Data Augmentation
Data augmentation is expanding a training dataset by creating modified copies of existing examples, such as rotated or cropped images or paraphrased sentences.
What is Data Augmentation?
Augmentation improves generalisation and robustness, especially when data is limited, by showing the model plausible variations.
How it works
Transformations are applied randomly to examples during training so the model rarely sees exactly the same input twice.
← All Artificial Intelligence & Superintelligence definitions · ← Back to the A–Z definitions