0
$\begingroup$

In the process of artificial data synthesis in Machine Learning applications, when we introduce noise to generate new data, is it possible that overfitting can occur because we are essentially creating variations of the same data example.

$\endgroup$

1 Answer 1

0
$\begingroup$

That's quite the opposite. By providing slightly different representations of the same data, you are forcing the net to learn a single representation for all the possibile variations, thus improving generalization.

$\endgroup$
2
  • $\begingroup$ But unless we perform the variations on all the datasets, it would cause bias wouldn't it? $\endgroup$ Jun 20, 2017 at 6:14
  • $\begingroup$ Which datasets? Can you be more precise? $\endgroup$
    – mp85
    Jun 20, 2017 at 6:16

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.