verne.freenode.net changed the topic of #mlpack to: http://www.mlpack.org/ -- We don't respond instantly... but we will respond. Give it a few minutes. Or hours. -- Channel logs: http://www.mlpack.org/irc/
< ironstark>
zoq: rcurtin: The output generated from the R script is of the type
< ironstark>
here is something I have cooked up. It works under the assumption that the very last decimal number parsed is the runtime
< ironstark>
does that seem correct?
mikeling has joined #mlpack
partobs-mdp has joined #mlpack
govg has quit [Quit: leaving]
govg has joined #mlpack
partobs-mdp has quit [Remote host closed the connection]
vivekp has quit [Ping timeout: 240 seconds]
vivekp has joined #mlpack
shikhar has joined #mlpack
kris1 has joined #mlpack
Erwan_ has joined #mlpack
kris1 has quit [Read error: Connection reset by peer]
kris1 has joined #mlpack
< zoq>
ironstark: Looks good, I would also include 'sec elapsed', since the pattern '\d+\.\d+' might match multiple times? So something like: "(\d+\.\d+).*sec elapsed"
< Erwan_>
rcurtin: do you use something special in your main code for deserializing ?
< Erwan_>
From further investigation, it seems that the problem I have come from trying to read the xml into a boost archive
< Erwan_>
It expects some tags to be there, like "<count></count>" for arrays
< Erwan_>
For every other data structure it's fine
< Erwan_>
but for the gmm model, it is replaced by the tag "<gaussians></gaussians>"
< Erwan_>
We boost tries to deserialize, it is unable to find the count tag, and thus throw the exception
< Erwan_>
If I manually add the tag, it then expect a <item_version> tag and so on
< Erwan_>
(btw if we exclude the serialization of the map of gmm, it serialize and deserialize fine)
kris1 has quit [Quit: kris1]
kris1 has joined #mlpack
kris1 has quit [Quit: kris1]
kris1 has joined #mlpack
kartik_ has joined #mlpack
< kartik_>
hi <zoq> i have completed the blog post and added the vanilla network on thyroid dataset. but it takes 4 mins to converge. Is it okay?
< zoq>
kartik_: 4 min on travis? Might be a a little bit long, we should see if we can get the test time down, instead of using the thyroid we could test iris.
kartik__ has joined #mlpack
< kartik__>
<zoq> no 4 mins on local machine
< zoq>
hm, okay, let us test the iris dataset
kartik_ has quit [Ping timeout: 260 seconds]
< kartik__>
<zoq> should i test it on iris and remove thyroid ?
< kartik__>
<zoq> about CNE what start pointers would u suggest?
< zoq>
we have to check the timings, but I guess we could delete the thyroid dataset
< zoq>
you mean NEAT?
< kartik__>
<zoq> yes NEAT
< zoq>
So I would start with the IsSameSpecies function including Disjoint and WeightDiff.
< zoq>
It looks like after some time there is only one species left.
< kartik__>
ohkay.
< zoq>
Where every Genome is the same ...
< zoq>
So my guess, the comparing function isn't correct or doesn't use the correct parameter.
< zoq>
As you already said, writting some debug functions might be helpful in the long run :)
< kartik__>
ohkay .. ill check that
< kartik__>
yes i will .. open a new PR and start with the functions
< zoq>
Okay, sounds good, let me know if you need any help.
< kartik__>
yes <zoq> i will go for the regular string stream and then saving it in a file
< kartik__>
in a specified format
< zoq>
You could also use the mlpack serialization feature.
< kartik__>
u said before to use serialization library
< kartik__>
yes exactly :D
< kartik__>
any start pointers for that
< kartik__>
?
< zoq>
Let's see if I can find the NEAT seralization code
< kartik__>
oh okae.. i will continue for 2-3 days reading the NEAT the papers and sethbling code and then will start with the feature.
< zoq>
kartik__: Okay, that gives me some time to find the code :)
< zoq>
ah, found it
< zoq>
Will push the code later to my fork
< kartik__>
on the way.. ill write the tutorial for it as well ..thanks a lot marcus.
< zoq>
okay, great
kartik__ has quit [Quit: Page closed]
travis-ci has joined #mlpack
< travis-ci>
mlpack/mlpack#3120 (master - c39a45b : Ryan Curtin): The build was broken.
mikeling has quit [Quit: Connection closed for inactivity]
shikhar has quit [Quit: WeeChat 1.7]
kris_ has quit [Quit: Connection closed for inactivity]
< zoq>
rcurtin: The last commit reminds me of something Kartik noticed a while ago, that some formulas aren't rendered e.g http://www.mlpack.org/docs/mlpack-2.2.4/doxygen/fmkstutorial.html right after: "algorithm which finds". haven't looked into the issue yet...
travis-ci has joined #mlpack
< travis-ci>
mlpack/mlpack#3125 (master - 39f9999 : Ryan Curtin): The build was fixed.