
上QQ阅读APP看书,第一时间看更新
Regularization parameters in linear regression and ridge/lasso regression
Adjusted R-squared in linear regression always penalizes, adding extra variables with less significance is one type of regularizing the data in linear regression, but it will adjust to the unique fit of the model. Whereas, in machine learning, many parameters are adjusted to regularize the overfitting problem. In the example of lasso/ridge regression penalty parameter (λ) adjusted to regularization, there are infinite values that can be applied to regularize the model in infinite ways:


Overall, there are many similarities between the statistical way and machine learning way of predicting the pattern.