1

I am new to the field of machine learning and am just learning linear regression, and I have some questions about this concept:

Does linear regression allow vector-valued target variables?

Does linear regression not assume that the features are uncorrelated?

Samir
  • 113
  • 1
    Your second question is addressed here; there is no such assumption. – Dave Jun 06 '22 at 22:32
  • 1
    The comment by @Dave answers your second question. Comverning the second question: regression with vector valued targets is called "multivariate regression". For linear regression, it essentially boils down to training a model for each component of the target variable. – cdalitz Jun 07 '22 at 04:43

1 Answers1

1

Does linear regression allow vector-valued target variables?

You can formulate that way. It'll be a parallel set of equations, $y=X\beta$, where $\beta$ is of size $f\times t$ ($f$ is number of features and $t$ is number of targets).

Does linear regression not assume that the features are uncorrelated?

I believe @Dave's comment and the associated post clears this question; but to reiterate, features can be correlated or uncorrelated. Linear regression does not assume anything.

gunes
  • 57,205