2
$\begingroup$

In the tutorial Parametric and Nonparametric Machine Learning Algorithms it says that parametric classifiers are faster than non-parametric classifiers. The reason that non-parametric classifiers are slower is because they often have far more parameters to train.

What else could cause this to be true?

$\endgroup$
  • $\begingroup$ Nearest neighbours is a non parametric classifier with essentially no training time. However at test time you have to find the nearest neighbour... $\endgroup$ – seanv507 Dec 18 '16 at 18:14
  • $\begingroup$ That appears to be somewhat oversimplifying.... Also the wording "non-parametric classifiers are slower is because they often have far more parameters" (emphasis mine) is a bit funny... In $k$-NN regression you learn a single parameter $k$, while in even a simple univariate least squares you learn two parameters (slope and intercept) so clearly it is not just the number of parameters you care about. $\endgroup$ – usεr11852 says Reinstate Monic Dec 18 '16 at 18:54
2
$\begingroup$

In a parametric classifier like ordinary least squares, a model is assumed. The training of the classifier does not need to learn the prediction model; only the coefficients for each term of the model. The assumed model could introduce bias depending on the skill of the modeler.

In nonparametric systems, neither the model nor the coefficients are assumed. Training involves the process of learning the other model and the coefficients.

$\endgroup$

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.