What is RandAugment and why is it a nice augmentation policy?
mediumAnswer
- Instead of learning an augmentation policy (AutoAugment), RandAugment randomly picks N transforms from a fixed pool (rotate, shear, color jitter, ...) each with a shared magnitude M.
- Two hyperparameters (N, M) instead of dozens — much easier to tune, and matches AutoAugment's accuracy on ImageNet.
- TrivialAugment (2021) simplifies further: apply one random transform with a random magnitude.
Check yourself — multiple choice
- Learned RL-based augmentation
- N random transforms with shared magnitude M — 2 hyperparameters, matches AutoAugment accuracy
- Static single augmentation
- Only for language
RandAugment: 2 knobs (N, M), matches learned policies — simple and effective.
#augmentation
Practise Deep Learning
214 interview questions in this topic.