You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Hello ,yes the development was stopped long time ago. This is because the original toolbox is not able to handle big ranges see here.
So, I no longer use FEAST for feature selection, even using the new update with integer discretization. I have tested on same dataset which is available on the issue, but even the new update doesn't work for my data.
Anyway, In my opinion a fast solution without modify many routines is implement a key-value based discretization on the discretize function of the FEAST library instead the current method that relies on reserve a vector with the length of the fully variable range. Key-value seems to be less memory hungry but more cpu intensive and more ram accesses.
Until now I'm working fully dedicated to my phD, this is why there is no changes on this repo since the first commit.
So, I no longer use FEAST for feature selection, even using the new update with integer discretization. I have tested on same dataset which is available on the issue, but even the new update doesn't work for my data.
hi,
i was just linked to your project. i see that development seemed to have stopped one year ago?
or is this still active?
i am the main developer of mlr here
https://github.com/mlr-org/mlr
and we would be interested to support FEAST in mlr as a fast and flexible way to perform feature filtering.
how much would be missing to release this in CRAN?
@smilesun @ajing
The text was updated successfully, but these errors were encountered: