Large bias:
Model cannot even fit the training examples => Underfitting
For bias,redesign your model:
- Add more features as input
- A more complex model
Large variance:
Fit the training data,but large error on testing data => Overfitting
For variance:
- More data (Very effective but not always practical)
- Regularization(smooth, bias increases)
There is usually a trade-off between bias and variance.
Model Selection
What you should NOT do:
Cross Validation
use validation set to select model
少在意model在public testing set 的表现
