Generalization Gap
The difference between a model's empirical error on the training set and its true error on unseen test data. It reflects the model's ability to transfer from training data to new data; an excessively large gap typically indicates overfitting.
Timeline (last 90 days)
The correlation coefficient between the value of the generalization bound of Caro et al. and the actual generalization gap is only r = 0.12, indicating very weak predictive power.
The researchers argue that the validity of a generalization bound should not be equated with its predictive power, and the two need to be evaluated separately.
In quantum machine learning, because factors such as the number of trainable parameters, circuit entanglement structure, and measurement noise act together, estimating the generalization gap is often more complex than for classical models.
All Facts (3)
The correlation coefficient between the value of the generalization bound of Caro et al. and the actual generalization gap is only r = 0.12, indicating very weak predictive power.
50%UnverifiedThe researchers argue that the validity of a generalization bound should not be equated with its predictive power, and the two need to be evaluated separately.
50%UnverifiedIn quantum machine learning, because factors such as the number of trainable parameters, circuit entanglement structure, and measurement noise act together, estimating the generalization gap is often more complex than for classical models.
50%