“This is the 16th day of my participation in the Gwen Challenge in November. Check out the details: The Last Gwen Challenge in 2021.”
After the deep learning algorithm model is established, the predicted results will often be over-fitting and over-fitting. In this case, we need to process the data to make the final results appropriate.
Overfitting processing
1. Get more training data
Using more training data is the most effective means to solve the over-fitting problem, because more samples can make the model learn more and more effective features and reduce the influence of noise.
2. Dimensionality reduction means discarding features that don’t help us predict correctly. You can manually select which features to keep, or you can use some model selection algorithm to help (such as PCA).
Regularization regularization technology retains all the features, but reduces the magnitude of the parameter, which can improve or reduce the overfitting problem.
4. Integrated learning Method Integrated learning is to integrate multiple models together to reduce the risk of overfitting a single model
Underfitting processing
1. Add new features
When features are insufficient or the correlation between existing features and sample labels is not strong, the model is prone to underfitting. By mining new features such as combination features, better results can often be achieved.
2. Increase model complexity
Simple models have poor learning ability, so the model can have stronger fitting power by increasing the complexity of the model. For example, high-order terms are added to linear models, and the number of network layers or neurons is added to neural network models.
3. Reduce the regularization coefficient
Regularization is used to prevent over-fitting, but when the model is under-fitting, it is necessary to reduce the regularization coefficient.
