
Autoencoder-based outlier detection can be improved by incorporating aleatoric uncertainty through Weighted Negative Logarithmic Likelihood (WNLL) and leveraging local relationships through Mean-Shift Scoring (MSS).


coremsg

mss-pae-improving-autoencoder-based-outlier-detection-by-incorporating-aleatoric-uncertainty-and-local-relationships


MSS-PAE: Improving Autoencoder-based Outlier Detection by Incorporating Aleatoric Uncertainty and Local Relationships


title_rewrite


By maximizing the difference between the distributions of training and test data, the proposed DOUST algorithm can achieve supervised-level outlier detection performance without any labeled anomalies.


maximizing-the-difference-between-training-and-test-distributions-for-effective-outlier-detection


Maximizing the Difference Between Training and Test Distributions for Effective Outlier Detection
