Font size
WorksheetsRidge, Lasso and Polynomial Regression
Total questions: 20
Worksheet time: 7mins
Ridge regression is mainly used to handle:
Non-linear relationships
Overfitting caused by multicollinearity
Outliers
Missing values
The penalty term in Ridge regression is based on:
Combination of L1 and L2
L1 norm (absolute values)
L2 norm (squared values)
No penalty
(a) regression can shrink some coefficients to zero, effectively performing feature selection.
(give answer in small case alphabet)
Adding higher-degree terms in polynomial regression can lead to:
(ans - give all character in small case)
(a)
(a) regression method cannot reduce coefficients to exactly zero.
In Ridge regression, as the penalty parameter (λ) increases, coefficients tend to:
Increase
Stay the same
Become negative
Shrink towards zero but never reach it
(a) regression is most suitable when you suspect that only a few features are important.
(a) regression method you use if you want to keep all features but reduce their effect.
Which parameter controls the strength of regularization in Ridge and Lasso regression?
α (alpha)
β (beta)
γ (gamma)
η (eta)
Lasso regression adds a penalty equal to:
Product of coefficients
Sum of squared coefficients
Sum of absolute coefficients
Log of coefficients
Which regression is useful when predictors are highly correlated? (a)
Polynomial regression models are linear in terms of:
Features
Both coefficients and features
Neither
Coefficients
The regularization parameter (α) in Ridge and Lasso controls:
Learning rate
Model accuracy
Penalty strength
Data normalization
In Ridge regression, large coefficients are:
Doubled
Ignored completely
Reduced in magnitude
Set to zero
Which regression type tends to produce smoother curves in high-degree models?
(a)
Lasso regression is useful for:
Removing correlated features
Feature selection
Visualization
Handling missing values
A disadvantage of Lasso is:
Cannot perform feature selection
May randomly drop important correlated variables
Cannot handle non-linear data
Too slow for small datasets
Ridge regression generally performs better than Lasso when features are (a) correlated.
(weakly / highly)
In Lasso regression, very large α results in:
All coefficients becoming large
No effect on coefficients
Many coefficients becoming exactly zero
Polynomial terms being added
Which regression can create curved decision boundaries by feature transformation?
(a)
