1
$\begingroup$

I have multiple predictors (ram, cpu, I/O etc.) (depended!) and looking for a model with multivariate outputs (I am trying to predict the same input fields given partial inputs). I have tried some models but cant find a model that will fit the data requirements. For example, lasso regression does not fit because it does not have a multi input-output version. Decision Tree regression does not fit because my inputs are dependent.. any suggestions?

Thanks in advance!

$\endgroup$
1
  • $\begingroup$ Lasso has the option Family="mgaussian". You can handle multivariate outputs with that. $\endgroup$
    – rook1996
    Jul 15 '18 at 9:59
0
$\begingroup$

You can use LASSO linear regression with multiple outputs (or with a multivariate output), simply writing $Y=X\beta+\varepsilon$ where $Y$ is a matrix such that each column of $Y$ corresponds to one output (or one coordinate of your multivariate output) and $\beta$ is a coefficient matrix whose columns correspond to coefficients giving the corresponding column of $Y$.

Or if you're in trouble with matrices, then simply fit one LASSO linear model per coordinate of your multivariate output.

$\endgroup$
1
  • $\begingroup$ all right! will try that $\endgroup$
    – Aline
    Jul 15 '18 at 11:11

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.