| --- Log opened Sun Feb 24 00:00:50 2013 | ||
| -!- alexlovesdata [~binder@e178001222.adsl.alicedsl.de] has joined #shogun | 01:14 | |
| -!- alexlovesdata [~binder@e178001222.adsl.alicedsl.de] has left #shogun [] | 01:22 | |
| shogun-buildbot_ | build #292 of nightly_default is complete: Failure [failed test] Build details are at http://www.shogun-toolbox.org/buildbot/builders/nightly_default/builds/292 | 04:00 |
|---|---|---|
| -!- sumit_ [75ea48a9@gateway/web/freenode/ip.117.234.72.169] has joined #shogun | 06:45 | |
| -!- sumit_ [75ea48a9@gateway/web/freenode/ip.117.234.72.169] has quit [Ping timeout: 245 seconds] | 08:06 | |
| -!- hoijui [~hoijui@dslb-092-078-043-220.pools.arcor-ip.net] has joined #shogun | 08:33 | |
| -!- lambday [0e8b614f@gateway/web/freenode/ip.14.139.97.79] has joined #shogun | 13:42 | |
| -!- wiking [~wiking@huwico/staff/wiking] has quit [Read error: Connection reset by peer] | 14:43 | |
| -!- n4nd0 [~nando@s83-179-44-135.cust.tele2.se] has joined #shogun | 15:06 | |
| -!- zxtx [~zv@c-24-18-130-24.hsd1.wa.comcast.net] has quit [Ping timeout: 256 seconds] | 15:39 | |
| -!- zxtx [~zv@c-24-18-130-24.hsd1.wa.comcast.net] has joined #shogun | 15:51 | |
| -!- n4nd0 [~nando@s83-179-44-135.cust.tele2.se] has quit [Quit: leaving] | 16:09 | |
| lambday | hello.. this is soumyajit.. I have been trying with string kernels... was using fisher's protein homology detection data (http://compbio.soe.ucsc.edu/discriminative/fisher-scop-data.tar.gz)... | 16:18 |
| lambday | got an error "ALPHABET does not contain all symbols in histogram"... even after removing the sequences which contained something other than [A-Z]... | 16:19 |
| lambday | was using PROTEIN alphabet... | 16:19 |
| lambday | when I ran it with ALPHANUM, it worked... but none of the sequences had any digits | 16:20 |
| lambday | and also, is the string subsequence kernel (http://jmlr.csail.mit.edu/papers/volume2/lodhi02a/lodhi02a.pdf or something similar) implemented in shogun? I was looking at the string kernels but couldn't find out :( | 16:28 |
| -!- sumit [73f91219@gateway/web/freenode/ip.115.249.18.25] has joined #shogun | 19:12 | |
| -!- sumit [73f91219@gateway/web/freenode/ip.115.249.18.25] has quit [Ping timeout: 245 seconds] | 19:16 | |
| -!- lambday [0e8b614f@gateway/web/freenode/ip.14.139.97.79] has quit [Quit: Page closed] | 19:17 | |
| -!- n4nd0 [~nando@s83-179-44-135.cust.tele2.se] has joined #shogun | 20:13 | |
| blackburn | hey n4nd0 | 20:25 |
| n4nd0 | blackburn: hi | 20:28 |
| n4nd0 | blackburn: how is it going? | 20:28 |
| blackburn | pretty fine | 20:28 |
| -!- nona994 [~nona@cpc5-haye17-2-0-cust251.haye.cable.virginmedia.com] has joined #shogun | 21:18 | |
| nona994 | Hi, I am desprate to work with WeightedDegreeRBFKernel but I can't figure its variables out. I went through its source code but it seems rather impossible to make a connection. | 21:25 |
| blackburn | nona994: hi nona | 21:29 |
| blackburn | nona994: what are your features? | 21:30 |
| blackburn | nona994: weighted degree rbf works with dense real-valued features only | 21:30 |
| nona994 | can you please explain what is a dense real-valued feature. I guess it's a matrix right? | 21:34 |
| blackburn | nona994: exactly | 21:34 |
| blackburn | just dense matrix | 21:34 |
| nona994 | n by n matrix right? | 21:35 |
| nona994 | these are arguaments for wd RBF 'int32_t size: is cache size, float64_t w: is ???, int32_t d: is degree for wd kernel, int32_t nof_prop: ??' | 21:37 |
| blackburn | nona994: no, n_features by n_vectors | 21:37 |
| blackburn | nona994: w is width (in means of exponential parameter coefficient) | 21:38 |
| blackburn | line 102 in .cpp | 21:38 |
| blackburn | nof_prop - 'number of properties' but not sure what it means | 21:39 |
| blackburn | I am afraid it is quite experimental | 21:40 |
| blackburn | if sonney2k was here he could tell us | 21:40 |
| nona994 | I see How can I pass a dense features | 21:40 |
| nona994 | Yes sure, he is the Oracle here | 21:42 |
| blackburn | nona994: is that a question about dense features? | 21:43 |
| nona994 | yes | 21:44 |
| nona994 | no | 21:44 |
| nona994 | I mean since you said it works with dense features I want to know how can I pass this dens feature | 21:45 |
| blackburn | I am confused | 21:45 |
| blackburn | well if you have real valued features just use it | 21:45 |
| blackburn | what is the data you are working with? | 21:45 |
| nona994 | I have protein sequences | 21:45 |
| nona994 | say they are strings | 21:46 |
| blackburn | so you have no real-valued features I guess | 21:46 |
| nona994 | nop | 21:46 |
| blackburn | then it looks quite impossible to use that kernel | 21:47 |
| nona994 | weighted degree (wd) works with strings and wd rbf include some specific measurement (real value) for each character or in my case amino acid | 21:48 |
| blackburn | in current implementation it just uses something similar to weighted degree kernel but for real values so I am unsure | 21:49 |
| nona994 | ok | 21:51 |
| nona994 | you are saying instead of strings it uses real-valued features and does the same computations and everything? | 21:53 |
| blackburn | nona994: yes | 21:53 |
| nona994 | so if I initiate it with a real-valued matrix as TRAINing set it has to work | 21:54 |
| nona994 | what is the difference between line 24 and 31 why dense features have l and r? | 21:56 |
| blackburn | nona994: treat it as rows and cols | 21:57 |
| blackburn | nona994: say l has 5 features | 21:57 |
| blackburn | err | 21:58 |
| blackburn | 5 vectors | 21:58 |
| blackburn | and r has 3 vectors | 21:58 |
| blackburn | then kernel matrix is 5 x 3 | 21:58 |
| nona994 | you mean dense features is 5 x 3 | 21:59 |
| blackburn | nona994: no, matrix of kernel between l and r is 5 x 3 | 21:59 |
| nona994 | i have got no idea about dense features :( | 22:10 |
| blackburn | nona994: well images, etc | 22:21 |
| blackburn | they are examples of dense features | 22:21 |
| nona994 | Are off to find them | 22:26 |
| nona994 | many thanks | 22:26 |
| nona994 | I am off to find them | 22:26 |
| -!- nona994 [~nona@cpc5-haye17-2-0-cust251.haye.cable.virginmedia.com] has quit [Quit: irc2go] | 22:58 | |
| -!- hoijui [~hoijui@dslb-092-078-043-220.pools.arcor-ip.net] has quit [Ping timeout: 244 seconds] | 23:41 | |
| --- Log closed Mon Feb 25 00:00:50 2013 | ||
Generated by irclog2html.py 2.10.0 by Marius Gedminas - find it at mg.pov.lt!