Selina91
commented
8 years ago Outlier robust (list is not complete):
- tree, random forest, clustering, k-nearest neighbors
- controlling outliers by using the robust Mahalanobis distance based on principal component analysis (PCA)
- robust regularized regression by using a weight through the Mahalanobis distance in principal component space
- By using the Gaussian kernel function in regularized regression
- penalized least trimmed squares criterion
Not robust (list is not complete):
- least square loss function
- regularized regression modeling with a lasso penalty
- mean squared error as error measure in general
http://link.springer.com/chapter/10.1007%2F978-3-642-01082-8_9#page-1 Neural networks and outliers