# Coursera吴恩达《构建机器学习项目》课程笔记（2）-- 机器学习策略（下）

### 1. Carrying out error analysis

• Fix pictures of dogs being recognized as cats
• Fix great cats(lions, panthers, etc…) being misrecognized
• Improve performance on blurry images

### 2. Cleaning up incorrectly labeled data

• Overall dev set error: 10%
• Errors due incorrect labels: 0.6%
• Errors due to other causes: 9.4%

• Overall dev set error: 2%
• Errors due incorrect labels: 0.6%
• Errors due to other causes: 1.4%

• Apply same process to your dev and test sets to make sure they continue to come from the same distribution
• Consider examining examples your algorithm got right as well as ones it got wrong
• Train and dev/test data may now come from slightly different distributions

### 3. Build your first system quickly then iterate

• Set up dev/test set and metric
• Build initial system quickly
• Use Bias/Variance analysis & Error analysis to prioritize next steps

### 5. Bias and Variance with mismatched data distributions

• Carry out manual error analysis to try to understand difference between training dev/test sets
• Make training data more similar; or collect more data similar to dev/test sets

### 7. Transfer learning

• Task A and B have the same input x.
• You have a lot more data for Task A than Task B.
• Low level features from A could be helpful for learning B.

\frac1m\sum_{i=1}^m\sum_{j=1}^cL(\hat y_j^{(i)},y_j^{(i)})

L(\hat y_j^{(i)},y_j^{(i)})=-y_j^{(i)}log\ \hat y_j^{(i)}-(1-y_j^{(i)})log\ (1-\hat y_j^{(i)})

• Training on a set of tasks that could benefit from having shared lower-level features.
• Usually: Amount of data you have for each task is quite similar.
• Can train a big enough neural network to do well on all the tasks.

### 10. Whether to use end-to-end deep learning

end-to-end深度学习有优点也有缺点。

• Let the data speak
• Less hand-designing of components needed

• May need large amount of data
• Excludes potentially useful hand-designed

0 条评论

## 相关文章

21660

41090

43490

26170

40080

### 学界 | 卷积神经网络做简单任务原来有 BUG？UBER AI Lab 来支招

AI 科技评论按：UBER AI Lab 最新研究发现卷积神经网络在处理坐标变换问题上存在惊人的「无能」缺陷，并提出了一种简单的 CoordConv 策略修复了...

8910

1.5K40

743110

15020

35970