Bayesian Linear Regression youtube video
- Why not use MLE? Overfitting!
- How to solve?
- Why not use MAP? NO representation of our uncertenty in w and Y
![MLE](https://i-blog.csdnimg.cn/blog_migrate/76149467b967901e19ee9e46237ab8c0.png)
![](https://i-blog.csdnimg.cn/blog_migrate/b4f6c7ff75fd27e3de893efe205d603b.png)
- Why Bayesian? Optimize loss fn.
- predicive distribution: Given us
=what we really want
- predicive distribution: Given us
Setup problem
Model: indep given
,
means variance
.
Assume a, b are known
Remark, repalce by
---> Non-linearilites in x by basis function
Likelihood:
其中, "Design Matrix"