wiking | :) | 14:04 |
@wiking | kk | 14:04 |
@wiking | but having the api cleaned would be good | 14:04 |
@wiking | actually | 14:04 |
@wiking | am i right | 14:04 |
@wiking | that then those motherfucking | 14:04 |
@HeikoS | maybe you can push micmn a bit more on that | 14:04 |
@wiking | macros | 14:05 |
@wiking | would go? | 14:05 |
@HeikoS | the draft he made is useful | 14:05 |
@wiking | because we dont need no more | 14:05 |
@HeikoS | could put that in, change an algorithm or two to use it | 14:05 |
@HeikoS | and then we can think about how to detect the types | 14:05 |
@wiking | macro it all the way the virtual functions | 14:05 |
@HeikoS | I know, we can use annotators | 14:05 |
@HeikoS | @float64_t | 14:05 |
@HeikoS | :D | 14:05 |
@wiking | lol | 14:05 |
@wiking | :D | 14:05 |
@wiking | whish | 14:05 |
@wiking | *wish | 14:06 |
@wiking | there's something coming | 14:06 |
@HeikoS | but basically we need to tell linalg: the next 10 operations are float64 | 14:06 |
@wiking | in c++20? :D | 14:06 |
@wiking | yeah | 14:06 |
@HeikoS | and it this is done via lazy eval, fine | 14:06 |
@HeikoS | Ill make lunch now | 14:06 |
@HeikoS | wiking: but yeah can discuss this more | 14:06 |
@HeikoS | I think micmn's patch is a good way to start this problem | 14:06 |
@HeikoS | tags where to start I dont know | 14:07 |
@HeikoS | as we failed this GSoC :) | 14:07 |
@HeikoS | cereal depends on tags | 14:07 |
@HeikoS | and then I am all game for TF integration | 14:07 |
-!- HeikoS [~heiko@host-92-0-169-11.as43234.net] has quit [Ping timeout: 248 seconds] | 14:13 | |
Trixis | wiking: shogun apache spark compatible version please:D | 14:48 |
@wiking | Trixis, what is the incompatilibyt measure? | 14:48 |
Trixis | wiking: nah i mean sparkify all the operations that can be sparkified :D | 14:48 |
@wiking | okok | 14:48 |
@wiking | gimme gimme ideas | 14:48 |
@wiking | and then i can haz it | 14:49 |
@wiking | but yeah we dont have batch support for shit | 14:49 |
@wiking | so you have to figure out yoruself | 14:49 |
@wiking | a parameter server | 14:49 |
@wiking | and batchify somehow your stuff | 14:49 |
@wiking | right? | 14:49 |
Trixis | (more just a dream for future really, ive been fuming that the cluster im using doesnt have spark) | 14:50 |
Trixis | wiking: it seems to be that for example kernel learning could benefit from parallelisation across multiple machines via spark/hadoop/whatever | 14:50 |
Trixis | *seems to me | 14:50 |
Trixis | since some kernels / transforms are very expensive to compute, and the full matrix can require up to O(n^2) of such operations | 14:51 |
@wiking | but what | 14:51 |
@wiking | you would take the submatrix | 14:51 |
@wiking | and do it batch by batch? | 14:52 |
@wiking | and then reduce it ? | 14:52 |
Trixis | wiking: um more of take the features, evaluate the individual kernel products, batch by batch, then reduce it into the matrix | 14:52 |
Trixis | wiking: alternatively if youre training a large amount if individual learners, be it for evaluation or some combined classifier, you might as well do it batch by batch, then reduce it, no? | 14:54 |
@wiking | :) | 15:03 |
@wiking | yeah that's doable | 15:03 |
@wiking | i guess only way is | 15:03 |
@wiking | that you have a view on the matrix | 15:03 |
@wiking | right?? | 15:03 |
@wiking | i mean Features | 15:04 |
@wiking | and then you do a batch | 15:04 |
@wiking | mikeling, around? | 15:05 |
mikeling | yes | 15:05 |
@wiking | ok so | 15:09 |
@wiking | in | 15:09 |
@wiking | just a sec trying to find the commit | 15:09 |
mikeling | ok | 15:10 |
@wiking | https://github.com/shogun-toolbox/shogun/pull/3906/commits/5af0493da417d0d8e0fcfcb13f0a70dbcd960b54 | 15:11 |
