5. Size of the Dev and Test sets
2023. 9. 14. 12:36ㆍGoogle ML Bootcamp/3. Structuring Machine Learning Projects
Train / test : 7/3
Train / Valid / Test : 6/2/2
if dataset >= 1,000,000
- 98/1/1
- 1% = 10,000 example. enough!
No test set might be okay..
- if your dev set size is too big
- else, overfitting to dev
- 일반적으로 추천하진 않는다.
'Google ML Bootcamp > 3. Structuring Machine Learning Projects' 카테고리의 다른 글
| 7. Why Human-level Performance? (0) | 2023.09.14 |
|---|---|
| 6. When to Change Dev/Test sets and Metrics? (0) | 2023.09.14 |
| 4. Train/Dev/Test Distributions (0) | 2023.09.14 |
| 3. Satisficing and Optimizing Metric (0) | 2023.09.14 |
| 2. Single Number Evaluation Metric (0) | 2023.09.14 |