Regularization
Regularization is one way of reducing the total error, by finding a good trade-off between bias and variance automatically. To do so, we simply add another term to the loss function, which will try to minimize the complexity of the model. The importance of this additional term can be controlled by a scalar, which is an example of a hyperparameter. In this video, we discuss the two most prominent methods, ridge, and lasso.
Watch video Regularization online without registration, duration hours minute second in high quality. This video was added by user Computational Thinking 16 November 2022, don't forget to share it with your friends and acquaintances, it has been viewed on our site 1,378 once and liked it 48 people.