Training Error
Training ErrorIt is the error that occurs in data training, which can be regarded as the average loss of the model on the training data.
The difference between training error and other errors
- Training error: The error of the model on the training set
- Validation error: The error of the model on the validation set
- Test error: The error of the model on the test set
- Generalization error: measures the generalization of the model
- In general, the relationship between them is: training error < validation error < test error ≒ generalization error
The role of training error
- Training set: used to fit the model, compare the size of the training error, and obtain the model parameters;
- Minimum training error: used to find a specific model and fit related parameters;
- Test set: used to evaluate the model. By calculating the deviation and variance of the model on the test set, the test error can be minimized.
- Summary: For the training error, we do not consider the bias and variance, but only need to minimize the training error.