IRC logs of #shogun for Thursday, 2019-05-02

« Wednesday, 2019-05-01 Index Friday, 2019-05-03 »
--- Log opened Thu May 02 00:00:57 2019
-!- durovo1 [~durovo@8e.68.2ea9.ip4.static.sl-reverse.com] has quit [Remote host closed the connection]04:51
-!- durovo [~durovo@8e.68.2ea9.ip4.static.sl-reverse.com] has joined #shogun04:51
-!- durovo [~durovo@8e.68.2ea9.ip4.static.sl-reverse.com] has quit [Read error: Connection reset by peer]05:38
-!- durovo [~durovo@8e.68.2ea9.ip4.static.sl-reverse.com] has joined #shogun05:38
-!- gf712 [bcd58b26@gateway/web/freenode/ip.188.213.139.38] has joined #shogun08:57
-!- geektoni [c1cdd253@gateway/web/freenode/ip.193.205.210.83] has joined #shogun09:13
-!- HeikoS [~heiko@147.red-176-84-171.dynamicip.rima-tde.net] has joined #shogun09:18
-!- mode/#shogun [+o HeikoS] by ChanServ09:18
gf712HeikoS: ping09:26
@HeikoSgf712: yo!09:27
@HeikoSgf712:  check out this if you have  a moment today09:28
@HeikoShttp://blog.detoni.me/2019/04/04/Observable-Machine-Learning-Models-Shogun/09:28
@HeikoSvery nice post by geektoni09:28
geektoniHeikoS: too kind :)09:28
@HeikoSno man it is really nice !09:28
gf712sure! I will have a look!09:29
gf712HeikoS: is there support for date format in shogun?09:30
@HeikoSgf712: nope09:30
@HeikoSgf712:  what for?09:30
gf712I am finishing writing the ARFF deserialiser09:30
@HeikoSsorry I mean why do you ask?09:30
@HeikoSah09:30
gf712and there are date fields09:30
@HeikoSmaybe there is some header only mini lib for stuff like that somewhere?09:30
gf712but I don't think anyone implements them09:30
gf712mhh maybe09:31
gf712it's not important09:31
gf71299% of files should be readable by the parser I have09:31
gf712I'll download a bunch and see how it does...09:31
@HeikoSgeektoni: check twitter :)09:33
@HeikoSgf712: ok :)09:33
@HeikoSI mean datetime -> features might be something usefil09:33
@HeikoSor datetime as index for data09:33
@HeikoSbut that is future stuff09:33
gf712cool! I would imagine they might need it for the env sciences09:35
gf712HeikoS: btw been working with Viktor on transferring everything to smart pointers09:35
gf712in the whole of shogun09:36
gf712no refcount09:36
gf712and no mom leaks :D09:36
gf712mem*09:36
geektoniHeikoS: \o/ that's really cool!09:36
@HeikoSgf712: yes viktor sent a text that xval works09:40
@HeikoSthis is so great09:40
@HeikoSgf712:  can you tell me some details of what you exactly did and what works/doesnt work09:40
@HeikoSgf712: how do you deal with inheritance/polymorphism for example?09:41
gf712ok, in C++ part most of it was replaced with a lib tooling script09:41
@HeikoSlike I pass shared_ptr<CBinaryLabels> to a method which wants shared_ptr<CLabels>09:41
gf712that turns out to be fine09:41
gf712we were surprised09:41
gf712you can even do09:41
gf712(shared_ptr<CLabels>*) shared_ptr<CBinaryLabels>*09:42
gf712so all the sg_add works fine09:42
gf712you can also use dynamic_pointer_cast for lvalues09:43
gf712i.e. std::dynamic_pointer_cast<CBinaryLabels>(a_labels_base_class_shared_ptr)09:43
gf712there are some tricky bits09:44
gf712when you have circular references09:44
gf712where you need to use weakref09:44
gf712weak_ptr09:44
@HeikoSah ok09:45
@HeikoScool09:45
@HeikoSreally nice09:45
gf712and you can't pass `this` inside a constructor to a function that takes shared_ptr09:45
gf712next step is swig09:45
@HeikoSah yes09:45
