# Packages or libraries for multiple-output learning

I have a learning problem from $X$ to $Y$ where:

• $X$ = $n$ input numeric vectors of $m$ dimensions
• $Y$ = $n$ output numeric vectors of $k$ dimensions

In other words:

I am hoping to collect a list of R packages or Python libraries for multiple-output problems for classification and regression.

For example, do any of the learning methods in caret support this functionality? What packages in general are available for this problem?

-
Are all $k$ elements of $Y$ numeric? –  curious_cat Mar 6 '13 at 5:08
Sounds like a many-to-many neural network might be one option. –  curious_cat Mar 6 '13 at 14:44
@curious_cat. They are numeric. I just updated the OP –  Amelio Vazquez-Reina Mar 6 '13 at 15:03
Since I cannot comment, I am looking for a similar solution to my problem i.e. multiple output. Does anyone know of any packages in R which supports multi outputs? My two outputs have a significant covariance. I have been predicting them separately using multiregression, trees and random forests, but I wish to predict them together. I am aware that this is possible using PLS and ANN, but is there any package like caret which does the aforementioned in a faster, compact way? –  cmh325 Mar 27 at 5:43

## 2 Answers

I know of the PLS R-package, which support multi-response regression. See "The pls Package: Principal Component and Partial Least Squares Regression in R", Journal of Statistical Software, Vol. 18, Issue 2, Jan 2007 for more information.

There is some more information at http://mevik.net/work/software/pls.html

-

I recommend the following resources (for some of those, you would need to do some detective work, obviously). I agree with @Innuo that PLS regression and pls R package are worth the look.

For Python:

For R:

I think that the following resources might also be quite helpful: this relevant discussion on CV, this technical report on multi-output learning via spectral filtering, this page on Output Kernel Learning, and this excellent tutorial on multi-target prediction.

-