参考
http://www.cnblogs.com/jerrylead/archive/2011/05/11/2043317.html
http://blog.csdn.net/puqutogether/article/details/43055633
http://blog.csdn.net/puqutogether/article/details/40889719
regression problem
- m
<<
<script type="math/tex" id="MathJax-Element-68"><<</script>n. the number of samples or instances is greater than the number of features dimension, leading to the X is singular.
- using factor analysis to solve.
- pca dimension
- overfitting (using pca or regularizer to avoid this issue)
- underfitting. the training data set is so small. using the local weighted linear regression.
single variable linear regression
multivariate linear regression
polynomial non-linear regression
logistic regression (non-linear)
http://blog.csdn.net/puqutogether/article/details/43191099
bayesian regression
bayesian estimation—MAP
introduce the prior for the parameters, so that the overfitting problem can be relieved. e.g. l2 norm prior
the prior can also be any extra knowledge, which is also called Occam’s razor.