Skip to content

Generalization

To measure the generalization, we define a generalization error1,

G=LP(f^)LE(f^),

where LP is the population loss, LE is the empirical loss, and f^ is our model by minimizing the empirical loss.

However, we do not know the actual joint probability p(x,y) of our dataset xi,yi. Thus the population loss is not known. In machine learning, we usually use cross validation where we split our dataset into train and test dataset. We approximate the population loss using the test dataset.


  1. Roelofs R. Measuring generalization and overfitting in machine learning. 2019.https://escholarship.org/uc/item/6j01x9mz


Contributors: LM