1. Bias Introduction: While Lasso can reduce variance, it can also introduce bias into the model, potentially leading to less accurate predictions if the true relationship between the predictors and the outcome is complex. 2. Collinearity Issue: Lasso may struggle with collinear features (features that are highly correlated). It tends to arbitrarily select one feature from a group of highly correlated features, which might not be ideal in some cases. 3. Tuning Challenges: Selecting the optimal \( \lambda \) parameter can be challenging and typically requires cross-validation, which is computationally intensive.
