# Full Stack Deep Learning
## Underfitting
- Bigger model
- Reduce regularization
## Overfitting
- More data, augmentation
- Dropout, early stopping
## Production
- Data pipeline
- Training, deployment
## Best Practices
- Error analysis
- Hyperparameter tuning
Full Stack Deep Learning
Full Stack Deep Learning (fullstackdeeplearning.com) — Notes for week 1 to week 12 (2021)
Reference: https://fullstackdeeplearning.com/spring2021
Underfitting vs Overfitting
Underfitting (reducing bias):
- Bigger model
- Reduce regularization
- Error analysis
- Different model architecture
- Tune hyper-parameters
- Add features
Overfitting (reducing variance):
- Add more training data
- Add normalization (batch norm, layer norm)
- Add data augmentation
- Increase regularization (dropout, L2, weight decay)
- Error analysis
- Choose different model architecture
- Tune hyper-parameters
- Early stopping
- Remove features
- Reduce model size
#StackDeepLearning #computervision #AI #deeplearning