gf712but meta for cpp all work09:45
@HeikoScool09:45
gf712unit tests most work09:45
@HeikoSshared_ptr is not available for all langs or?09:45
gf712except serialisation09:45
gf712most do09:45
@HeikoSok i mean whatever09:46
gf712the ones that are stable do09:46
@HeikoSas long as python works09:46
gf712from the top of my head I think it is just Lua that doesn't work09:46
gf712but if needed we can have a look09:46
gf712and implement it for that09:46
gf712also swig is being actively developed now09:46
gf712so maybe there will be more shared_ptr stuff09:47
gf712in any case, then it's the serialisation stuff09:47
gf712which is victors other pr09:47
gf712and then we can transfer my openml stuff to that09:48
gf712Viktor is doing the REST stuff with curl09:48
gf712and then can use openml in c++ + swig :DS09:49
gf712:D09:49
@HeikoSgreat great09:50
@HeikoSbtw I just got an email asking about your contract09:50
@HeikoSI think the turing will start a leaving process now for you09:50
@HeikoS(annoyingly)09:50
@HeikoSand then if we get the proposal accepted, it would be re-vived09:50
@HeikoS(rather than doing the easy thing of just extending your contract)09:51
gf712ah ok! they haven't sent me anything... oh well that's fine09:52
gf712Viktor talked to someone in legal yesterday and they said there will be a decision on the project at the end of the month09:52
-!- durovo [~durovo@8e.68.2ea9.ip4.static.sl-reverse.com] has quit [Read error: Connection reset by peer]09:52
@HeikoSyeah09:52
@HeikoSthats what the submission system said09:53
-!- durovo [~durovo@8e.68.2ea9.ip4.static.sl-reverse.com] has joined #shogun09:53
gf712ah ok, good09:53
@HeikoSnow we just need to have fingers crossed the reviewers will be happy09:53
gf712do you know how competitive it is?09:53
gf712i.e. number of applicants vs number of funded projects09:53
@HeikoSI think they want to fund 6 concurrently to start with09:55
@HeikoSand then have multiple submission rounds09:55
@HeikoSas for numbers, you saw the number of posters at the ws09:55
@HeikoSbut one thing i am quite confident about: we wrote a strong proposal imo09:55
@HeikoSI saw a few others and they were not nearly as concrete09:56
@HeikoSbut there is a big big political component with this as well09:56
@HeikoSand there idk where we stand, as I know a few peeps at the Turing but I am not senior, and they dont know anyone else in the team09:56
@HeikoSbut we will see09:57
@HeikoSI think they are generally quite positive towards us09:57
@HeikoSbut that is the PIs at the Turing, and idk who reviews09:57
@HeikoSlet's hope it's not cambridge people09:57
gf712well, at least we have other people in the proposal, it will definitely help10:01
gf712it would be very cool working with methodologist on a concrete problem10:01
gf712but yea, fingers crossed..10:01
@HeikoSwe have scott and two other co-pis10:05
@HeikoSso very nice10:05
@HeikoSembedding into the turing10:05
@HeikoSif it happens10:05
@HeikoSwith this Turing Way book etc10:05
gf712yea, would be cool being one their development tools, would get so much feedback about api design and so on...10:06
@HeikoSexactly10:07
@HeikoSalso interesting for you in terms of job perspective or?10:08
@HeikoSbeing exposed to scientists who need good tools etc10:08
@HeikoSI remember thats something you wanna do10:08
