Model capacity Machine Learning
noun phrase
Definition: The complexity of problems a model can learn; capacity generally increases with the number of parameters [Google Machine Learning Glossary].
Example in context: “While data augmentation is flexible, easy to implement and can be combined with other approaches, the DNN requires additional model capacity to capture the desired symmetries, which in addition are only approximated rather than guaranteed.” [Rath, Condurache 2024]
Related terms: VC dimension, underfitting, overfitting