Difference between revisions of "L1 and L2 Regularization"
| Line 9: | Line 9: | ||
* add more data | * add more data | ||
* use [[Data Augmentation]] | * use [[Data Augmentation]] | ||
| − | * use [[Batch | + | * use [[Batch Norm(alization) & Standardization]] |
* use architectures that generalize well | * use architectures that generalize well | ||
* reduce architecture complexity | * reduce architecture complexity | ||
Revision as of 18:32, 2 January 2019
Youtube search... ...Google search
Mathematically speaking, L1 is just the sum of the weights as a regularization term in order to prevent the coefficients to fit so perfectly to overfit. There is also L2 regularization. where L2 is the sum of the square of the weights.
Good practices for addressing the Overfitting Challenge:
- add more data
- use Data Augmentation
- use Batch Norm(alization) & Standardization
- use architectures that generalize well
- reduce architecture complexity
- add Regularization
- L1 and L2 Regularization - update the general cost function by adding another term known as the regularization term.
- Dropout - at every iteration, it randomly selects some nodes and temporarily removes the nodes (along with all of their incoming and outgoing connections)
- Data Augmentation
- Early Stopping