-!- Saurabh7_ [Saurabh7@gateway/shell/panicbnc/x-qjqdjvhmwkfngaea] has quit [Ping timeout: 255 seconds] | 15:11 | |
@wiking | here you removed CMath::permute | 15:11 |
@wiking | but didn't add it anywhere | 15:11 |
@wiking | so currently as i've merged the pr into feature/random-refactor | 15:11 |
@wiking | src/shogun/CMakeFiles/libshogun.dir/clustering/KMeansBase.cpp fails | 15:12 |
@wiking | as it calls CMath::permute | 15:12 |
@wiking | git grep CMath::permute|wc -l | 15:12 |
@wiking | 38 | 15:12 |
@wiking | so there are 38 places it's references | 15:12 |
@wiking | *referenced | 15:12 |
-!- iglesiasg [~iglesiasg@217.119.234.214] has joined #shogun | 15:12 | |
-!- mode/#shogun [+o iglesiasg] by ChanServ | 15:12 | |
@wiking | mikeling, what was there the idea? :) | 15:13 |
@wiking | i mean now i'll force push that branch as i rebased it over the latest develop | 15:13 |
@wiking | so you should send in a PR that fixes that | 15:13 |
@wiking | ok | 15:13 |
@wiking | ? | 15:13 |
mikeling | mmmm, wait a second | 15:14 |
mikeling | ok, sure. I will send a pr for that | 15:14 |
@wiking | k lemme test the rest | 15:15 |
mikeling | but actually , do i really remove the CMath::permute? I do remember we had discuss about that | 15:15 |
@wiking | because the rebase wasn't an easy story | 15:15 |
@wiking | mikeling, my idea is to move it to CRandom | 15:15 |
mikeling | you said maybe we should keep Random somehow for permute | 15:15 |
@wiking | or that doesn't exist anymore? | 15:16 |
@wiking | if not then we can move it to a utility function? :) | 15:16 |
@wiking | i mean honestly | 15:17 |
@wiking | maybe it's just keep it there? :) | 15:17 |
@wiking | mikeling, i can actually amend your last commit | 15:17 |
@wiking | and just leave it there for the time being | 15:17 |
@wiking | :P | 15:17 |
@wiking | mikeling, any objections? | 15:17 |
mikeling | no | 15:18 |
@wiking | k | 15:18 |
@wiking | i'll try to come up with the fixtures | 15:18 |
mikeling | ok, thank you :) | 15:18 |
Trixis | wiking: hm i guess yeah | 15:21 |
@wiking | ok feature/random-refactor is rebased and should be good | 15:55 |
@sukey | [https://github.com/shogun-toolbox/shogun] vigsterkr pushed 12 commits: | 15:55 |
@sukey | https://github.com/shogun-toolbox/shogun/commit/306f0c6fe06349ecfff05725c49f58699094edb8 | 15:55 |
@sukey | https://github.com/shogun-toolbox/shogun/commit/5d4c0372b7e33941db28bcb733cb6753fd07f5f6 | 15:55 |
@wiking | now it's only about the fixtures | 15:55 |
@sukey | https://github.com/shogun-toolbox/shogun/commit/276a74d62ef785d6d3162b24dfeb7b8f45915a9e | 15:55 |
@sukey | https://github.com/shogun-toolbox/shogun/commit/c81f0df50e9c4fd88122061f6ede83409fac2451 | 15:55 |
@sukey | https://github.com/shogun-toolbox/shogun/commit/de234f6a778e36932d2a9860afc234bfe64abb05 | 15:55 |
@sukey | https://github.com/shogun-toolbox/shogun/commit/849740409ce85de7710fb234b2927fe6a6d510ef | 15:55 |
@sukey | https://github.com/shogun-toolbox/shogun/commit/c0a7ef3c19117fc8fec60e9022907930208dee55 | 15:55 |
@sukey | https://github.com/shogun-toolbox/shogun/commit/02dfb1cb65238bdc60e2409aa74905cc0da55f92 | 15:55 |
@sukey | https://github.com/shogun-toolbox/shogun/commit/aa9cfc300d36bf89fcadd74e6e207f904183da44 | 15:55 |
@sukey | https://github.com/shogun-toolbox/shogun/commit/cc3eaaa340a7a9592fcc1bbd59adaf2109a3c55c | 15:55 |
@sukey | https://github.com/shogun-toolbox/shogun/commit/35984d3be8faa336755ca24f362bc76067971b2d | 15:55 |
@sukey | https://github.com/shogun-toolbox/shogun/commit/4f6831ce4739b772865b2a6a8afacf439af00b52 | 15:55 |
@wiking | sukey, flip | 15:55 |
@sukey | (ノಥ益ಥ)ノ ┻━┻ | 15:55 |
@sukey | [https://github.com/shogun-toolbox/shogun] New commit https://github.com/shogun-toolbox/shogun/commit/58ef0574871acab3bbb322236036226fb27d7165 by vigsterkr | 16:08 |
