I have multivariate variables that I want to regress to a single target label. For some reason, using partial least squares regression (projected to a single component) gives much better prediction than ordinary least squares regression. I have made sure that the input is scaled the same in both methods. But since for my problem I am limited to a neural network in Tensorflow, how can I make a neural network replicate the result from plsregression computed with NIPALS?

  • $\begingroup$ I am not sure if that is possible in Tensorflow using standard functions, but did you look at the steps of NIPALS algoritm? It is quite simple. In fact, if you are using a single component it is even easier. $\endgroup$ – theGD May 9 at 11:22

Your Answer

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

Browse other questions tagged or ask your own question.