MSS-PAE: Saving Autoencoder-based Outlier Detection from Unexpected Reconstruction

arXiv:2304.00709v2 Announce Type: replace
Abstract: AutoEncoders (AEs) are commonly used for machine learning tasks due to their intrinsic learning ability. This unique characteristic can be capitalized for Outlier Detection (OD). However conventional AE-based methods face the issue of overconfident decisions and unexpected reconstruction results of outliers, limiting their performance in OD. To mitigate these issues, the Mean Squared Error (MSE) and Negative Logarithmic Likelihood (NLL) were first analyzed, and the importance of incorporating aleatoric uncertainty to AE-based OD was elucidated. …

abstract arxiv autoencoder autoencoders cs.lg decisions detection face however intrinsic issue machine machine learning outlier outliers performance results saving tasks type