2
$\begingroup$

What are the advantages using Principal Components as input for neural networks compared to using "normal input" for neural networks?

$\endgroup$
  • $\begingroup$ a) variable selection to reduce overfitting (assuming largest PCs preserve 'signal' b) improved gradient descent (if sphered input data, ie covariance of input is identity) $\endgroup$ – seanv507 Dec 29 '16 at 11:34
1
$\begingroup$

You can read a nice description of PCA here.

The advantage of PCA (one method of dimension reduction) in the context of ML (ie. neural networks) would be to reduce the risk of over fitting and reduce computational complexity.

You could also phrase your question as: what are the alternatives to PCA? For PCA vs Partial Least Squares, see this Cross Validated post For PCA vs random subspace projection see this Cross Validated post

$\endgroup$
1
$\begingroup$

I think this question has been asked countless number of times here... You want PCA because you want to reduce your input size and thus improve training time and less likely to overfit.

$\endgroup$

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.