[machine learning notes] how to solve over fitting and under fitting

1. How to solve the problem of over fitting ：

1. simplified model ： Models with fewer parameters can be selected , It can also reduce the number of attributes in the training data , Or constraint model
2. Collect more training data
3. Reduce the noise in the training data （ for example , Fix data errors and eliminate outliers ）

2. How to solve the problem of under fitting ：

1. Choose one with more parameters , A more powerful model
2. Provide better feature sets for learning algorithms
3. Reduce constraints in the model （ for example ： Reduce regularization hyperparameters ）