This is my blog.
正则化，一开始我看英文都没有想中文是啥
嘿嘿嘿，被嘲笑了（不对，应该是嘤嘤嘤
就大概是将值限定在[-1,1]，不会因为取值而影响权重吧
然后也防止特征太多，导致这条线歪歪扭扭，缺少泛化性！
Lesson 7 Solving the problem of overfitting
“underfitting” also calls “high bias”
“overfitting” also calls “high variance “
So, how to solve the problem? There are two main methods.
Ruduce the number of features
- Manually select which features to keep.(delete the features that doesn’t matter)
- Use a model selection algorithm
Regularization
- Keep all the features, but reduce the magnitude of parameters θj.
- Regularization works well when we have a lot of slightly useful features.
转载请注明出处，谢谢。
愿 我是你的小太阳
