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/
< rcurtin[m]>
I would agree with that. I am not looking forward to learning about UEFI, for instance...
< shrit[m]>
maybe it would be better to acquire an old server hardware with two or more XEON on one motherboard with 64 GB RAM with SSD that should me enough for next years, but I do not know a lot of places where this kind of hardware can be found even if it is 10 years old back
< shrit[m]>
be
< RishabhGarg108Gi>
Hii all, I was writing a test case in which I have to check for the equality of two armadillo vectors. Here the vectors can have `arma::datum::nan`. So this is making it complicated to check for equality because a `nan` is not equal to anything (not even itself). Can someone suggest how to check the equality of two such vectors. Thanks! :)
ImQ009 has joined #mlpack
< NippunSharmaGitt>
@RishabhGarg108 you can replace nan with a large value and then check the equality
< NippunSharmaGitt>
You can also drop nan values altogether and then check afterwards
< NippunSharmaGitt>
But i think dropping will not be possible so former one would be the better choice
< RishabhGarg108Gi>
@NippunSharma , I did try to replace those null values with infinity, but I was still not able to do it. I will give it a try once again, but I am not sure that would work and also I don't think that it is the most efficient way of doing it.
< RishabhGarg108Gi>
One way I was thinking of doing it is that if I can somehow recognise if a particular value is a null value, for example lets say a vector `V` has null value at index `i` and I could do `isnull( V(i) )`, then that would definitely solve the problem here.
< RishabhGarg108Gi>
It turns out that armadillo does it in some way because we have a `has_null` method for matrices that returns true if a matrix contains any null value
< NippunSharmaGitt>
@RishabhGarg108 instead of replacing with infinity try replacing with `std::numeric_limits<T>::max()` where T is the `double` or `int` type
< NippunSharmaGitt>
Does mlpack have the support to map non-numeric variables to numeric values inside a pandas dataframe automatically or do we have to specify some parameter ?
ashwin has joined #mlpack
ashwin has quit [Remote host closed the connection]
< rcurtin[m]>
Nippun Sharma (Gitter): to go from a Pandas dataframe to an arma::mat mapping categorical variables is done; see `to_matrix()` in `src/mlpack/bindings/python/`; but there are no bindings that *return* a categorical matrix, so the other direction (arma::mat with categorical variables to pandas dataframe) has never needed to be implemented
< NippunSharmaGitt>
Thanks @rcurtin
< NippunSharmaGitt>
would it be better to have a parameter that uses `to_matrix()` or `to_matrix_with_info()` that automatically does the mapping instead of manually calling the utility function ?
< rcurtin[m]>
I don't understand what you mean
< rcurtin[m]>
the utility function is not manually called; it is automatically called from the generated bindings
< NippunSharmaGitt>
So here the problem is that the dataframe contains a categorical feature inside and an error is thrown due to this
< NippunSharmaGitt>
I think this error is because `to_matrix()` is called in backend here
< NippunSharmaGitt>
whereas `to_matrix_with_info()` can be used to map the categorical features inside the dataframe
< NippunSharmaGitt>
So it might be better if we give the user a choice to call between `to_matrix` and `to_matrix_with_info` ? It might also be that I am missing something here ?
ImQ009_ has joined #mlpack
ImQ009 has quit [Ping timeout: 260 seconds]
< abernauer[m]>
Belated Happy Holidays to everyone!