Please enable JavaScript.

Coggle requires JavaScript to display documents.

Overfitting (preventing overfitting (general (L2 Norm: sum of the squares…

- - - - iterates training and testing
      - compare fold accuracies between models
  - - - fitting graph plots against compexity
  - - - how few instances are willing to tolerate a leaf?
      - keep the tree from getting to big
  - - - use nested holdout projecedure to first pick the best individual feature
    - - functions fit better with larger weights
      - incorp L2 into standard least squares inear regression
        
        get ridge regression
        
        Use L1 Norm (sum of abs. values), get lasso.
      - near equivalent to support vector machine learning
        
        difference: uses hinge loss instead of likelihood in its optimization
- - - - can be complex in different ways
- - - - compare predicted values with hidden true values