The Lambda (๐) Parameter
The Lambda (๐) Parameter also known as the regularization rate is used to tune the overall impact of regularization on the complexity error. I have already covered the complexity error and regularization in one of my previous articles.
The idea behind this is that we use this parameter to penalize the complexity error according to the problem statement we have at hand. In case if we have a problem from a medical or defense domain then we are okay to have a complex model and hence we would want to give less punishment to the complexity error and on the other hand, if we have a recommendation engine to be built we would want the model to be less complex (more simple) and hence the punishment on complexity error should be large.
This punishment for complexity error is the lambda (๐) parameter or the regularization rate.
