ChanServ changed the topic of #mlpack to: "mlpack: a fast, flexible machine learning library :: We don't always respond instantly, but we will respond; please be patient :: Logs at http://www.mlpack.org/irc/
vansika__ has quit [Ping timeout: 265 seconds]
vansika__ has joined #mlpack
< jonpsy[m]>
kartikdutt18: yo, you there?
< kartikdutt18[m]>
Hi jonpsy , Here now.
< kartikdutt18[m]>
DatasetX/Y are just templates for arma types
< jonpsy[m]>
What else are expecting besides arma::mat
< jonpsy[m]>
And your thoughts on 2.?
< kartikdutt18[m]>
Makes sense.
< kartikdutt18[m]>
The field can also can contain both vectors and matrices I guess.
< jonpsy[m]>
Aye, that's true..
< jonpsy[m]>
We can template the choice b/w vectors and matrices. BTW are you sure it's vector and not row<>?
< jonpsy[m]>
* kartikdutt18we we can template the choice btw vectors and matrices. BTW are you sure it's vector and not row<>?
< jonpsy[m]>
(Please tag me so i get notification :) )
dendre has quit [Read error: Connection reset by peer]
< jonpsy[m]>
<kartikdutt18[m] "The field can also can contain b"> You're using DataSetX& to toggle between field<vector> or field<mat> that's what you're saying?
< jonpsy[m]>
Field :vec: and field :mat:
< kartikdutt18[m]>
I meant field can contain any object (mat, vec) etc.
< jonpsy[m]>
Yes
< jonpsy[m]>
So that's what DatasetX is for?
< kartikdutt18[m]>
It's just a template can be arma::mat, arma::field or any arma type
< jonpsy[m]>
I see.
< jonpsy[m]>
Okay, I'll push soon with the updates lemme know your thoughts on the PR.
vansika__ has quit [Ping timeout: 260 seconds]
vansika__ has joined #mlpack
< kartikdutt18[m]>
Great. Looking forward to it.
vansika__ has quit [Read error: Connection reset by peer]
vansika__ has joined #mlpack
< jonpsy[m]>
kartikdutt18: , hi sorry for the bother. I've noticed field can put vector in multiple columns or in a single column. How would I know which type the user passed in?
ImQ009 has joined #mlpack
< jonpsy[m]>
is this the correct way? or ```trainData(idx,0)```, sorry again :D
< jonpsy[m]>
or even trainData[idx] for that matter
ImQ009 has quit [Read error: Connection reset by peer]
ImQ009 has joined #mlpack
< kartikdutt18[m]>
Number of cols shouldn't be an issue, Since the way I used field type is each column corresponds to a vector having different shape / elements. The implementation should (atleast that's why I think) split number of columns (Irrespective what each column holds). For more info on field type you can [use armadillo's documentation](http://arma.sourceforge.net/docs.html#field). The difference here is that you can't use
< kartikdutt18[m]>
insert, you need to pre allocate a field type and fill each field column
< jonpsy[m]>
ok , so each field column is a vector right? that's the standard we're following?
< kartikdutt18[m]>
Each field column is vector where I use it but it can be anything. I don't think it would be different if each column held mat but I'm fine with a field vec type implementation as well.
< jonpsy[m]>
hmm, maybe we can template that if the implementation is same
< kartikdutt18[m]>
Agreed.
< jonpsy[m]>
and check std::enable_if to only have vec and mat for input
< kartikdutt18[m]>
Sounds good. :)
< AbishaiEbenezer4>
Hi mlpack!
< AbishaiEbenezer4>
I wanted to know about the naive bayes classifier.
< AbishaiEbenezer4>
I see that the gaussian classifier has been implemented, but i don't think multinomial and bernoulli types have been implemented, although they are quite simple.
< AbishaiEbenezer4>
could i please have any clarifications on this?
< rcurtin>
AbishaiEbenezer4: you're right, not sure what to add to that
< rcurtin>
if you'd like to implement those other types, it would be a nice contribution
< rcurtin>
I would suggest finding a way to do it that maximizes code reuse among the different distribution types
< rcurtin>
so, ideally, NaiveBayesClassifier gets an extra template parameter specifying the distribution type---and everything else about the external API stays the same
< rcurtin>
but, I'm not sure if you were looking to contribute, so if you're not, no worries :-D