gf712yea, I want to sit somewhere in between software devel and research10:08
gf712so I can switch a bit around10:08
gf712in terms of job, yes would definitely be good10:09
@HeikoScool10:10
@HeikoShope it works out !:)10:10
gf712yea me too!10:10
@HeikoSgf712:  oh btw10:14
@HeikoSmaybe you know this10:14
@HeikoSI have a problem with eigen and overflow10:14
@HeikoSI have a Map around a large array of float32 numbers10:15
@HeikoSeach number is large10:15
@HeikoSI want to compute the sum of the numbers10:15
@HeikoSbut this number is too large for float3210:15
@HeikoSso I would need to tell eigen to use float64 or floatmax as return type10:15
@HeikoSwhen doing .sum()10:15
@HeikoSyou know whether that is possible or how?10:15
gf712mhhh I am not sure10:25
gf712I know you can cast in Eigen...10:26
gf712matrix.cast <float64> ();10:26
gf712but I am not sure how to tell it to just cast for return value...10:27
gf712HeikoS: can you do (matrix_a + matrix_b).cast<float64_t>() ?10:28
@HeikoSis that zero cost?10:28
@HeikoSgf712: because I dont want to store it as float6410:29
gf712I am not sure10:36
gf712you would have to test10:37
gf712but if you chain it all without storing it in a MatrixXd I don't think it would get evaluated?10:38
@HeikoSok cool let me try10:44
-!- geektoni [c1cdd253@gateway/web/freenode/ip.193.205.210.83] has quit [Ping timeout: 256 seconds]11:14
@HeikoSgf712: remind me, did you need funding for travel/accommodation for berlin ws?11:25
-!- geektoni [c1cdd253@gateway/web/freenode/ip.193.205.210.83] has joined #shogun11:33
gf712HeikoS: yes, but I'll try to get ati /ucl to pay for it11:36
@HeikoSok11:36
@HeikoSi excluded you in the budget for now11:36
gf712yup, that's fine11:42
-!- HeikoS [~heiko@147.red-176-84-171.dynamicip.rima-tde.net] has quit [Ping timeout: 244 seconds]12:00
-!- HeikoS [~heiko@147.red-176-84-171.dynamicip.rima-tde.net] has joined #shogun12:04
-!- mode/#shogun [+o HeikoS] by ChanServ12:04
-!- HeikoS [~heiko@147.red-176-84-171.dynamicip.rima-tde.net] has quit [Quit: Leaving.]12:11
-!- geektoni [c1cdd253@gateway/web/freenode/ip.193.205.210.83] has quit [Quit: Page closed]12:29
-!- geektoni [973eb7d5@gateway/web/freenode/ip.151.62.183.213] has joined #shogun14:47
-!- besser82 [~besser82@fedora/besser82] has quit [Quit: Freedom, Friends, Features, First [fedoraproject.org]]15:36
-!- besser82 [~besser82@fedora/besser82] has joined #shogun15:37
-!- mode/#shogun [+o besser82] by ChanServ15:37
-!- gf712 [bcd58b26@gateway/web/freenode/ip.188.213.139.38] has quit [Ping timeout: 256 seconds]15:47
-!- gf712 [bcd58b26@gateway/web/freenode/ip.188.213.139.38] has joined #shogun16:27
-!- gf712 [bcd58b26@gateway/web/freenode/ip.188.213.139.38] has quit [Quit: Page closed]17:00
-!- geektoni [973eb7d5@gateway/web/freenode/ip.151.62.183.213] has quit [Ping timeout: 256 seconds]17:08
-!- durovo [~durovo@8e.68.2ea9.ip4.static.sl-reverse.com] has quit [Read error: Connection reset by peer]22:17
-!- durovo [~durovo@8e.68.2ea9.ip4.static.sl-reverse.com] has joined #shogun22:18
-!- durovo [~durovo@8e.68.2ea9.ip4.static.sl-reverse.com] has quit [Read error: Connection reset by peer]22:34
-!- durovo [~durovo@8e.68.2ea9.ip4.static.sl-reverse.com] has joined #shogun22:34
--- Log closed Fri May 03 00:00:59 2019
« Wednesday, 2019-05-01 Index Friday, 2019-05-03 »

Generated by irclog2html.py 2.10.0 by Marius Gedminas - find it at mg.pov.lt!