|
Home
|
- Overfitting is your worst enemy
- Train, Test (Quiz), Validate
- Out-of-sample in the real world is subtle
- Loss functions
- Feature generation is critical
- Think about the problem!!
- How might you transform the features?
- Do you want a scale-invariant method or not?
- What else could you measure?
- Is semi-supervised learning possible?
- Are there surrogate labels you might use?
- Feature Blocks
- Different feature sets need different regularization
- One solution: block-stagewise regression
- Combinations of multiple methods (“ensemble methods”) are usually the most accurate
- Missing data
- missing at random or not requires different imputation
- Explanation/Insight is often important
- visualization: word clouds, PCA, MDS
- MDS: given an {$n x n$} matrix of distances between points, find a new (usually 2-D) representation of each of the points that as closely as possible preserves that distance matrix
- Look at the data
- posts, images scoring highest in some feature or outcome
- variable importance
- How “important” is each feature for the prediction?
- Correlation is not causality
Back to Lectures
|
