choosing a value of $\lambda$ for logistic regression with regularisation

Question

I understand the $\lambda$ term is used to avoid an overfitting in many models, including logistic regression. Can you help me how to choose which $\lambda$ to use and what will be its corresponding interpretation?

Thanks

Answer 1

It is the weight that we placed on "simplicty", the bigger it is, the less we focus on training accuracy.

We usually tune it, we first prepare the training set and validation set. We then try various values of $\lambda$ and choose the version that has the best performance on the validation set. Alternatively, we use $k$-fold cross validation.