Glossary term
Glossary term
Training and Fine-Tuning
See pooling.
Created for this library
A computer vision team uses subsampling in its CNN to reduce spatial dimensions through pooling layers.
A research team subsamples training data per epoch to keep training time bounded on very large datasets.
An NLP team subsamples frequent tokens during training so the model is not dominated by stop words.
Definition source: Google for Developers Machine Learning Glossary | Creative Commons Attribution 4.0 License