-!- HeikoS [~heiko@host-92-0-169-11.as43234.net] has joined #shogun | 16:33 | |
-!- mode/#shogun [+o HeikoS] by ChanServ | 16:33 | |
-!- witness [uid10044@gateway/web/irccloud.com/x-qjwxxgenczokfsxk] has joined #shogun | 16:47 | |
-!- travis-ci [~travis-ci@ec2-54-211-49-115.compute-1.amazonaws.com] has joined #shogun | 17:04 | |
travis-ci | it's Viktor Gal's turn to pay the next round of drinks for the massacre he caused in shogun-toolbox/shogun: https://travis-ci.org/shogun-toolbox/shogun/builds/262674922 | 17:04 |
-!- travis-ci [~travis-ci@ec2-54-211-49-115.compute-1.amazonaws.com] has left #shogun [] | 17:04 | |
-!- olinguyen [81615ad9@gateway/web/freenode/ip.129.97.90.217] has joined #shogun | 17:29 | |
olinguyen | HeikoS: hey | 17:30 |
-!- travis-ci [~travis-ci@ec2-54-211-49-115.compute-1.amazonaws.com] has joined #shogun | 17:49 | |
travis-ci | it's Viktor Gal's turn to pay the next round of drinks for the massacre he caused in shogun-toolbox/shogun: https://travis-ci.org/shogun-toolbox/shogun/builds/262680160 | 17:49 |
-!- travis-ci [~travis-ci@ec2-54-211-49-115.compute-1.amazonaws.com] has left #shogun [] | 17:49 | |
@HeikoS | olinguyen: hi there | 17:49 |
olinguyen | I can give you a rundown of the notebook here | 17:50 |
olinguyen | I plotted the heart rate of 10 random patients that did not survive after 24 hours (shown in red), and 10 that did (shown in blue) to try and see if there's a significant difference. | 17:50 |
@HeikoS | shall we have a chat? | 17:51 |
@HeikoS | as in video? | 17:51 |
olinguyen | sure | 17:51 |
olinguyen | let me setup my laptop | 17:52 |
@HeikoS | ok | 17:52 |
@HeikoS | olinguyen: https://github.com/olinguyen/gsoc2017-shogun-dataproject/blob/master/Improved%20Model.ipynb | 17:56 |
@HeikoS | olinguyen: https://pandas.pydata.org/pandas-docs/stable/generated/pandas.Series.resample.html | 18:06 |
@sukey | [https://github.com/shogun-toolbox/shogun] Pull Request https://github.com/shogun-toolbox/shogun/pull/3954 synchronized by olinguyen | 18:32 |
-!- mikeling [uid89706@gateway/web/irccloud.com/x-vsniquqegldoincf] has quit [Quit: Connection closed for inactivity] | 18:38 | |
@wiking | ok i guess this is now mergable then https://github.com/shogun-toolbox/shogun/pull/3954 | 19:22 |
@wiking | right? | 19:23 |
@wiking | if so i'll do a squash merge | 19:23 |
@HeikoS | olinguyen: you still around? | 20:12 |
@HeikoS | olinguyen: check your email | 20:13 |
@HeikoS | https://gist.github.com/karlnapf/f4ff9a69e5baede838212c2a9f5947ab | 20:13 |
@HeikoS | let me know if you have questions on this | 20:13 |
olinguyen | ok, going through it now | 20:13 |
@HeikoS | if not, try to replicate this for multiple patients and try to train a model on it, testing it on different patients that you trained on | 20:13 |
olinguyen | thanks for taking the time to do this | 20:13 |
@HeikoS | olinguyen: I gotta leave now | 20:13 |
@HeikoS | olinguyen: no worries | 20:13 |
-!- HeikoS [~heiko@host-92-0-169-11.as43234.net] has quit [Ping timeout: 255 seconds] | 20:19 | |
-!- olinguyen [81615ad9@gateway/web/freenode/ip.129.97.90.217] has quit [Ping timeout: 260 seconds] | 21:28 | |
shogitter | (olinguyen) Cool | 22:26 |
-!- olinguyen [81615ad9@gateway/web/freenode/ip.129.97.90.217] has joined #shogun | 22:27 | |
shogitter | (olinguyen) test | 22:27 |
-!- witness [uid10044@gateway/web/irccloud.com/x-qjwxxgenczokfsxk] has quit [Quit: Connection closed for inactivity] | 22:37 | |
-!- Saurabh7_ [Saurabh7@gateway/shell/panicbnc/x-xmqojrmdxdpwiivs] has joined #shogun | 23:51 | |
--- Log closed Thu Aug 10 00:00:32 2017 |
Generated by irclog2html.py 2.10.0 by Marius Gedminas - find it at mg.pov.lt!