3. Basic recipe for machine learning
2023. 9. 10. 15:46ㆍGoogle ML Bootcamp/2. Improving Deep Neural Networks
1. check high bias ?
- training data performance
- and then, more iteration or change to bigger network
2. check high variance ?
- dev set performance
- more data always work
- Regularization or chage to bigger network
1-2를 반복.
딥러닝 시대 이전. "bias vs variance trade off"
- 빅데이터 + 딥러닝모델 시대에서는 bias를 크게 해치지 않으면서 variance를 줄일 수 있다!
'Google ML Bootcamp > 2. Improving Deep Neural Networks' 카테고리의 다른 글
6. Dropout Regularization (0) | 2023.09.10 |
---|---|
5. Why Regularization Reduces Overfitting (0) | 2023.09.10 |
4. Regularization (0) | 2023.09.10 |
2. Bias / Variance (0) | 2023.09.10 |
1. Train / Dev / Test sets (0) | 2023.09.10 |