Glossary term
Glossary term
Evaluation and Benchmarks
The complexity of problems that a model can learn. The more complex the problems that a model can learn, the higher the model's capacity. A model's capacity typically increases with the number of model parameters. For a formal definition of classification model capacity, see VC dimension.
Created for this library
A research team increases model capacity until the loss curve no longer improves to find the right size for its dataset.
An ML platform team caps model capacity for on-device deployment based on latency and battery constraints.
A startup picks model capacity based on the inference budget rather than maximum accuracy because cost per call dictates production design.
Definition source: Google for Developers Machine Learning Glossary | Creative Commons Attribution 4.0 License