- [[lasso feature selection]] # Idea Feature selection can be used to reduce variance and overfitting (see [[bias-variance tradeoff]]). Stepwise regression - computationally efficient but it's a greedy algorithm - locally optimal - globally suboptimal [[lasso feature selection]] - # References - [Algorithms for automatic model selection - Cross Validated](https://stats.stackexchange.com/questions/20836/algorithms-for-automatic-model-selection) - [feature selection - What are the advantages of stepwise regression? - Cross Validated](https://stats.stackexchange.com/questions/218208/what-are-the-advantages-of-stepwise-regression/218223#218223)